diff --git a/-NFIT4oBgHgl3EQf8ysQ/content/2301.11403v1.pdf b/-NFIT4oBgHgl3EQf8ysQ/content/2301.11403v1.pdf new file mode 100644 index 0000000000000000000000000000000000000000..bbd6be2d9afdf0095220ea96ae2b8197c66c6cb6 --- /dev/null +++ b/-NFIT4oBgHgl3EQf8ysQ/content/2301.11403v1.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1933b3ba87c0903ae49421f0e29b18900f814201af6af1133ed9b22e4b820331 +size 601372 diff --git a/-NFIT4oBgHgl3EQf8ysQ/vector_store/index.pkl b/-NFIT4oBgHgl3EQf8ysQ/vector_store/index.pkl new file mode 100644 index 0000000000000000000000000000000000000000..3ca5dc47d2dffeebe87362f0cc6f9896c979918e --- /dev/null +++ b/-NFIT4oBgHgl3EQf8ysQ/vector_store/index.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5bb315e469dd1352311830db577dd1c47e3368352e1cf7fba7b2d93c92400663 +size 106596 diff --git a/-tE3T4oBgHgl3EQfrgrp/vector_store/index.faiss b/-tE3T4oBgHgl3EQfrgrp/vector_store/index.faiss new file mode 100644 index 0000000000000000000000000000000000000000..53bd1f92deedc5bbda55987150f6b532ca59f24a --- /dev/null +++ b/-tE3T4oBgHgl3EQfrgrp/vector_store/index.faiss @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db416830c04055c1d7813166480058675bc23c227597522185743942062adca8 +size 6619181 diff --git a/-tE4T4oBgHgl3EQf4A1b/content/2301.05310v1.pdf b/-tE4T4oBgHgl3EQf4A1b/content/2301.05310v1.pdf new file mode 100644 index 0000000000000000000000000000000000000000..5aa5afd224a6ba2d5d9e15ca970ea891304a9b4e --- /dev/null +++ b/-tE4T4oBgHgl3EQf4A1b/content/2301.05310v1.pdf @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c86025141869211b3e8cedec1f5a7afe9c84dd79d357c3e9fd82788cc5c9422 +size 637122 diff --git a/-tE4T4oBgHgl3EQf4A1b/vector_store/index.faiss b/-tE4T4oBgHgl3EQf4A1b/vector_store/index.faiss new file mode 100644 index 0000000000000000000000000000000000000000..516b2f2bf9c93cfde824be6b9b354a02c0861dac --- /dev/null +++ b/-tE4T4oBgHgl3EQf4A1b/vector_store/index.faiss @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22c086fe7fccc11f111177055bd2201a9951e99b9e7990f9c068673189659345 +size 2752557 diff --git a/-tE4T4oBgHgl3EQf4A1b/vector_store/index.pkl b/-tE4T4oBgHgl3EQf4A1b/vector_store/index.pkl new file mode 100644 index 0000000000000000000000000000000000000000..868dac65d0ac9fa922dd158f19e89525a9c4c3ba --- /dev/null +++ b/-tE4T4oBgHgl3EQf4A1b/vector_store/index.pkl @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4811f496c6824b559618794dd90bded27cbc824eaefe39997459a1b69029d8b7 +size 121103 diff --git a/.gitattributes b/.gitattributes index 6b84acecefa33b9b08e5104e311ef5a02b9738a7..b0cf18de476ece028092f2a449c4cd1cfa0cb0a8 100644 --- a/.gitattributes +++ b/.gitattributes @@ -6481,3 +6481,87 @@ uNAyT4oBgHgl3EQf0fm1/content/2301.00720v1.pdf filter=lfs diff=lfs merge=lfs -tex 5NE5T4oBgHgl3EQfPA41/content/2301.05501v1.pdf filter=lfs diff=lfs merge=lfs -text QdFJT4oBgHgl3EQf2y0y/content/2301.11657v1.pdf filter=lfs diff=lfs merge=lfs -text QtE0T4oBgHgl3EQfkQGy/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +otE1T4oBgHgl3EQfOwPh/content/2301.03020v1.pdf filter=lfs diff=lfs merge=lfs -text +RdAyT4oBgHgl3EQft_l_/content/2301.00605v1.pdf filter=lfs diff=lfs merge=lfs -text +b9E1T4oBgHgl3EQfKwO_/content/2301.02969v1.pdf filter=lfs diff=lfs merge=lfs -text +UNE2T4oBgHgl3EQfXQdn/content/2301.03842v1.pdf filter=lfs diff=lfs merge=lfs -text +5NE5T4oBgHgl3EQfPA41/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +QNE0T4oBgHgl3EQf1QJC/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +ydE0T4oBgHgl3EQftQH5/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +ANAzT4oBgHgl3EQfTPxG/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +Q9E4T4oBgHgl3EQf_A6p/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +C9E0T4oBgHgl3EQfyQLe/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +d9AzT4oBgHgl3EQfn_2x/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +zdAzT4oBgHgl3EQfefwA/content/2301.01435v1.pdf filter=lfs diff=lfs merge=lfs -text +6dE1T4oBgHgl3EQf7AUK/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +UNE2T4oBgHgl3EQfXQdn/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +MtFOT4oBgHgl3EQf1jQj/content/2301.12939v1.pdf filter=lfs diff=lfs merge=lfs -text +w9AzT4oBgHgl3EQfQftS/content/2301.01199v1.pdf filter=lfs diff=lfs merge=lfs -text +QdFJT4oBgHgl3EQf2y0y/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +BdE5T4oBgHgl3EQfTA_5/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +WtAyT4oBgHgl3EQf8_rY/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +6dE1T4oBgHgl3EQf7AUK/content/2301.03528v1.pdf filter=lfs diff=lfs merge=lfs -text +89AzT4oBgHgl3EQfFPox/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +LNE1T4oBgHgl3EQfGwM1/content/2301.02917v1.pdf filter=lfs diff=lfs merge=lfs -text +ztAyT4oBgHgl3EQfPPYT/content/2301.00019v1.pdf filter=lfs diff=lfs merge=lfs -text +ANAzT4oBgHgl3EQfTPxG/content/2301.01245v1.pdf filter=lfs diff=lfs merge=lfs -text +F9E1T4oBgHgl3EQf-wbm/content/2301.03574v1.pdf filter=lfs diff=lfs merge=lfs -text +udE0T4oBgHgl3EQfsQEM/content/2301.02575v1.pdf filter=lfs diff=lfs merge=lfs -text +5dE3T4oBgHgl3EQfpQo9/content/2301.04640v1.pdf filter=lfs diff=lfs merge=lfs -text +1tE0T4oBgHgl3EQfuQHv/content/2301.02604v1.pdf filter=lfs diff=lfs merge=lfs -text +nNE2T4oBgHgl3EQfJgZo/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +5dE3T4oBgHgl3EQfpQo9/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +-tE4T4oBgHgl3EQf4A1b/content/2301.05310v1.pdf filter=lfs diff=lfs merge=lfs -text +-NFIT4oBgHgl3EQf8ysQ/content/2301.11403v1.pdf filter=lfs diff=lfs merge=lfs -text +DtAyT4oBgHgl3EQf4vpJ/content/2301.00791v1.pdf filter=lfs diff=lfs merge=lfs -text +uNAyT4oBgHgl3EQf0fm1/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +w9AzT4oBgHgl3EQfQftS/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +b9E1T4oBgHgl3EQfKwO_/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +zdAzT4oBgHgl3EQfefwA/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +N9AyT4oBgHgl3EQftPlh/content/2301.00591v1.pdf filter=lfs diff=lfs merge=lfs -text +N9AyT4oBgHgl3EQftPlh/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +otE1T4oBgHgl3EQfOwPh/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +LNE1T4oBgHgl3EQfGwM1/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +-tE3T4oBgHgl3EQfrgrp/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +MtFOT4oBgHgl3EQf1jQj/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +CdE1T4oBgHgl3EQfpwUV/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +RdAyT4oBgHgl3EQft_l_/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +G9E2T4oBgHgl3EQfTge-/content/2301.03804v1.pdf filter=lfs diff=lfs merge=lfs -text +FtE0T4oBgHgl3EQfRAB3/content/2301.02201v1.pdf filter=lfs diff=lfs merge=lfs -text +mdFAT4oBgHgl3EQfcR2T/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +PtAzT4oBgHgl3EQfIvs5/content/2301.01066v1.pdf filter=lfs diff=lfs merge=lfs -text +c9FKT4oBgHgl3EQfqS4B/content/2301.11873v1.pdf filter=lfs diff=lfs merge=lfs -text +F9E1T4oBgHgl3EQf-wbm/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +4tE4T4oBgHgl3EQfBAt_/content/2301.04848v1.pdf filter=lfs diff=lfs merge=lfs -text +ZdE2T4oBgHgl3EQfEgbW/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +h9FAT4oBgHgl3EQf9h7F/content/2301.08757v1.pdf filter=lfs diff=lfs merge=lfs -text +kdFLT4oBgHgl3EQfcy-f/content/2301.12084v1.pdf filter=lfs diff=lfs merge=lfs -text +nNE2T4oBgHgl3EQfJgZo/content/2301.03692v1.pdf filter=lfs diff=lfs merge=lfs -text +i9E2T4oBgHgl3EQfdAdf/content/2301.03901v1.pdf filter=lfs diff=lfs merge=lfs -text +DtAyT4oBgHgl3EQf4vpJ/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +VdFIT4oBgHgl3EQfgyug/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +F9AyT4oBgHgl3EQfSveK/content/2301.00092v1.pdf filter=lfs diff=lfs merge=lfs -text +QNE0T4oBgHgl3EQf1QJC/content/2301.02696v1.pdf filter=lfs diff=lfs merge=lfs -text +CdAzT4oBgHgl3EQfiP0X/content/2301.01495v1.pdf filter=lfs diff=lfs merge=lfs -text +ZdE2T4oBgHgl3EQfEgbW/content/2301.03637v1.pdf filter=lfs diff=lfs merge=lfs -text +y9FKT4oBgHgl3EQfMS25/content/2301.11750v1.pdf filter=lfs diff=lfs merge=lfs -text +-tE4T4oBgHgl3EQf4A1b/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +4tE4T4oBgHgl3EQfBAt_/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +h9FAT4oBgHgl3EQf9h7F/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +z9E1T4oBgHgl3EQfkwQJ/content/2301.03276v1.pdf filter=lfs diff=lfs merge=lfs -text +CdAzT4oBgHgl3EQfiP0X/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +xdAzT4oBgHgl3EQfdvz8/content/2301.01427v1.pdf filter=lfs diff=lfs merge=lfs -text +kdFLT4oBgHgl3EQfcy-f/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +j9E4T4oBgHgl3EQfTQyV/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +mdE2T4oBgHgl3EQfeQfB/content/2301.03915v1.pdf filter=lfs diff=lfs merge=lfs -text +i9E2T4oBgHgl3EQfdAdf/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +U9FJT4oBgHgl3EQfNyw_/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +RtAzT4oBgHgl3EQf0P7C/content/2301.01782v1.pdf filter=lfs diff=lfs merge=lfs -text +G9E2T4oBgHgl3EQfTge-/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +xdAzT4oBgHgl3EQfdvz8/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +o9FAT4oBgHgl3EQfeB1V/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +z9E1T4oBgHgl3EQfkwQJ/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +Z9FQT4oBgHgl3EQffTZu/content/2301.13339v1.pdf filter=lfs diff=lfs merge=lfs -text +g9E0T4oBgHgl3EQfXgDe/content/2301.02295v1.pdf filter=lfs diff=lfs merge=lfs -text +y9FKT4oBgHgl3EQfMS25/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +z9AyT4oBgHgl3EQfPPZM/content/2301.00020v1.pdf filter=lfs diff=lfs merge=lfs -text diff --git a/19FQT4oBgHgl3EQfFDWe/content/tmp_files/2301.13240v1.pdf.txt b/19FQT4oBgHgl3EQfFDWe/content/tmp_files/2301.13240v1.pdf.txt new file mode 100644 index 0000000000000000000000000000000000000000..0771c750731a3dfc3929562da373237c56ee9f54 --- /dev/null +++ b/19FQT4oBgHgl3EQfFDWe/content/tmp_files/2301.13240v1.pdf.txt @@ -0,0 +1,4229 @@ +AdS super gluon scattering up to two loops: +A position space approach +Zhongjie Huanga,b, Bo Wanga,b, Ellis Ye Yuana,b, Xinan Zhouc +aZhejiang Institute of Modern Physics, School of Physics, Zhejiang University, +Hangzhou, Zhejiang 310058, China +bJoint Center for Quanta-to-Cosmos Physics, Zhejiang University, +Hangzhou, Zhejiang 310058, China +cKavli Institute for Theoretical Sciences, University of Chinese Academy of Sciences, +Beijing 100190, China. +E-mail: eyyuan@zju.edu.cn, b w@zju.edu.cn, zjhuang@zju.edu.cn, +xinan.zhou@ucas.ac.cn +Abstract: We carry out a bootstrap study of four-point correlators in 4d N = 2 SCFTs +which are dual to super Yang-Mills on AdS5×S3. We focus on the simplest 1 +2-BPS operators +which correspond to the super gluons in the massless current multiplet. Our computation +is based on an ansatz in position space which is inspired by a hidden symmetry structure +manifest in the leading terms of the Lorentzian singularities of the correlators. By using +other consistency conditions, we completely fix the super gluon correlators at one and two +loops in the bulk genus expansion, up to possible counterterms. Our results reveal a number +of interesting properties enriched by the color structures. In particular, the implication of +hidden conformal symmetry on the full super gluon reduced correlator exhibits an analogous +pattern as in the AdS5 × S5 supergravity correlators recently computed up to two loops. +arXiv:2301.13240v1 [hep-th] 30 Jan 2023 + +Contents +1 +Introduction +2 +2 +Preliminaries +5 +2.1 +Four-point correlators +6 +2.2 +Projectors and color decomposition +7 +2.3 +Spectrum and conformal block decomposition +9 +3 +Leading logarithmic singularities +11 +3.1 +Recursion by unitarity +12 +3.2 +Hidden conformal symmetry +14 +4 +One-loop correlator +16 +4.1 +Ansatz +16 +4.2 +Constraints +20 +4.3 +Results at one loop +21 +4.4 +Comparison with the Mellin space result +25 +5 +Two-loop correlator +26 +5.1 +Color structures at two loops +26 +5.2 +Ansatz and constraints +28 +5.3 +Results at two loops +32 +6 +Outlook +34 +A Single-valued multiple polylogarithms as basis functions +35 +B Analytic result of the one-loop reduced correlator +38 +C Bulk-point limit +40 +D Recursion of twist-4 data at log2 u +43 +– 1 – + +1 +Introduction +The AdS/CFT correspondence maps correlation functions of local operators in the CFT +to on-shell scattering amplitudes in AdS. In the holographic limit, these observables are +expanded in powers of 1/c with respect to the large central charge. At the leading order, the +holographic correlators are just given by the generalized free field theory due to the large N +factorization and they can be computed simply by Wick contractions. However, to extract +nontrivial dynamical information one needs to go to higher orders in 1/c . Computing these +subleading contributions is in general intractable from the CFT side alone as the theory +is strongly coupled. The weakly coupled dual description makes it possible, at least in +principle, as holographic correlators can be computed as amplitudes at various loop orders +by using the AdS generalization of the standard Feynman diagram expansion. However, +it should be noted that such a recipe is rather impractical to use beyond the few simplest +cases [1–5], due to the proliferation of diagrams and complicated AdS vertices [6]. In fact, +just at the tree level, i.e., at order 1/c, the computation of general four-point functions +remained an unsolved problem for almost two decades. +A much better strategy, initiated in [7, 8], is the bootstrap approach, which led to the +complete tree-level four-point functions of 1 +2-BPS operators with arbitrary Kaluza-Klein +(KK) levels for IIB supergravity in AdS5 × S5. The bootstrap approach exploits both the +amplitude intuition from the bulk and the superconformal constraints from the boundary, +and is currently the most efficient method for computing holographic correlators. At the +moment, there is already a wealth of results at tree level. +For example, general four- +point functions of arbitrary 1 +2-BPS operators have been computed in closed forms in all +maximally superconformal theories [9, 10], as well as in theories with half the amount of +maximal superconformal symmetry [11–13].1 By contrast, our understanding for loop level +correlators is much more limited, even in the paradigmatic example of IIB supergravity +on AdS5 × S5. The first one-loop correlator was computed in [15, 16] for the stress tensor +multiplet in position space and later in Mellin space [17]. The calculation was generalized +to four-point functions with higher KK levels in [18–20]. However, explicit one-loop results +are still case-by-case with the exception for the ⟨22pp⟩ family in [20]. At two loops and +higher, the situation is more difficult. The strategy at one loop, which is based on the AdS +unitarity method [21], now requires the additional input of multi-trace operators. Such +information is not yet available in the literature.2 +Therefore, one can in principle only +compute a part of the correlator that corresponds to the iterated s-channel cuts in flat +space [24, 25]. However, it turns out that this difficulty can be overcome at two loops +by formulating an ansatz that is structured by an observed extra hidden symmetry in the +leading Lorentzian singularities, together with additional physical constraints such as the +behavior in the flat-space limit [26]. In this way, the four-point two-loop correlator of stress +tensor multiplets has also been bootstrapped [26, 27]. +1See [14] for a recent review. +2For example, at two loops there are exchange contributions from triple-trace operators. These can be +in principle extracted from tree-level five-point functions. However, only five-point functions of the form +⟨pp222⟩ have been computed [22, 23] while extracting the data requires all ⟨pqr22⟩ five-point functions. +– 2 – + +In this paper, we continue to explore the loop-level calculation of holographic correla- +tors. However, instead of considering correlators of super gravitons, we will focus on super +gluons of SYM in AdS. More precisely, we consider a decoupling sector of certain 4d N = 2 +SCFTs in the holographic limit. These SCFTs can be engineered by using either a stack +of N D3-branes probing F-theory singularities [28, 29] or D3-branes with probe D7-branes +[30]. The near horizon geometries in both cases include an AdS5 ×S3 subspace which hosts +localized degrees of freedom corresponding to the gluons. In the limit of N → ∞, the gluon +degrees of freedom effectively decouple from the graviton degrees of freedom living in the +full 10d bulk via 1/N suppressions in the vertices [13]. The resulting physics in 8d is the +same regardless of the model we choose. Strictly speaking, the decoupling happens only +at the leading order and correlators at subleading orders include gravity contributions as +well. However, in this paper we will choose to turn off gravity to all orders in 1/N and our +goal is to compute the super gluon four-point correlators in this SYM theory in AdS5 × S3 +to two loops. +The motivations for considering super gluon correlators in such a setup are two fold. +First, as we already mentioned, holographic correlators are on-shell scattering amplitudes +in AdS. It is natural to wonder if various remarkable properties of flat-space amplitudes +admit generalizations in curved backgrounds. In particular, does the double copy relation +[31], which famously states gravity is the “square” of YM, still holds in AdS? To this end, +it makes sense to decouple gravity and study the amplitudes of just SYM in AdS. In fact, +analysis of this model at tree level already showed evidence for such a generalization at +four points [32]. Here we will compute the loop corrections of the super gluon four-point +functions which will serve as the starting point for exploring further generalizations of +double copy at higher genus. Second, the super gluon case also provides a useful playground +for acquiring deeper understandings of various results from the supergravity setup. The +position space method for computing loop-level correlators so far have only been tested +in AdS5 × S5 and it is a priori unclear whether it can be applied to other backgrounds. +In this paper, we will show that such a method can be successfully applied to AdS5 × S3 +and leads to similar results to the supergravity case. In the process, we also provide a +nontrivial consistency check of the one-loop result which was previously obtained in [33] +using Mellin space techniques. Moreover, the various different color structures allow us to +have a more refined understanding of the dynamical structures of the correlators which are +similar in the two cases, whereas in the supergravity case all structures are mixed up due +to the absence of colors. +Let us briefly outline our strategy and the key results of the paper. Our approach is +similar to that of [15, 19, 26]. We first make an ansatz in position space which requires a set +of building block functions. Due to the similarity with the supergravity case at tree level, we +assume that single-valued multiple polylogarithms (SVMPLs) continue to be a good basis +for the super gluon correlators at one and two loops. In other words, the correlators are +assumed to be linear combinations of SVMPLs with rational functions of the cross ratios as +coefficients. However, this turns out to be a bit too general. In the supergravity case, the +existence of a tree-level 10d superconformal symmetry [34] highlights a special eighth-order +differential operator ∆(8) which relates the correlators of the top and bottom components +– 3 – + +of the super graviton multiplet. By unitarity this symmetry extends to the leading part +of the Lorentzian singularities at arbitrary loops. Using this operator at loop levels, the +supergravity correlators can be more succinctly written in terms of the pre-correlators L +[19, 26, 27] +H1-loop +sugra = ∆(8)L1-loop +sugra + 1 +4Htree +sugra , +(1.1a) +H2-loop +sugra = +� +∆(8)�2 +L2-loop +sugra + 5 +4H1-loop +sugra − 1 +16Htree +sugra , +(1.1b) +together with additional lower-order correlators. A similar 8d superconformal symmetry +also appears in the tree-level super gluon correlators [13] and the role of ∆(8) is replaced +by a fourth-order operator ∆(4). In analogy with the supergravity case, we assume that +similar pre-correlators can also be defined for super gluons +H1-loop +SYM = ∆(4)L1-loop +SYM + ¯Htree +SYM , +(1.2a) +H2-loop +SYM = +� +∆(4)�2 +L2-loop +SYM + � +H1-loop +SYM + � +Htree +SYM , +(1.2b) +where ¯Htree +SYM and � +Htree +SYM are “tree-like” correlators and � +H1-loop +SYM +is a “one-loop-like” corre- +lator. We will be more precise about the meaning of “tree-like” and “one-loop-like”. But +for the moment it suffices to say they are characterized by the transcendental degrees of +SVMPLs expected at each loop order. Then the position space ansatz in terms of SVMPLs +is formulated in terms of the pre-correlators L and the lower-order objects Hi, in parallel +with the supergravity story. Note that, unlike supergravity, super gluon correlators have +different color structures. Therefore, we make such an ansatz for each independent color +structure and assume the correlator to be a linear combination of all these structures. To +perform the bootstrap, we impose a number of consistency conditions. These are +• Leading logarithmic singularities +• Crossing symmetry +together with a few other constraints. Here the leading logarithmic singularities rely only +on the tree-level data and can be computed at any loop order. At two loops, the additional +constraints further include comparison with the scattering amplitude in a proper flat-space +limit that can be computed independently using flat-space techniques, and the data of +twist-4 operators which can be extracted from the tree and one-loop correlators. Imposing +these constraints, we find that all parameters in the ansatz are fixed except for those +corresponding to the counterterms needed for the UV divergences. Moreover, the tree-like +and one-loop-like terms turn out to be exactly the tree-level and one-loop correlators except +for simple replacements for the color structures. +The rest of the paper is organized as follows. We review in Section 2 some preliminaries +of super gluon four-point functions which include the superconformal kinematics, color +structure and superconformal block decomposition. In Section 3 we review how the leading +logarithmic singularities can be constructed from the tree-level data and compute them in +closed forms using hidden conformal symmetry. In Section 4 we introduce the position +– 4 – + +space method and demonstrate it by bootstrapping the one-loop correlator. In Section 5 +we apply the method to the two-loop correlator and obtain the full answer by imposing +constraints. In Section 6 we outline a few future directions. The paper also has several +appendices where we include further technical details. +In Appendix A we give a brief +review of the properties of SVMPLs. Appendix B contains the complete analytic result +for the reduced correlator at one loop. The details of the flat-space two-loop amplitude +are presented in Appendix C. In Appendix D we discuss the computations related to the +twist-4 data. +2 +Preliminaries +In this paper, we consider holographic correlators corresponding to super gluon scattering +in AdS. To be concrete, we consider SYM in AdS5 ×S3 which arises as a decoupling sector +of certain 4d N = 2 SCFTs. One can construct these SCFTs from a stack of N D3-branes, +by either using them to probe F-theory singularities [28, 29] or by adding a few probe +D7-branes [30]. In either case, in the near horizon limit there is an AdS5 × S3 subspace in +the total ten dimensional spacetime which is locally AdS5 ×S5. On this subspace there are +localized degrees of freedom transforming as an N = 1 vector multiplet and in the adjoint +representation of certain flavor group GF of the boundary CFT. Here GF depends on the +theory and is a gauge group from the bulk perspective.3 Since the N = 1 vector multiplet +contains fields with Lorentz spin at most 1, its KK reduction with respect to S3 also leads +to fields with the same maximal spin. These are the massless and massive AdS gluons +and their super partners. Because of the bound on spins all the KK modes have to reside +in 1 +2-BPS multiplets by the 4d N = 2 representation theory. Their conformal dimensions +are fully fixed by R-symmetry and therefore are independent of the bulk theory. More +precisely, the superconformal primaries of these 1 +2-BPS are scalar operators Ok labelled by +an integer k = 2, 3, . . .. They have conformal dimension ∆ = k and transform in the spin-k +2 +representation of the SU(2)R R-symmetry group. Moreover, they transform in the adjoint +representaiton of the flavor group GF . We will call the fields dual to these superprimaries +the super gluons. The real spinning gluon fields are superconformal descendants in the +multiplets. +By contrast, the super gravitons and their super partners live in the full ten dimen- +sional spacetime. +Unlike the supergluons, their KK spectrum depends on the specific +theory. However, an interesting fact of all these 4d N = 2 SCFTs is that there is a hier- +archy in the couplings at large N. For example, the cubic coupling of three super gluons +(or their superconformal descendants) is of order 1/ +√ +N, while the coupling involving two +super gluons and one super graviton is of order 1/N. Therefore, for large N the lead- +ing contribution to the super gluon correlators comes only from the 8d SYM. Subleading +corrections in 1/N will in general contain graviton contributions as well. +As mentioned in the introduction, we will continue to study loop corrections of the +four-point correlator of the super gluon operator O2 in the AdS5 × S3 SYM. Although this +does not give the full answer for this correlator in N = 2 SCFTs, it makes sense from +3Therefore, in the following we will use “flavor”, “color” and “gauge” interchangeably. +– 5 – + +the perspective of exploring curved space generalizations of gauge theory amplitudes. This +section serves to provide some preliminary features of this correlator, which will be used +in our bootstrap computation. +2.1 +Four-point correlators +With all indices restored, the super gluon operator O2 has the form +OI;a1a2 +2 +(x) , +(2.1) +where I = 1, . . . , dim(GF ) is the flavor symmetry index and ai = 1, 2 are the SU(2)R R- +symmetry indices. It is convenient to contract the R-symmetry indices with two-dimensional +polarization spinors +OI +2(x; v) = OI;a1a2 +2 +va1va2 . +(2.2) +The four-point function +GI1I2I3I4(xi; vi) = ⟨OI1 +2 (x1; v1)OI2 +2 (x2; v2)OI3 +2 (x3; v3)OI4 +2 (x4; v4)⟩ +(2.3) +is therefore a function of both the spacetime coordinates xi and the internal space spinors +vi. Exploiting the bosonic symmetries, i.e. conformal symmetry and R-symmetry, we can +write the correlator as a function of the cross ratios +GI1I2I3I4 = (v1 · v2)2 (v3 · v4)2 +x4 +12x4 +34 +GI1I2I3I4(u, v; α) +(2.4) +where xij = xi − xj, vi · vj = va +i vb +jϵab (ϵab being the 2d Levi–Civita symbol), and the cross +ratios are +u = x2 +12x2 +34 +x2 +13x2 +24 += z¯z , +v = x2 +23x2 +14 +x2 +13x2 +24 += (1 − z)(1 − ¯z) , +α = (v1 · v3) (v2 · v4) +(v1 · v2) (v3 · v4) . +(2.5) +In addition, the ferminonic generators in the superconformal algebra impose further con- +straints known as the superconformal Ward identities [35] +(x∂x − α∂α) GI1I2I3I4(z, ¯z; α) +�� +α=1/x = 0 , +x = z or ¯z. +(2.6) +Solving these identities, we can decompose the correlator into two parts +GI1I2I3I4(z, ¯z; α) = GI1I2I3I4 +0 +(z, ¯z; α) + RHI1I2I3I4(z, ¯z) , +(2.7) +where +R = (1 − zα)(1 − ¯zα) +z¯z +. +(2.8) +Note that our definiton of R is different from that of [13] by a z¯z in the denominator. The +first term GI1I2I3I4 +0 +is protected, while dynamical information of the correlator is encoded in +the reduced correlator HI1I2I3I4. In our case of O2 correlator, HI1I2I3I4 is simply a function +of the spacetime cross ratios {z, ¯z} (or equivalently {u, v}) and is free of the R-symmetry +cross ratio α. +– 6 – + +We study the expansion of the correlator with respect to the large flavor central charge +CJ 4. For convenience, we use the small parameter aF = 6/CJ , with respect to which the +expansion reads +HI1I2I3I4 +2222 +≡ H = H(0) + aF H(1) + a2 +F H(2) + a3 +F H(3) + · · · . +(2.9) +This expansion has a nice interpretation from the bulk point of view. The leading con- +tribution H(0) is associated with the disconnected part of scattering in AdS and can be +evaluated by generalized free field theory. The first correction H(1) is the tree-level scat- +tering of the super gluons, which has been obtained in [13]. The higher-order correction +H(L+1) corresponds to scattering at L loops, where the one-loop case has been computed +in [33] using Mellin space techniques. +2.2 +Projectors and color decomposition +Since we are studying gluon scattering, as usual the correlator H(L+1) at each perturbative +order splits into various color factors and their corresponding dynamical factors 5 +� +H(L+1)�I1I2I3I4 = +� +C +CI1I2I3I4H(L+1) +C +. +(2.10) +A color factor CI1I2I3I4 is constructed out of the structure constants fIJK of the gauge +group according to the topology of a diagram that may arise at the given loop order +according to Feynman rules (or Witten rules in AdS), and so the summation above carries +over all possible topologies at L loops. +The dynamical factors H(L+1) +C +are functions of +kinematic variables {z, ¯z}, and with the above decomposition they only rely on diagram +topologies as well, regardless of any specific choice of the gauge group GF . +The decomposition (2.10) is not the most convenient for practical computations as the +color factors CI1I2I3I4 are highly redundant. So instead one often seeks for other types +of color decompositions. Because our computation requires the input from CFT data of +the spectrum and the coefficients arising in OPEs, it is preferable to decompose the color +factors in a way that resembles the conformal block expansion. This can be fulfilled by +specifying a particular channel (say the s-channel) and introduce an operation P I1I2|I3I4 +a +that picks out irreducible represetation a of the flavor group from the tensor products of +two adjoints adj ⊗ adj. This is called an s-channel projector, and by definition it satisfies +the symmetry properties +P I1I2|I3I4 +a += (−1)|Ra|P I2I1|I3I4 +a +, +P I1I2|I3I4 +a += P I3I4|I1I2 +a +, +(2.11) +where |Ra| stands for the parity of representation a, and the idempotency condition +P I1I2|I5I6 +a +P I6I5|I3I4 +b += δabP I1I2|I3I4 +a +. +(2.12) +4The flavor central charge CJ appears in the flavor current two-point functions as ⟨J I +µ (x)J J +ν (0)⟩ = +CJ +2π2 +δIJ (δµν−2 xµxν +x2 +) +x6 +. Moreover, via supersymmetry, it is related to the three-point function coefficient C2 +2,2,2 +of ⟨O2O2O2⟩ by CJ = 1 +6C2 +2,2,2. +5In the context of scattering amplitudes these coefficients of color factors are more frequently called kine- +matic factors (when referring to numerators in Feynman diagrams). In this paper we call them dynamical +factors to remind the readers that they contains the dynamical information of the theory. +– 7 – + +In particular from (2.12) we also have P I1I2|I3I4 +a +P I1I2|I3I4 +b += δabdim(Ra). Therefore ev- +ery color factor appearing in (2.10) receives a unique decomposition onto the s-channel +projectors +CI1I2I3I4 = +� +a∈adj⊗adj +P I1I2|I3I4 +a +Ca , +(2.13) +with coefficients Ca, or equivalently +Ca = dim(Ra)−1P I1I2|I3I4 +a +CI1I2I3I4 . +(2.14) +The efficiency of these projectors comes from the fact that the set of irreducible rep- +resentations arising in adj ⊗ adj depends only on the gauge group GF but not on the +perturbative order. As a result, the color decomposition of the reduced correlator H as +well as any term H(L+1) in the expansion (2.9) can be carried out in a uniform manner. +Generically, we have +HI1I2I3I4 = +� +a∈adj⊗adj +P I1I2|I3I4 +a +Ha , +(2.15) +and H(L+1) follows similarly. Furthermore, the idempotency condition (2.12) also makes +the recursive relation between different loop levels very simple, as will be further illustrated +in the next section. +As a simple example for the use of projectors, let us quickly review the tree-level +correlator H(1), which was computed in [13]. Its takes the following form +H(1) = csH(1) +s ++ ctH(1) +t ++ cuH(1) +u , +(2.16) +with +H(1) +s += u3 +3 +� +2∂u + (1 + v)∂u∂v + u∂2 +u +� ¯D1111, +(2.17a) +H(1) +t += −u3 +3 +� +2∂v + v∂2 +v + (1 + u)∂u∂v +� ¯D1111, +(2.17b) +H(1) +u += u3 +3 +� +2∂v + v∂2 +v − 2∂u + (u − v)∂u∂v − u∂2 +u +� ¯D1111 . +(2.17c) +Here ¯D1111 is an example of the ¯D-functions which are contact Witten diagrams in AdS 6 +¯D1111(z, ¯z) = +1 +z − ¯z +� +2Li2(z) − 2Li2(¯z) + log(z¯z) log +�1 − z +1 − ¯z +�� +. +(2.18) +cs/t/u are color factors built from structure constants +(cs)I1I2I3I4 = fI1I2JfJ I3I4, +(ct)I1I2I3I4 = fI1I4JfJ I2I3, +(cu)I1I2I3I4 = fI1I3JfJ I4I2, +(2.19) +which are diagrammatically depicted in Figure 1. Note again in this decomposition the +kinematic factors H(1) +s/t/u are independent of the gauge group GF . +When decomposing using the projectors, let us assume that we are working with the +gauge group E8. In this case adj ⊗ adj includes altogether five irreducible representations +6For a review of the precise definition and general properties of ¯D-functions, see Appendix C of [14]. +– 8 – + +I1 +I4 +I3 +I2 +cs = +ct = +I1 +I4 +I3 +I2 +cu = +I1 +I4 +I3 +I2 +Figure 1: Tree color structures cs, ct and cu. +1, 3875, 27000, 248 (adj), and 30380. The former three representations are parity even +and the latter two are paritty odd. +Note that (cs)I1I2I3I4 already represents the exchange of the adjoint representation 248 +in the s-channel, it is therefore proportional to the projector P I1I2|I3I4 +248 +, and we have +(cs)a ≡ P I1I2|I3I4 +248 +(cs)I1I2I3I4 = ψ2h∨(0 +↑ +1 +, +0 +↑ +3875 +, +0 +↑ +27000 +, 1 +↑ +248 +, +0 +↑ +30380 +)T , +(2.20) +where h∨ is the dual Coxeter number, ψ2 is the length squared of the longest root. By +contrast the decomposition of ct, cu involves a mixture of different s-channel projectors +(ct)a = −ψ2h∨ +� +1, 1 +5, − 1 +30, 1 +2, 0 +�T +, +(cu)a = ψ2h∨ +� +1, 1 +5, − 1 +30, −1 +2, 0 +�T +. +(2.21a) +One easily sees that the Jacobi identity (cs)a + (ct)a + (cu)a = 0 is satisfied. Consequently +the coefficients in the projector decomposition of the whole tree-level correlator H(1) are +H(1) +1 +=ψ2h∨ � +−H(1) +t ++ H(1) +u +� +, +(2.22a) +H(1) +3875 =ψ2h∨ +5 +� +−H(1) +t ++ H(1) +u +� +, +(2.22b) +H(1) +27000 = − ψ2h∨ +30 +� +−H(1) +t ++ H(1) +u +� +, +(2.22c) +H(1) +248 =ψ2h∨ +2 +� +2H(1) +s +− H(1) +t +− H(1) +u +� +, +(2.22d) +H(1) +30380 =0. +(2.22e) +Quite remarkably, at this specific level the coefficients of projectors with equal parity are +in fact the same up to some overall constant factors, as was observed in a more general +setup in [13]. +2.3 +Spectrum and conformal block decomposition +As mentioned before our computation partly relies on the existing data of operators ob- +tained from lower loops, so it is helpful to have a quick look at the structure of OPE +and the related block expansion. Thanks to the 4d N = 2 superconformal symmetry, the +correlator GI1I2I3I4 admits a decomposition into superconformal blocks in correspondence +to the exchanges of different superconformal multiplets in the four-point function. The +– 9 – + +relevant sueprmultiplets are listed in Table 1 and a complete classification can be found in +[35]. The OPE of two 1 +2-BPS multipelts B1 contains the following supermultiplets +B1 ⊗ B1 ≃ +2 +� +p=0 +Bp ⊕ +� +ℓ≥0 +� +� +1 +� +p=0 +Cp,( ℓ +2 , ℓ +2) +� +∆ +A∆ +0,( ℓ +2 , ℓ +2) +� +� . +(2.23) +Here Bp and Cp,( ℓ +2 , ℓ +2 ) are protected multiplets and their twists τ = ∆ − ℓ are bounded +from above by the allowed R-symmetry charges. In contrast, there is no upper bound on +the twists of the long multiplets A∆ +0,( ℓ +2 , ℓ +2 ) and their dimensions are not protected. Instead, +they have a lower bound in the holographic limit as they are double-trace (and more +generally multi-trace) operators formed by single-trace operators.7 Let us also note that +the superprimaries of the long multiplets are only allowed to be R-symmetry singlets in +order for the representations of the entire multiplet to fit into the four-point function. The +long multiplets play a key role in the paper as the loop corrections correspond to precisely +the contribution of these multipelts. +Multiplet +Label +SU(2)R +Dimension ∆ and spin ℓ +Half-BPS +BR +R/2 +∆ = 2R, ℓ = 0 +Semi-short +CR,(ℓ/2,ℓ/2) +R/2 +∆ = 2 + 2R + ℓ +Long +A∆ +R,(ℓ/2,ℓ/2) +R/2 +∆ ≥ 2 + 2R + ℓ +Table 1: Supermultiplets that appear in the fusion rules of two B’s for N = 2 SCFTs. +We will focus on the reduced correlator H which has already taken superconformal +symmetry into account. In this way the superconformal block decomposition simply reduces +to just the ordinary conformal block decomposition. As superconformal symmetry and +gauge symmetry commute, this directly passes through the color projector decomposition, +and in terms of each component in (2.15) this reads [35] +Ha(z, ¯z) = +� +τa,ℓ +aagτa+2,ℓ(z, ¯z) , +(2.24) +where τa and ℓ sum over the spectrum of the supermultiplets. Note that the shift in τ by +2 in the ordinary conformal block gτ+2,ℓ is a consequence of the superconformal symmetry. +The detailed expression of these blocks is [36] +gτ,ℓ = +z¯z +¯z − z +� +k τ−2 +2 (z)k τ+2ℓ +2 (¯z) − k τ−2 +2 (¯z)k τ+2ℓ +2 (z) +� +, +kh(z) = zh 2F1(h, h, 2h, z). +(2.25) +Since the long multiplets are not protected, in the limit of N → ∞ their twists as well +as OPE coefficients receive perturbative corrections with respect to small aF +τa =τ0 + aF γ(1) +a ++ a2 +F γ(2) +a ++ . . . , +(2.26a) +aa(τ, ℓ) =a(0) +a ++ aF a(1) +a ++ a2 +F a(2) +a ++ . . . . +(2.26b) +7This bound is stronger than the unitarity bound in Table 1. +– 10 – + +Substituting the above expansion into (2.24) gives the following series expansion for Ha +Ha = +� +τ0,ℓ +a(0) +a gτ0+2,ℓ(z, ¯z) +� +�� +� +H(0) +a ++aF +� +τ0,ℓ +� +a(0) +a γ(1) +a ∂τ0 + a(1) +a +� +gτ0+2,ℓ(z, ¯z) +� +�� +� +H(1) +a ++ a2 +F +� +τ0,ℓ +�1 +2a(0) +a (γ(1) +a )2∂2 +τ0 + (a(1) +a γ(1) +a ++ a(0) +a γ(2) +a )∂τ0 + a(2) +a +� +gτ0+2,ℓ(z, ¯z) +� +�� +� +H(2) +a ++ . . . . +(2.27) +The first term H(0) +a +receives contributions only from long operators whose a(0) +a +are non- +vanishing. From large N factorization, H(0) +a +is given by the disconnected correlator and +these contributing operators can only be double-trace operators. However, these operators +are degenerate at the classical level. For instance, among the double-trace operators +: O2□n−2∂ℓO2 : , : O3□n−3∂ℓO3 : , . . . , : On∂ℓOn : +(2.28) +all have classical twist τ (0) = 2n and spin ℓ. Consequently, each term in (2.27) should not +be literally understood as the contribution from a single operator, but rather in an averaged +sense. Moreover, at higher orders in aF there are also higher-trace operators appearing in +the OPE 8, which can have the same twist as the double-trace operators and will enter +the mixing as well. +Therefore, in a precise description it is necessary to use an extra +label i to distinguish different operators in the degeneracy. Then the coefficient a(0) +a γ(1) +a +should in fact be understood as ⟨a(0) +a γ(1) +a ⟩ ≡ � +i a(0) +i,aγ(1) +i,a , and a(0) +a +� +γ(1) +a +�2 +as ⟨a(0) +a +� +γ(1) +a +�2 +⟩ ≡ +� +i a(0) +i,a +� +γ(1) +i,a +�2 +, and so on. +3 +Leading logarithmic singularities +As an analytic function of the kinematic variables z and ¯z, a conformal correlator can in +principle be constructed out of its singularities by dispersion-type relations, in a similar +way as the dispersion relation that generates a four-point scattering amplitude from its +physical channel discontinuities. For generic CFTs such relations were formulated in [37]. +This means that the defining data for a correlator is necessarily encoded in its singularities. +While our computation does not rely on the dispersion relations, these data still provide a +vital input in determining the loop-level corrections to the reduced correlator H. +When viewed in the perturbative expansion (2.27) these singularities are sourced at +small u by the log(u) factors arising from the derivatives acting on the conformal block. +Recall in the definition (2.27) that gτ,ℓ(z, ¯z) ∝ uτ/2, so at each order ap +F the reduced +correlator can be organized in terms of powers of log(u) +H(p) +a (z, ¯z) = +1 +2pp! logp(u) +� +τ0,ℓ +⟨a(0) +a (γ(1) +a )p⟩ gτ0,ℓ(z, ¯z) + +� +terms with logk
0, the final result for the +planar double box integral Ipdb +1234 in 8 − 2ϵ dimensions reads +Ipdb +1234 =(−s)1−2ϵ +� +1 +144 +1 +ϵ2 + 79 + 2x +1728 +1 +ϵ + −48 + 3303x + 242x2 +20736x ++ (1 + x)(2 − 15x + x2) +432x2 +(G0 − G1) + (1 − 6x + 34x2 + 67x3)π2 +2592(x − 1)x3 +− 1 − 5x + 25x2 + 3x3 +216x3 +� +G0,0 − G0,1 − G1,0 + G1,1 + 5π2 +12 +� +− −1 + 8x + 17x2 +108(x − 1)x2 +� +G0,0,0 − G0,0,1 − G0,1,0 + G0,1,1 + π2 +6 (2G0 + G1) − ζ3 +� +− +3 + x +18(x − 1)2 +� +G0,0,0,0 − G0,0,0,1 − G0,0,1,0 + G0,0,1,1 + π2 +6 (2G0,0 + G0,1) − ζ3G0 + 17π4 +360 +�� +, +(C.6) +where we use the abbreviation G⃗a ≡ G⃗a(x−1), and the dimensionless parameter x is related +to the Mandelstam variables s, t or the scattering angle θ by +x ≡ 1 + t +s = 1 + cos θ +2 +. +(C.7) +15The authors are grateful to Lilin Yang for sharing data for these integrals in 4−2ϵ dimensions, together +with a set of master integrals needed for dimensional recursion which were selected following [66, 67]. The +dimensional recursion is performed using the Mathematica package LiteRed [68]. +– 41 – + +In the same region the result for the non-planar double box integral Indb +1234 reads +Indb +1234 =(−s)1−2ϵ +� +1 +288 +1 +ϵ2 + +37 +1728 +1 +ϵ + −60 − 3607x + 3625x2 − 36x3 + 18x4 +51840(x − 1)x ++ (−12 + 42x − 47x2 + 2x3 − 63x4 + 18x5) +25920(x − 1)2 +(G1 + iπ) +− 60 − 328x + 727x2 − 798x3 + 399x4 +25920(x − 1)2x2 +(G0 − G1) + +π2 +1728 +− 30 − 173x + 428x2 − 600x3 + 525x4 − 270x5 + 90x6 +12960(x − 1)3x3 +(G0,0 + iπG0) ++ 30 − 113x + 160x2 − 107x3 + 36x4 + 6x5 +12960(x − 1)x3 +� +G0,1 + G1,0 + iπ(G0 + G1) − π2 +2 +� ++ (x − 1)3(−30 − 7x + 4x2 + 9x3 + 9x4) +12960x3 +(G1,1 + iπG1) − (3x − 2)(3x − 1)ζ3 +720(x − 1)2x2 ++ (2x − 1)(2 − 5x + 5x2) +2160(−1 + x)2x2 +� +G0,0,1 + G0,1,0 − 2G1,0,0 + iπ(G0,0 + G0,1 − 2G1,0) − π2 +2 G0 + iπ3 +2 +� ++ (x − 1)3(2 + x) +2160x2 +� +G1,0,1 + G1,1,0 − 2G0,1,1 + iπ(G1,0 + G1,1 − 2G0,1) − π2 +2 G1 +�� +. +(C.8) +The remaining integrals in (C.3) can be obtained from (C.6) and (C.8) by permuting the +particle labels. For readers’ convenience we also record these results in the ancillary file. +Note that the expressions from permutations may live in different physical regions, and +so before assembling them together one needs to analytically continued the Mandelstam +variables to the same region following the standard iε prescription. For our computation +we finally lands on the region s, t < 0 and u > 0, in which (C.6) and (C.8) directly apply. +In order to perform the comparison in the bulk-point limit we need to analytically +continue our ansatz for the correlator from Euclidean region to physical region as well. +Following [27], our prescription is to continue z counter-clockwisely around 0 and ¯z clock- +wisely around 1, and then set z = ¯z + 2ω¯z√1 − ¯z. The bulk-point limit z → ¯z is then +reached by setting ω → 0. As was already pointed out in Section 5.2, when taking this +limit at two loops [∆(4)]2L(3) dominates over the modification terms, and the action of ∆(4) +reduces to a simple multiplication factor. Therefore it suffices to compare A2-loop with the +pre-correlator L(3). The detailed connection between these two objects is +lim +ω→0 (z − ¯z)5L(3) � +z⟲0, ¯z⟲1���� +z=¯z+2ω¯z√1−¯z = 48π2(1 − ¯z)2¯z6s−2 ˜ +A2-loop(x) +��� +x=1/¯z , +(C.9) +Note the above relation is already expressed in terms of the reduced amplitude ˜ +A2-loop. +Although the full amplitude A2-loop contains an extra factor stAtree +1234 including polarization +vectors, in practical computation we do not have to bother manipulating it. The reason is +that comparison with the leading log data in (5.15) already fully determines contributions +of some MPLs of highest transcendental weight, and matching them with the corresponding +MPL contributions in ˜ +A2-loop easily determines the factors appearing in the above relation +– 42 – + +(C.9) (e.g., matching coefficients of G0,0,0,0 or G1,1,1,1 in the limit). Comparison between +the remaining contributions then generates constraints for the undetermined variables in +the ansatz. +[46] proposed a simpler connection than (C.9), the relation between discontinuities +dDisc H and Disc A, in the context of graviton scattering. Similar connection should also +apply to the gluon scattering studied here. Ideally one would not expect much difference +between the comparison at the level of the full correlator and that of the discontinuity, +because in principle the correlator H can be reconstructed from its double discontinuity +dDisc H through Lorentz inversion [47], which is the CFT counterpart of the dispersion +relation relating A and its discontinuity Disc A in flat space. However, at loop level these +dispersion relations can be polluted by the presence of finite spin contributions to the cor- +relator/amplitude, which imposes extra data in addition to the discontinuities. Therefore +one expects that constraints from (C.9) are stronger. +D +Recursion of twist-4 data at log2 u +The reduced correlator H can be organized in terms of power expansions in log u in the small +u limit, where the power of log u goes up to n + 1 at n loops. The coefficient functions +of these log u powers encode different combinations of the expansion coefficients of the +CFT data with respect to aF , which are schematically listed in Table 2. The goal of this +appendix is to compute the combination ⟨a(1) +a (γ(1) +a )2 + 2a(0) +a γ(1) +a γ(2) +a ⟩ for twist-4 operators. +This data contributes to the two-loop correlator in the log2 u coefficient, as explicitly shown +in (5.22), and we use them as one of the inputs in our bootstrap algorithm. As is clear from +the expression, only tree-level and one-loop correlators are needed. Moreover, the twist-4 +operators are free of operator mixing. This fact makes it possible to extract their CFT +data from just the O2 correlators alone. The angle brackets ⟨. . .⟩ will also be dropped as +they are no longer necessary in this case. +order +a1 +F +a2 +F +a3 +F +log(u)0 +⟨a(1) +a ⟩ +⟨a(2) +a ⟩ +⟨a(3) +a ⟩ +log(u)1 +⟨a(0) +a γ(1) +a ⟩ +⟨a(1) +a γ(1) +a ++ a(0) +a γ(2)⟩ +⟨a(2) +a γ(1) +a ++ a(1) +a γ(2) +a ++ a(0) +a γ(3) +a ⟩ +log(u)2 +⟨a(0) +a (γ(1) +a )2⟩ +⟨a(1)(γ(1) +a )2 + 2a(0) +a γ(1) +a γ(2) +a ⟩ +log(u)3 +⟨a(0) +a (γ(1) +a )3⟩ +Table 2: The OPE data encoded in the coefficient functions of different log u powers for +correlators up to two loops. +The most efficient way to extract the CFT data from explicit correlators is to use the +Lorentzian inversion formula [47]. Applying it to the tree-level correlator H(1), we obtain +– 43 – + +the following data for the twist-4 operator with spin ℓ +a(0) +a γ(1) +a +=(−ct + (−1)ℓcu)a +2Γ(ℓ + 3)2 +Γ(2ℓ + 5) , +(D.1) +a(1) +a +=(−ct + (−1)ℓcu)a +Γ(ℓ + 3)2 +Γ(2ℓ + 5) +� +2ψ(0)(ℓ + 3)−2ψ(0)(2ℓ + 5)+ 2 +3 + (−1)ℓ +6 +� +. +(D.2) +Here for the color part (#)a is defined by projectors as in (2.20). +Similarly, from the +one-loop correlator H(2), we can extract the following combinations +a(0) +a (γ(1) +a )2 =(dst + (−1)ℓdsu)a +Γ(ℓ + 3)2 +Γ(2ℓ + 5) +8 +(ℓ + 1)(ℓ + 4) , +(D.3) +a(1) +a γ(1) +a ++ a(0) +a γ(2) +a +=(dst + (−1)ℓdsu)a +Γ(ℓ + 3)2 +Γ(2ℓ + 5) +� +−4 +� +2ℓ3 + 23ℓ2 + 65ℓ + 32 +� +ℓ(ℓ + 1)2(ℓ + 4)2(ℓ + 5) ++ 2 +� +12ψ(0)(ℓ + 3) − 12ψ(0)(2ℓ + 5) + 11 +� +3(ℓ + 1)(ℓ + 4) +� +− (1 + (−1)ℓ)(dtu)a +Γ(ℓ + 3)2 +Γ(2ℓ + 5) +2 +� +5ℓ2 + 25ℓ + 24 +� +3ℓ(ℓ + 1)(ℓ + 4)(ℓ + 5) . +(D.4) +By comparing (D.1) and (D.3), we get +γ(1) +a += (−ct + (−1)ℓcu)a +2 +(ℓ + 1)(ℓ + 4) . +(D.5) +We could also solve for a(0) +a , a(1) +a , γ(2) +a . However, to compute the wanted data it is sufficient +to consider the following combination +a(1) +a (γ(1) +a )2 + 2a(0) +a γ(1) +a γ(2) +a += 2 +� +a(1) +a γ(1) +a ++ a(0) +a γ(2) +a +� � +γ(1) +a +� +− +� +a(1) +a +� � +γ(1) +a +�2 +, +(D.6) +and get +a(1) +a (γ(1) +a )2 + 2a(0) +a γ(1) +a γ(2) +a += − (1 + (−1)ℓ)(fs)a +Γ(ℓ + 3)2 +Γ(2ℓ + 5) +16 +� +5ℓ2 + 25ℓ + 24 +� +3ℓ(ℓ + 1)2(ℓ + 4)2(ℓ + 5) ++ (es1 + (−1)ℓes2)a +Γ(ℓ + 3)2 +Γ(2ℓ + 5) +� +−32 +� +2ℓ3 + 23ℓ2 + 65ℓ + 32 +� +ℓ(ℓ + 1)3(ℓ + 4)3(ℓ + 5) ++ 8 +� +12ψ(0)(ℓ + 3) − 12ψ(0)(2ℓ + 5) + 18 − (−1)ℓ� +3(ℓ + 1)2(ℓ + 4)2 +� +. +(D.7) +Note this result should only be trusted down to ℓ = 2. This is because it uses the one-loop +data. The presence of the contact counterterm at one loop spoils the analyticity in spin at +ℓ = 0. +– 44 – + +References +[1] D. Z. Freedman, S. D. Mathur, A. Matusis, and L. Rastelli, “Correlation functions in the +CFT(d) / AdS(d+1) correspondence,” Nucl. Phys. B546 (1999) 96–118, +arXiv:hep-th/9804058 [hep-th]. +[2] E. D’Hoker, D. Z. Freedman, S. D. Mathur, A. Matusis, and L. Rastelli, “Graviton exchange +and complete four point functions in the AdS / CFT correspondence,” Nucl. Phys. B562 +(1999) 353–394, arXiv:hep-th/9903196 [hep-th]. +[3] G. Arutyunov and S. Frolov, “Four point functions of lowest weight CPOs in N=4 SYM(4) +in supergravity approximation,” Phys. Rev. D62 (2000) 064016, arXiv:hep-th/0002170 +[hep-th]. +[4] G. Arutyunov, F. Dolan, H. Osborn, and E. Sokatchev, “Correlation functions and massive +Kaluza-Klein modes in the AdS / CFT correspondence,” Nucl. Phys. B 665 (2003) 273–324, +arXiv:hep-th/0212116. +[5] G. Arutyunov and E. Sokatchev, “On a large N degeneracy in N=4 SYM and the AdS / CFT +correspondence,” Nucl. Phys. B663 (2003) 163–196, arXiv:hep-th/0301058 [hep-th]. +[6] G. Arutyunov and S. Frolov, “Scalar quartic couplings in type IIB supergravity on AdS(5) x +S**5,” Nucl. Phys. B579 (2000) 117–176, arXiv:hep-th/9912210 [hep-th]. +[7] L. Rastelli and X. Zhou, “Mellin amplitudes for AdS5 × S5,” Phys. Rev. Lett. 118 no. 9, +(2017) 091602, arXiv:1608.06624 [hep-th]. +[8] L. Rastelli and X. Zhou, “How to Succeed at Holographic Correlators Without Really +Trying,” JHEP 04 (2018) 014, arXiv:1710.05923 [hep-th]. +[9] L. F. Alday and X. Zhou, “All Tree-Level Correlators for M-theory on AdS7 × S4,” Phys. +Rev. Lett. 125 no. 13, (2020) 131604, arXiv:2006.06653 [hep-th]. +[10] L. F. Alday and X. Zhou, “All Holographic Four-Point Functions in All Maximally +Supersymmetric CFTs,” Phys. Rev. X 11 no. 1, (2021) 011056, arXiv:2006.12505 +[hep-th]. +[11] L. Rastelli, K. Roumpedakis, and X. Zhou, “AdS3 × S3 Tree-Level Correlators: Hidden +Six-Dimensional Conformal Symmetry,” JHEP 10 (2019) 140, arXiv:1905.11983 [hep-th]. +[12] S. Giusto, R. Russo, A. Tyukov, and C. Wen, “The CFT6 origin of all tree-level 4-point +correlators in AdS3 × S3,” Eur. Phys. J. C 80 no. 8, (2020) 736, arXiv:2005.08560 +[hep-th]. +[13] L. F. Alday, C. Behan, P. Ferrero, and X. Zhou, “Gluon Scattering in AdS from CFT,” +JHEP 06 (2021) 020, arXiv:2103.15830 [hep-th]. +[14] A. Bissi, A. Sinha, and X. Zhou, “Selected topics in analytic conformal bootstrap: A guided +journey,” Phys. Rept. 991 (2022) 1–89, arXiv:2202.08475 [hep-th]. +[15] F. Aprile, J. M. Drummond, P. Heslop, and H. Paul, “Quantum Gravity from Conformal +Field Theory,” JHEP 01 (2018) 035, arXiv:1706.02822 [hep-th]. +[16] L. F. Alday and A. Bissi, “Loop Corrections to Supergravity on AdS5 × S5,” Phys. Rev. +Lett. 119 no. 17, (2017) 171601, arXiv:1706.02388 [hep-th]. +[17] L. F. Alday, “On genus-one string amplitudes on AdS5 × S5,” JHEP 04 (2021) 005, +arXiv:1812.11783 [hep-th]. +[18] F. Aprile, J. M. Drummond, P. Heslop, and H. Paul, “Loop corrections for Kaluza-Klein +– 45 – + +AdS amplitudes,” JHEP 05 (2018) 056, arXiv:1711.03903 [hep-th]. +[19] F. Aprile, J. Drummond, P. Heslop, and H. Paul, “One-loop amplitudes in AdS5 × S5 +supergravity from N = 4 SYM at strong coupling,” JHEP 03 (2020) 190, +arXiv:1912.01047 [hep-th]. +[20] L. F. Alday and X. Zhou, “Simplicity of AdS Supergravity at One Loop,” JHEP 09 (2020) +008, arXiv:1912.02663 [hep-th]. +[21] O. Aharony, L. F. Alday, A. Bissi, and E. Perlmutter, “Loops in AdS from Conformal Field +Theory,” JHEP 07 (2017) 036, arXiv:1612.03891 [hep-th]. +[22] V. Gon¸calves, R. Pereira, and X. Zhou, “20′ Five-Point Function from AdS5 × S5 +Supergravity,” JHEP 10 (2019) 247, arXiv:1906.05305 [hep-th]. +[23] V. Gon¸calves, C. Meneghelli, R. Pereira, J. V. Boas, and X. Zhou, “To appear,”. +[24] A. Bissi, G. Fardelli, and A. Georgoudis, “Towards All Loop Supergravity Amplitudes on +AdS5 × S5,” arXiv:2002.04604 [hep-th]. +[25] A. Bissi, G. Fardelli, and A. Georgoudis, “All loop structures in Supergravity Amplitudes on +AdS5 × S5 from CFT,” arXiv:2010.12557 [hep-th]. +[26] Z. Huang and E. Y. Yuan, “Graviton Scattering in AdS5 × S5 at Two Loops,” +arXiv:2112.15174 [hep-th]. +[27] J. M. Drummond and H. Paul, “Two-loop supergravity on AdS5 × S5 from CFT,” JHEP 08 +(2022) 275, arXiv:2204.01829 [hep-th]. +[28] A. Fayyazuddin and M. Spalinski, “Large N superconformal gauge theories and supergravity +orientifolds,” Nucl. Phys. B 535 (1998) 219–232, arXiv:hep-th/9805096. +[29] O. Aharony, A. Fayyazuddin, and J. M. Maldacena, “The Large N limit of N=2, N=1 field +theories from three-branes in F theory,” JHEP 07 (1998) 013, arXiv:hep-th/9806159. +[30] A. Karch and E. Katz, “Adding flavor to AdS / CFT,” JHEP 06 (2002) 043, +arXiv:hep-th/0205236. +[31] Z. Bern, J. J. M. Carrasco, and H. Johansson, “Perturbative Quantum Gravity as a Double +Copy of Gauge Theory,” Phys. Rev. Lett. 105 (2010) 061602, arXiv:1004.0476 [hep-th]. +[32] X. Zhou, “Double Copy Relation in AdS Space,” Phys. Rev. Lett. 127 no. 14, (2021) 141601, +arXiv:2106.07651 [hep-th]. +[33] L. F. Alday, A. Bissi, and X. Zhou, “One-loop gluon amplitudes in AdS,” JHEP 02 (2022) +105, arXiv:2110.09861 [hep-th]. +[34] S. Caron-Huot and A.-K. Trinh, “All Tree-Level Correlators in AdS5×S5 Supergravity: +Hidden Ten-Dimensional Conformal Symmetry,” JHEP 01 (2019) 196, arXiv:1809.09173 +[hep-th]. +[35] M. Nirschl and H. Osborn, “Superconformal Ward identities and their solution,” Nucl. Phys. +B 711 (2005) 409–479, arXiv:hep-th/0407060. +[36] F. A. Dolan and H. Osborn, “Conformal four point functions and the operator product +expansion,” Nucl. Phys. B 599 (2001) 459–496, arXiv:hep-th/0011040. +[37] D. Carmi and S. Caron-Huot, “A Conformal Dispersion Relation: Correlations from +Absorption,” JHEP 09 (2020) 009, arXiv:1910.12123 [hep-th]. +[38] F. Aprile, J. M. Drummond, P. Heslop, and H. Paul, “Unmixing Supergravity,” JHEP 02 +– 46 – + +(2018) 133, arXiv:1706.08456 [hep-th]. +[39] R. E. Cutkosky, “Singularities and discontinuities of Feynman amplitudes,” J. Math. Phys. 1 +(1960) 429–433. +[40] Z. Bern, L. J. Dixon, D. C. Dunbar, and D. A. Kosower, “One loop n point gauge theory +amplitudes, unitarity and collinear limits,” Nucl. Phys. B 425 (1994) 217–260, +arXiv:hep-ph/9403226. +[41] Z. Bern, L. J. Dixon, D. C. Dunbar, and D. A. Kosower, “Fusing gauge theory tree +amplitudes into loop amplitudes,” Nucl. Phys. B 435 (1995) 59–101, +arXiv:hep-ph/9409265. +[42] J. M. Drummond, R. Glew, and M. Santagata, “BCJ relations in AdS5 × S3 and the +double-trace spectrum of super gluons,” arXiv:2202.09837 [hep-th]. +[43] F. A. Dolan and H. Osborn, “Conformal Partial Waves: Further Mathematical Results,” +arXiv:1108.6194 [hep-th]. +[44] C.-M. Chang and Y.-H. Lin, “Carving Out the End of the World or (Superconformal +Bootstrap in Six Dimensions),” JHEP 08 (2017) 128, arXiv:1705.05392 [hep-th]. +[45] J. Maldacena, D. Simmons-Duffin, and A. Zhiboedov, “Looking for a bulk point,” JHEP 01 +(2017) 013, arXiv:1509.03612 [hep-th]. +[46] L. F. Alday and S. Caron-Huot, “Gravitational S-matrix from CFT dispersion relations,” +JHEP 12 (2018) 017, arXiv:1711.02031 [hep-th]. +[47] S. Caron-Huot, “Analyticity in Spin in Conformal Theories,” JHEP 09 (2017) 078, +arXiv:1703.00278 [hep-th]. +[48] N. Ceplak, S. Giusto, M. R. R. Hughes, and R. Russo, “Holographic correlators with +multi-particle states,” JHEP 09 (2021) 204, arXiv:2105.04670 [hep-th]. +[49] W.-J. Ma and X. Zhou, “Scattering bound states in AdS,” JHEP 08 (2022) 107, +arXiv:2204.13419 [hep-th]. +[50] A. B. Goncharov, “Multiple polylogarithms, cyclotomy and modular complexes,” Math. Res. +Lett. 5 (1998) 497–516, arXiv:1105.2076 [math.AG]. +[51] A. B. Goncharov, “Multiple polylogarithms and mixed Tate motives,” arXiv:math/0103059. +[52] J. M. Drummond and H. Paul, “One-loop string corrections to AdS amplitudes from CFT,” +JHEP 03 (2021) 038, arXiv:1912.07632 [hep-th]. +[53] C. Duhr and F. Dulat, “PolyLogTools — polylogs for the masses,” JHEP 08 (2019) 135, +arXiv:1904.07279 [hep-th]. +[54] F. Chavez and C. Duhr, “Three-mass triangle integrals and single-valued polylogarithms,” +JHEP 11 (2012) 114, arXiv:1209.2722 [hep-ph]. +[55] F. Brown, “Single-valued multiple polylogarithms in one variable,” C.R.Acad.Sci.Paris I338 +(2004) 527. +[56] L. J. Dixon, C. Duhr, and J. Pennington, “Single-valued harmonic polylogarithms and the +multi-Regge limit,” JHEP 10 (2012) 074, arXiv:1207.0186 [hep-th]. +[57] Z. Bern, J. S. Rozowsky, and B. Yan, “Two loop four gluon amplitudes in N=4 +superYang-Mills,” Phys. Lett. B 401 (1997) 273–282, arXiv:hep-ph/9702424. +[58] A. V. Kotikov, “Differential equations method: New technique for massive Feynman +– 47 – + +diagrams calculation,” Phys. Lett. B 254 (1991) 158–164. +[59] A. V. Kotikov, “Differential equation method: The Calculation of N point Feynman +diagrams,” Phys. Lett. B 267 (1991) 123–127. [Erratum: Phys.Lett.B 295, 409–409 (1992)]. +[60] E. Remiddi, “Differential equations for Feynman graph amplitudes,” Nuovo Cim. A 110 +(1997) 1435–1452, arXiv:hep-th/9711188. +[61] J. M. Henn, “Multiloop integrals in dimensional regularization made simple,” Phys. Rev. +Lett. 110 (2013) 251601, arXiv:1304.1806 [hep-th]. +[62] C. G. Papadopoulos, “Simplified differential equations approach for Master Integrals,” JHEP +07 (2014) 088, arXiv:1401.6057 [hep-ph]. +[63] O. V. Tarasov, “Connection between Feynman integrals having different values of the +space-time dimension,” Phys. Rev. D 54 (1996) 6479–6490, arXiv:hep-th/9606018. +[64] R. N. Lee, “Space-time dimensionality D as complex variable: Calculating loop integrals +using dimensional recurrence relation and analytical properties with respect to D,” Nucl. +Phys. B 830 (2010) 474–492, arXiv:0911.0252 [hep-ph]. +[65] R. N. Lee, “Calculating multiloop integrals using dimensional recurrence relation and +D-analyticity,” Nucl. Phys. B Proc. Suppl. 205-206 (2010) 135–140, arXiv:1007.2256 +[hep-ph]. +[66] J. Chen, X. Jiang, X. Xu, and L. L. Yang, “Constructing canonical Feynman integrals with +intersection theory,” Phys. Lett. B 814 (2021) 136085, arXiv:2008.03045 [hep-th]. +[67] J. Chen, X. Jiang, C. Ma, X. Xu, and L. L. Yang, “Baikov representations, intersection +theory, and canonical Feynman integrals,” JHEP 07 (2022) 066, arXiv:2202.08127 +[hep-th]. +[68] R. N. Lee, “LiteRed 1.4: a powerful tool for reduction of multiloop integrals,” J. Phys. Conf. +Ser. 523 (2014) 012059, arXiv:1310.1145 [hep-ph]. +– 48 – + diff --git a/19FQT4oBgHgl3EQfFDWe/content/tmp_files/load_file.txt b/19FQT4oBgHgl3EQfFDWe/content/tmp_files/load_file.txt new file mode 100644 index 0000000000000000000000000000000000000000..cad19bf6a7c55d8a524771f86aba8731743cd53f --- /dev/null +++ b/19FQT4oBgHgl3EQfFDWe/content/tmp_files/load_file.txt @@ -0,0 +1,1692 @@ +filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf,len=1691 +page_content='AdS super gluon scattering up to two loops: A position space approach Zhongjie Huanga,b, Bo Wanga,b, Ellis Ye Yuana,b, Xinan Zhouc aZhejiang Institute of Modern Physics, School of Physics, Zhejiang University, Hangzhou, Zhejiang 310058, China bJoint Center for Quanta-to-Cosmos Physics, Zhejiang University, Hangzhou, Zhejiang 310058, China cKavli Institute for Theoretical Sciences, University of Chinese Academy of Sciences, Beijing 100190, China.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' E-mail: eyyuan@zju.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='cn, b w@zju.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='cn, zjhuang@zju.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='cn, xinan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='zhou@ucas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='cn Abstract: We carry out a bootstrap study of four-point correlators in 4d N = 2 SCFTs which are dual to super Yang-Mills on AdS5×S3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' We focus on the simplest 1 2-BPS operators which correspond to the super gluons in the massless current multiplet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Our computation is based on an ansatz in position space which is inspired by a hidden symmetry structure manifest in the leading terms of the Lorentzian singularities of the correlators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' By using other consistency conditions, we completely fix the super gluon correlators at one and two loops in the bulk genus expansion, up to possible counterterms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Our results reveal a number of interesting properties enriched by the color structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In particular, the implication of hidden conformal symmetry on the full super gluon reduced correlator exhibits an analogous pattern as in the AdS5 × S5 supergravity correlators recently computed up to two loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='13240v1 [hep-th] 30 Jan 2023 Contents 1 Introduction 2 2 Preliminaries 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1 Four-point correlators 6 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2 Projectors and color decomposition 7 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='3 Spectrum and conformal block decomposition 9 3 Leading logarithmic singularities 11 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1 Recursion by unitarity 12 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2 Hidden conformal symmetry 14 4 One-loop correlator 16 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1 Ansatz 16 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2 Constraints 20 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='3 Results at one loop 21 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='4 Comparison with the Mellin space result 25 5 Two-loop correlator 26 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1 Color structures at two loops 26 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2 Ansatz and constraints 28 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='3 Results at two loops 32 6 Outlook 34 A Single-valued multiple polylogarithms as basis functions 35 B Analytic result of the one-loop reduced correlator 38 C Bulk-point limit 40 D Recursion of twist-4 data at log2 u 43 – 1 – 1 Introduction The AdS/CFT correspondence maps correlation functions of local operators in the CFT to on-shell scattering amplitudes in AdS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In the holographic limit, these observables are expanded in powers of 1/c with respect to the large central charge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' At the leading order, the holographic correlators are just given by the generalized free field theory due to the large N factorization and they can be computed simply by Wick contractions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, to extract nontrivial dynamical information one needs to go to higher orders in 1/c .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Computing these subleading contributions is in general intractable from the CFT side alone as the theory is strongly coupled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The weakly coupled dual description makes it possible, at least in principle, as holographic correlators can be computed as amplitudes at various loop orders by using the AdS generalization of the standard Feynman diagram expansion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, it should be noted that such a recipe is rather impractical to use beyond the few simplest cases [1–5], due to the proliferation of diagrams and complicated AdS vertices [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In fact, just at the tree level, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=', at order 1/c, the computation of general four-point functions remained an unsolved problem for almost two decades.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' A much better strategy, initiated in [7, 8], is the bootstrap approach, which led to the complete tree-level four-point functions of 1 2-BPS operators with arbitrary Kaluza-Klein (KK) levels for IIB supergravity in AdS5 × S5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The bootstrap approach exploits both the amplitude intuition from the bulk and the superconformal constraints from the boundary, and is currently the most efficient method for computing holographic correlators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' At the moment, there is already a wealth of results at tree level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' For example, general four- point functions of arbitrary 1 2-BPS operators have been computed in closed forms in all maximally superconformal theories [9, 10], as well as in theories with half the amount of maximal superconformal symmetry [11–13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1 By contrast, our understanding for loop level correlators is much more limited, even in the paradigmatic example of IIB supergravity on AdS5 × S5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The first one-loop correlator was computed in [15, 16] for the stress tensor multiplet in position space and later in Mellin space [17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The calculation was generalized to four-point functions with higher KK levels in [18–20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, explicit one-loop results are still case-by-case with the exception for the ⟨22pp⟩ family in [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' At two loops and higher, the situation is more difficult.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The strategy at one loop, which is based on the AdS unitarity method [21], now requires the additional input of multi-trace operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Such information is not yet available in the literature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2 Therefore, one can in principle only compute a part of the correlator that corresponds to the iterated s-channel cuts in flat space [24, 25].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, it turns out that this difficulty can be overcome at two loops by formulating an ansatz that is structured by an observed extra hidden symmetry in the leading Lorentzian singularities, together with additional physical constraints such as the behavior in the flat-space limit [26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In this way, the four-point two-loop correlator of stress tensor multiplets has also been bootstrapped [26, 27].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 1See [14] for a recent review.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 2For example, at two loops there are exchange contributions from triple-trace operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' These can be in principle extracted from tree-level five-point functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, only five-point functions of the form ⟨pp222⟩ have been computed [22, 23] while extracting the data requires all ⟨pqr22⟩ five-point functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' – 2 – In this paper, we continue to explore the loop-level calculation of holographic correla- tors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, instead of considering correlators of super gravitons, we will focus on super gluons of SYM in AdS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' More precisely, we consider a decoupling sector of certain 4d N = 2 SCFTs in the holographic limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' These SCFTs can be engineered by using either a stack of N D3-branes probing F-theory singularities [28, 29] or D3-branes with probe D7-branes [30].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The near horizon geometries in both cases include an AdS5 ×S3 subspace which hosts localized degrees of freedom corresponding to the gluons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In the limit of N → ∞, the gluon degrees of freedom effectively decouple from the graviton degrees of freedom living in the full 10d bulk via 1/N suppressions in the vertices [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The resulting physics in 8d is the same regardless of the model we choose.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Strictly speaking, the decoupling happens only at the leading order and correlators at subleading orders include gravity contributions as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, in this paper we will choose to turn off gravity to all orders in 1/N and our goal is to compute the super gluon four-point correlators in this SYM theory in AdS5 × S3 to two loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The motivations for considering super gluon correlators in such a setup are two fold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' First, as we already mentioned, holographic correlators are on-shell scattering amplitudes in AdS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' It is natural to wonder if various remarkable properties of flat-space amplitudes admit generalizations in curved backgrounds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In particular, does the double copy relation [31], which famously states gravity is the “square” of YM, still holds in AdS?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' To this end, it makes sense to decouple gravity and study the amplitudes of just SYM in AdS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In fact, analysis of this model at tree level already showed evidence for such a generalization at four points [32].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Here we will compute the loop corrections of the super gluon four-point functions which will serve as the starting point for exploring further generalizations of double copy at higher genus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Second, the super gluon case also provides a useful playground for acquiring deeper understandings of various results from the supergravity setup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The position space method for computing loop-level correlators so far have only been tested in AdS5 × S5 and it is a priori unclear whether it can be applied to other backgrounds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In this paper, we will show that such a method can be successfully applied to AdS5 × S3 and leads to similar results to the supergravity case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In the process, we also provide a nontrivial consistency check of the one-loop result which was previously obtained in [33] using Mellin space techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Moreover, the various different color structures allow us to have a more refined understanding of the dynamical structures of the correlators which are similar in the two cases, whereas in the supergravity case all structures are mixed up due to the absence of colors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Let us briefly outline our strategy and the key results of the paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Our approach is similar to that of [15, 19, 26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' We first make an ansatz in position space which requires a set of building block functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Due to the similarity with the supergravity case at tree level, we assume that single-valued multiple polylogarithms (SVMPLs) continue to be a good basis for the super gluon correlators at one and two loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In other words, the correlators are assumed to be linear combinations of SVMPLs with rational functions of the cross ratios as coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, this turns out to be a bit too general.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In the supergravity case, the existence of a tree-level 10d superconformal symmetry [34] highlights a special eighth-order differential operator ∆(8) which relates the correlators of the top and bottom components – 3 – of the super graviton multiplet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' By unitarity this symmetry extends to the leading part of the Lorentzian singularities at arbitrary loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Using this operator at loop levels, the supergravity correlators can be more succinctly written in terms of the pre-correlators L [19, 26, 27] H1-loop sugra = ∆(8)L1-loop sugra + 1 4Htree sugra , (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1a) H2-loop sugra = � ∆(8)�2 L2-loop sugra + 5 4H1-loop sugra − 1 16Htree sugra , (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1b) together with additional lower-order correlators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' A similar 8d superconformal symmetry also appears in the tree-level super gluon correlators [13] and the role of ∆(8) is replaced by a fourth-order operator ∆(4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In analogy with the supergravity case, we assume that similar pre-correlators can also be defined for super gluons H1-loop SYM = ∆(4)L1-loop SYM + ¯Htree SYM , (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2a) H2-loop SYM = � ∆(4)�2 L2-loop SYM + � H1-loop SYM + � Htree SYM , (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2b) where ¯Htree SYM and � Htree SYM are “tree-like” correlators and � H1-loop SYM is a “one-loop-like” corre- lator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' We will be more precise about the meaning of “tree-like” and “one-loop-like”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' But for the moment it suffices to say they are characterized by the transcendental degrees of SVMPLs expected at each loop order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Then the position space ansatz in terms of SVMPLs is formulated in terms of the pre-correlators L and the lower-order objects Hi, in parallel with the supergravity story.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Note that, unlike supergravity, super gluon correlators have different color structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Therefore, we make such an ansatz for each independent color structure and assume the correlator to be a linear combination of all these structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' To perform the bootstrap, we impose a number of consistency conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' These are Leading logarithmic singularities Crossing symmetry together with a few other constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Here the leading logarithmic singularities rely only on the tree-level data and can be computed at any loop order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' At two loops, the additional constraints further include comparison with the scattering amplitude in a proper flat-space limit that can be computed independently using flat-space techniques, and the data of twist-4 operators which can be extracted from the tree and one-loop correlators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Imposing these constraints, we find that all parameters in the ansatz are fixed except for those corresponding to the counterterms needed for the UV divergences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Moreover, the tree-like and one-loop-like terms turn out to be exactly the tree-level and one-loop correlators except for simple replacements for the color structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The rest of the paper is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' We review in Section 2 some preliminaries of super gluon four-point functions which include the superconformal kinematics, color structure and superconformal block decomposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In Section 3 we review how the leading logarithmic singularities can be constructed from the tree-level data and compute them in closed forms using hidden conformal symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In Section 4 we introduce the position – 4 – space method and demonstrate it by bootstrapping the one-loop correlator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In Section 5 we apply the method to the two-loop correlator and obtain the full answer by imposing constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In Section 6 we outline a few future directions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The paper also has several appendices where we include further technical details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In Appendix A we give a brief review of the properties of SVMPLs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Appendix B contains the complete analytic result for the reduced correlator at one loop.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The details of the flat-space two-loop amplitude are presented in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In Appendix D we discuss the computations related to the twist-4 data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 2 Preliminaries In this paper, we consider holographic correlators corresponding to super gluon scattering in AdS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' To be concrete, we consider SYM in AdS5 ×S3 which arises as a decoupling sector of certain 4d N = 2 SCFTs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' One can construct these SCFTs from a stack of N D3-branes, by either using them to probe F-theory singularities [28, 29] or by adding a few probe D7-branes [30].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In either case, in the near horizon limit there is an AdS5 × S3 subspace in the total ten dimensional spacetime which is locally AdS5 ×S5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' On this subspace there are localized degrees of freedom transforming as an N = 1 vector multiplet and in the adjoint representation of certain flavor group GF of the boundary CFT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Here GF depends on the theory and is a gauge group from the bulk perspective.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='3 Since the N = 1 vector multiplet contains fields with Lorentz spin at most 1, its KK reduction with respect to S3 also leads to fields with the same maximal spin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' These are the massless and massive AdS gluons and their super partners.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Because of the bound on spins all the KK modes have to reside in 1 2-BPS multiplets by the 4d N = 2 representation theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Their conformal dimensions are fully fixed by R-symmetry and therefore are independent of the bulk theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' More precisely, the superconformal primaries of these 1 2-BPS are scalar operators Ok labelled by an integer k = 2, 3, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='. They have conformal dimension ∆ = k and transform in the spin-k 2 representation of the SU(2)R R-symmetry group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Moreover, they transform in the adjoint representaiton of the flavor group GF .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' We will call the fields dual to these superprimaries the super gluons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The real spinning gluon fields are superconformal descendants in the multiplets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' By contrast, the super gravitons and their super partners live in the full ten dimen- sional spacetime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Unlike the supergluons, their KK spectrum depends on the specific theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, an interesting fact of all these 4d N = 2 SCFTs is that there is a hier- archy in the couplings at large N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' For example, the cubic coupling of three super gluons (or their superconformal descendants) is of order 1/ √ N, while the coupling involving two super gluons and one super graviton is of order 1/N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Therefore, for large N the lead- ing contribution to the super gluon correlators comes only from the 8d SYM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Subleading corrections in 1/N will in general contain graviton contributions as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' As mentioned in the introduction, we will continue to study loop corrections of the four-point correlator of the super gluon operator O2 in the AdS5 × S3 SYM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Although this does not give the full answer for this correlator in N = 2 SCFTs, it makes sense from 3Therefore, in the following we will use “flavor”, “color” and “gauge” interchangeably.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' – 5 – the perspective of exploring curved space generalizations of gauge theory amplitudes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' This section serves to provide some preliminary features of this correlator, which will be used in our bootstrap computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1 Four-point correlators With all indices restored, the super gluon operator O2 has the form OI;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='a1a2 2 (x) , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='1) where I = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' , dim(GF ) is the flavor symmetry index and ai = 1, 2 are the SU(2)R R- symmetry indices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' It is convenient to contract the R-symmetry indices with two-dimensional polarization spinors OI 2(x;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' v) = OI;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='a1a2 2 va1va2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2) The four-point function GI1I2I3I4(xi;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' vi) = ⟨OI1 2 (x1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' v1)OI2 2 (x2;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' v2)OI3 2 (x3;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' v3)OI4 2 (x4;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' v4)⟩ (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='3) is therefore a function of both the spacetime coordinates xi and the internal space spinors vi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Exploiting the bosonic symmetries, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' conformal symmetry and R-symmetry, we can write the correlator as a function of the cross ratios GI1I2I3I4 = (v1 · v2)2 (v3 · v4)2 x4 12x4 34 GI1I2I3I4(u, v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' α) (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='4) where xij = xi − xj, vi · vj = va i vb jϵab (ϵab being the 2d Levi–Civita symbol), and the cross ratios are u = x2 12x2 34 x2 13x2 24 = z¯z , v = x2 23x2 14 x2 13x2 24 = (1 − z)(1 − ¯z) , α = (v1 · v3) (v2 · v4) (v1 · v2) (v3 · v4) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='5) In addition, the ferminonic generators in the superconformal algebra impose further con- straints known as the superconformal Ward identities [35] (x∂x − α∂α) GI1I2I3I4(z, ¯z;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' α) �� α=1/x = 0 , x = z or ¯z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='6) Solving these identities, we can decompose the correlator into two parts GI1I2I3I4(z, ¯z;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' α) = GI1I2I3I4 0 (z, ¯z;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' α) + RHI1I2I3I4(z, ¯z) , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='7) where R = (1 − zα)(1 − ¯zα) z¯z .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='8) Note that our definiton of R is different from that of [13] by a z¯z in the denominator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The first term GI1I2I3I4 0 is protected, while dynamical information of the correlator is encoded in the reduced correlator HI1I2I3I4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In our case of O2 correlator, HI1I2I3I4 is simply a function of the spacetime cross ratios {z, ¯z} (or equivalently {u, v}) and is free of the R-symmetry cross ratio α.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' – 6 – We study the expansion of the correlator with respect to the large flavor central charge CJ 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' For convenience, we use the small parameter aF = 6/CJ , with respect to which the expansion reads HI1I2I3I4 2222 ≡ H = H(0) + aF H(1) + a2 F H(2) + a3 F H(3) + · · · .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='9) This expansion has a nice interpretation from the bulk point of view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The leading con- tribution H(0) is associated with the disconnected part of scattering in AdS and can be evaluated by generalized free field theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The first correction H(1) is the tree-level scat- tering of the super gluons, which has been obtained in [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The higher-order correction H(L+1) corresponds to scattering at L loops, where the one-loop case has been computed in [33] using Mellin space techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='2 Projectors and color decomposition Since we are studying gluon scattering, as usual the correlator H(L+1) at each perturbative order splits into various color factors and their corresponding dynamical factors 5 � H(L+1)�I1I2I3I4 = � C CI1I2I3I4H(L+1) C .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='10) A color factor CI1I2I3I4 is constructed out of the structure constants fIJK of the gauge group according to the topology of a diagram that may arise at the given loop order according to Feynman rules (or Witten rules in AdS), and so the summation above carries over all possible topologies at L loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The dynamical factors H(L+1) C are functions of kinematic variables {z, ¯z}, and with the above decomposition they only rely on diagram topologies as well, regardless of any specific choice of the gauge group GF .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The decomposition (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='10) is not the most convenient for practical computations as the color factors CI1I2I3I4 are highly redundant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' So instead one often seeks for other types of color decompositions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Because our computation requires the input from CFT data of the spectrum and the coefficients arising in OPEs, it is preferable to decompose the color factors in a way that resembles the conformal block expansion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' This can be fulfilled by specifying a particular channel (say the s-channel) and introduce an operation P I1I2|I3I4 a that picks out irreducible represetation a of the flavor group from the tensor products of two adjoints adj ⊗ adj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' This is called an s-channel projector, and by definition it satisfies the symmetry properties P I1I2|I3I4 a = (−1)|Ra|P I2I1|I3I4 a , P I1I2|I3I4 a = P I3I4|I1I2 a , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='11) where |Ra| stands for the parity of representation a, and the idempotency condition P I1I2|I5I6 a P I6I5|I3I4 b = δabP I1I2|I3I4 a .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='12) 4The flavor central charge CJ appears in the flavor current two-point functions as ⟨J I µ (x)J J ν (0)⟩ = CJ 2π2 δIJ (δµν−2 xµxν x2 ) x6 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Moreover, via supersymmetry, it is related to the three-point function coefficient C2 2,2,2 of ⟨O2O2O2⟩ by CJ = 1 6C2 2,2,2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 5In the context of scattering amplitudes these coefficients of color factors are more frequently called kine- matic factors (when referring to numerators in Feynman diagrams).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In this paper we call them dynamical factors to remind the readers that they contains the dynamical information of the theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' – 7 – In particular from (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='12) we also have P I1I2|I3I4 a P I1I2|I3I4 b = δabdim(Ra).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Therefore ev- ery color factor appearing in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='10) receives a unique decomposition onto the s-channel projectors CI1I2I3I4 = � a∈adj⊗adj P I1I2|I3I4 a Ca , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='13) with coefficients Ca, or equivalently Ca = dim(Ra)−1P I1I2|I3I4 a CI1I2I3I4 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='14) The efficiency of these projectors comes from the fact that the set of irreducible rep- resentations arising in adj ⊗ adj depends only on the gauge group GF but not on the perturbative order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' As a result, the color decomposition of the reduced correlator H as well as any term H(L+1) in the expansion (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='9) can be carried out in a uniform manner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Generically, we have HI1I2I3I4 = � a∈adj⊗adj P I1I2|I3I4 a Ha , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='15) and H(L+1) follows similarly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Furthermore, the idempotency condition (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='12) also makes the recursive relation between different loop levels very simple, as will be further illustrated in the next section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' As a simple example for the use of projectors, let us quickly review the tree-level correlator H(1), which was computed in [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Its takes the following form H(1) = csH(1) s + ctH(1) t + cuH(1) u , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='16) with H(1) s = u3 3 � 2∂u + (1 + v)∂u∂v + u∂2 u � ¯D1111, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='17a) H(1) t = −u3 3 � 2∂v + v∂2 v + (1 + u)∂u∂v � ¯D1111, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='17b) H(1) u = u3 3 � 2∂v + v∂2 v − 2∂u + (u − v)∂u∂v − u∂2 u � ¯D1111 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='17c) Here ¯D1111 is an example of the ¯D-functions which are contact Witten diagrams in AdS 6 ¯D1111(z, ¯z) = 1 z − ¯z � 2Li2(z) − 2Li2(¯z) + log(z¯z) log �1 − z 1 − ¯z �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='18) cs/t/u are color factors built from structure constants (cs)I1I2I3I4 = fI1I2JfJ I3I4, (ct)I1I2I3I4 = fI1I4JfJ I2I3, (cu)I1I2I3I4 = fI1I3JfJ I4I2, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='19) which are diagrammatically depicted in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Note again in this decomposition the kinematic factors H(1) s/t/u are independent of the gauge group GF .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' When decomposing using the projectors, let us assume that we are working with the gauge group E8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In this case adj ⊗ adj includes altogether five irreducible representations 6For a review of the precise definition and general properties of ¯D-functions, see Appendix C of [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' – 8 – I1 I4 I3 I2 cs = ct = I1 I4 I3 I2 cu = I1 I4 I3 I2 Figure 1: Tree color structures cs, ct and cu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 1, 3875, 27000, 248 (adj), and 30380.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The former three representations are parity even and the latter two are paritty odd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Note that (cs)I1I2I3I4 already represents the exchange of the adjoint representation 248 in the s-channel, it is therefore proportional to the projector P I1I2|I3I4 248 , and we have (cs)a ≡ P I1I2|I3I4 248 (cs)I1I2I3I4 = ψ2h∨(0 ↑ 1 , 0 ↑ 3875 , 0 ↑ 27000 , 1 ↑ 248 , 0 ↑ 30380 )T , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='20) where h∨ is the dual Coxeter number, ψ2 is the length squared of the longest root.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' By contrast the decomposition of ct, cu involves a mixture of different s-channel projectors (ct)a = −ψ2h∨ � 1, 1 5, − 1 30, 1 2, 0 �T , (cu)a = ψ2h∨ � 1, 1 5, − 1 30, −1 2, 0 �T .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='21a) One easily sees that the Jacobi identity (cs)a + (ct)a + (cu)a = 0 is satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Consequently the coefficients in the projector decomposition of the whole tree-level correlator H(1) are H(1) 1 =ψ2h∨ � −H(1) t + H(1) u � , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='22a) H(1) 3875 =ψ2h∨ 5 � −H(1) t + H(1) u � , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='22b) H(1) 27000 = − ψ2h∨ 30 � −H(1) t + H(1) u � , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='22c) H(1) 248 =ψ2h∨ 2 � 2H(1) s − H(1) t − H(1) u � , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='22d) H(1) 30380 =0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='22e) Quite remarkably, at this specific level the coefficients of projectors with equal parity are in fact the same up to some overall constant factors, as was observed in a more general setup in [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='3 Spectrum and conformal block decomposition As mentioned before our computation partly relies on the existing data of operators ob- tained from lower loops, so it is helpful to have a quick look at the structure of OPE and the related block expansion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Thanks to the 4d N = 2 superconformal symmetry, the correlator GI1I2I3I4 admits a decomposition into superconformal blocks in correspondence to the exchanges of different superconformal multiplets in the four-point function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The – 9 – relevant sueprmultiplets are listed in Table 1 and a complete classification can be found in [35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The OPE of two 1 2-BPS multipelts B1 contains the following supermultiplets B1 ⊗ B1 ≃ 2 � p=0 Bp ⊕ � ℓ≥0 � � 1 � p=0 Cp,( ℓ 2 , ℓ 2) � ∆ A∆ 0,( ℓ 2 , ℓ 2) � � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='23) Here Bp and Cp,( ℓ 2 , ℓ 2 ) are protected multiplets and their twists τ = ∆ − ℓ are bounded from above by the allowed R-symmetry charges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In contrast, there is no upper bound on the twists of the long multiplets A∆ 0,( ℓ 2 , ℓ 2 ) and their dimensions are not protected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Instead, they have a lower bound in the holographic limit as they are double-trace (and more generally multi-trace) operators formed by single-trace operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='7 Let us also note that the superprimaries of the long multiplets are only allowed to be R-symmetry singlets in order for the representations of the entire multiplet to fit into the four-point function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The long multiplets play a key role in the paper as the loop corrections correspond to precisely the contribution of these multipelts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Multiplet Label SU(2)R Dimension ∆ and spin ℓ Half-BPS BR R/2 ∆ = 2R, ℓ = 0 Semi-short CR,(ℓ/2,ℓ/2) R/2 ∆ = 2 + 2R + ℓ Long A∆ R,(ℓ/2,ℓ/2) R/2 ∆ ≥ 2 + 2R + ℓ Table 1: Supermultiplets that appear in the fusion rules of two B’s for N = 2 SCFTs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' We will focus on the reduced correlator H which has already taken superconformal symmetry into account.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' In this way the superconformal block decomposition simply reduces to just the ordinary conformal block decomposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' As superconformal symmetry and gauge symmetry commute, this directly passes through the color projector decomposition, and in terms of each component in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='15) this reads [35] Ha(z, ¯z) = � τa,ℓ aagτa+2,ℓ(z, ¯z) , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='24) where τa and ℓ sum over the spectrum of the supermultiplets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Note that the shift in τ by 2 in the ordinary conformal block gτ+2,ℓ is a consequence of the superconformal symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' The detailed expression of these blocks is [36] gτ,ℓ = z¯z ¯z − z � k τ−2 2 (z)k τ+2ℓ 2 (¯z) − k τ−2 2 (¯z)k τ+2ℓ 2 (z) � , kh(z) = zh 2F1(h, h, 2h, z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='25) Since the long multiplets are not protected, in the limit of N → ∞ their twists as well as OPE coefficients receive perturbative corrections with respect to small aF τa =τ0 + aF γ(1) a + a2 F γ(2) a + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='26a) aa(τ, ℓ) =a(0) a + aF a(1) a + a2 F a(2) a + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='26b) 7This bound is stronger than the unitarity bound in Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' – 10 – Substituting the above expansion into (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='24) gives the following series expansion for Ha Ha = � τ0,ℓ a(0) a gτ0+2,ℓ(z, ¯z) � �� � H(0) a +aF � τ0,ℓ � a(0) a γ(1) a ∂τ0 + a(1) a � gτ0+2,ℓ(z, ¯z) � �� � H(1) a + a2 F � τ0,ℓ �1 2a(0) a (γ(1) a )2∂2 τ0 + (a(1) a γ(1) a + a(0) a γ(2) a )∂τ0 + a(2) a � gτ0+2,ℓ(z, ¯z) � �� � H(2) a + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='27) The first term H(0) a receives contributions only from long operators whose a(0) a are non- vanishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' From large N factorization, H(0) a is given by the disconnected correlator and these contributing operators can only be double-trace operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' However, these operators are degenerate at the classical level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' For instance, among the double-trace operators : O2□n−2∂ℓO2 : , : O3□n−3∂ℓO3 : , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' , : On∂ℓOn : (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='28) all have classical twist τ (0) = 2n and spin ℓ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Consequently, each term in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='27) should not be literally understood as the contribution from a single operator, but rather in an averaged sense.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Moreover, at higher orders in aF there are also higher-trace operators appearing in the OPE 8, which can have the same twist as the double-trace operators and will enter the mixing as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Therefore, in a precise description it is necessary to use an extra label i to distinguish different operators in the degeneracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Then the coefficient a(0) a γ(1) a should in fact be understood as ⟨a(0) a γ(1) a ⟩ ≡ � i a(0) i,aγ(1) i,a , and a(0) a � γ(1) a �2 as ⟨a(0) a � γ(1) a �2 ⟩ ≡ � i a(0) i,a � γ(1) i,a �2 , and so on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' 3 Leading logarithmic singularities As an analytic function of the kinematic variables z and ¯z, a conformal correlator can in principle be constructed out of its singularities by dispersion-type relations, in a similar way as the dispersion relation that generates a four-point scattering amplitude from its physical channel discontinuities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' For generic CFTs such relations were formulated in [37].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' This means that the defining data for a correlator is necessarily encoded in its singularities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' While our computation does not rely on the dispersion relations, these data still provide a vital input in determining the loop-level corrections to the reduced correlator H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' When viewed in the perturbative expansion (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='27) these singularities are sourced at small u by the log(u) factors arising from the derivatives acting on the conformal block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' Recall in the definition (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content='27) that gτ,ℓ(z, ¯z) ∝ uτ/2, so at each order ap F the reduced correlator can be organized in terms of powers of log(u) H(p) a (z, ¯z) = 1 2pp!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'} +page_content=' logp(u) � τ0,ℓ ⟨a(0) a (γ(1) a )p⟩ gτ0,ℓ(z, ¯z) + � terms with logk
0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' the final result for the planar double box integral Ipdb 1234 in 8 − 2ϵ dimensions reads Ipdb 1234 =(−s)1−2ϵ � 1 144 1 ϵ2 + 79 + 2x 1728 1 ϵ + −48 + 3303x + 242x2 20736x + (1 + x)(2 − 15x + x2) 432x2 (G0 − G1) + (1 − 6x + 34x2 + 67x3)π2 2592(x − 1)x3 − 1 − 5x + 25x2 + 3x3 216x3 � G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 − G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 − G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + 5π2 12 � − −1 + 8x + 17x2 108(x − 1)x2 � G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 − G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 − G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + π2 6 (2G0 + G1) − ζ3 � − 3 + x 18(x − 1)2 � G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 − G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 − G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + π2 6 (2G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1) − ζ3G0 + 17π4 360 �� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='6) where we use the abbreviation G⃗a ≡ G⃗a(x−1), and the dimensionless parameter x is related to the Mandelstam variables s, t or the scattering angle θ by x ≡ 1 + t s = 1 + cos θ 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='7) 15The authors are grateful to Lilin Yang for sharing data for these integrals in 4−2ϵ dimensions, together with a set of master integrals needed for dimensional recursion which were selected following [66, 67].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The dimensional recursion is performed using the Mathematica package LiteRed [68].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' – 41 – In the same region the result for the non-planar double box integral Indb 1234 reads Indb 1234 =(−s)1−2ϵ � 1 288 1 ϵ2 + 37 1728 1 ϵ + −60 − 3607x + 3625x2 − 36x3 + 18x4 51840(x − 1)x + (−12 + 42x − 47x2 + 2x3 − 63x4 + 18x5) 25920(x − 1)2 (G1 + iπ) − 60 − 328x + 727x2 − 798x3 + 399x4 25920(x − 1)2x2 (G0 − G1) + π2 1728 − 30 − 173x + 428x2 − 600x3 + 525x4 − 270x5 + 90x6 12960(x − 1)3x3 (G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + iπG0) + 30 − 113x + 160x2 − 107x3 + 36x4 + 6x5 12960(x − 1)x3 � G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + iπ(G0 + G1) − π2 2 � + (x − 1)3(−30 − 7x + 4x2 + 9x3 + 9x4) 12960x3 (G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + iπG1) − (3x − 2)(3x − 1)ζ3 720(x − 1)2x2 + (2x − 1)(2 − 5x + 5x2) 2160(−1 + x)2x2 � G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 − 2G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + iπ(G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 − 2G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0) − π2 2 G0 + iπ3 2 � + (x − 1)3(2 + x) 2160x2 � G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 − 2G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 + iπ(G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0 + G1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1 − 2G0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1) − π2 2 G1 �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='8) The remaining integrals in (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='3) can be obtained from (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='6) and (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='8) by permuting the particle labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' For readers’ convenience we also record these results in the ancillary file.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Note that the expressions from permutations may live in different physical regions, and so before assembling them together one needs to analytically continued the Mandelstam variables to the same region following the standard iε prescription.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' For our computation we finally lands on the region s, t < 0 and u > 0, in which (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='6) and (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='8) directly apply.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' In order to perform the comparison in the bulk-point limit we need to analytically continue our ansatz for the correlator from Euclidean region to physical region as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Following [27], our prescription is to continue z counter-clockwisely around 0 and ¯z clock- wisely around 1, and then set z = ¯z + 2ω¯z√1 − ¯z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The bulk-point limit z → ¯z is then reached by setting ω → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' As was already pointed out in Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='2, when taking this limit at two loops [∆(4)]2L(3) dominates over the modification terms, and the action of ∆(4) reduces to a simple multiplication factor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Therefore it suffices to compare A2-loop with the pre-correlator L(3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The detailed connection between these two objects is lim ω→0 (z − ¯z)5L(3) � z⟲0, ¯z⟲1���� z=¯z+2ω¯z√1−¯z = 48π2(1 − ¯z)2¯z6s−2 ˜ A2-loop(x) ��� x=1/¯z , (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='9) Note the above relation is already expressed in terms of the reduced amplitude ˜ A2-loop.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Although the full amplitude A2-loop contains an extra factor stAtree 1234 including polarization vectors, in practical computation we do not have to bother manipulating it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The reason is that comparison with the leading log data in (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='15) already fully determines contributions of some MPLs of highest transcendental weight, and matching them with the corresponding MPL contributions in ˜ A2-loop easily determines the factors appearing in the above relation – 42 – (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='9) (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=', matching coefficients of G0,0,0,0 or G1,1,1,1 in the limit).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Comparison between the remaining contributions then generates constraints for the undetermined variables in the ansatz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [46] proposed a simpler connection than (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='9), the relation between discontinuities dDisc H and Disc A, in the context of graviton scattering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Similar connection should also apply to the gluon scattering studied here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Ideally one would not expect much difference between the comparison at the level of the full correlator and that of the discontinuity, because in principle the correlator H can be reconstructed from its double discontinuity dDisc H through Lorentz inversion [47], which is the CFT counterpart of the dispersion relation relating A and its discontinuity Disc A in flat space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' However, at loop level these dispersion relations can be polluted by the presence of finite spin contributions to the cor- relator/amplitude, which imposes extra data in addition to the discontinuities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Therefore one expects that constraints from (C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='9) are stronger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' D Recursion of twist-4 data at log2 u The reduced correlator H can be organized in terms of power expansions in log u in the small u limit, where the power of log u goes up to n + 1 at n loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The coefficient functions of these log u powers encode different combinations of the expansion coefficients of the CFT data with respect to aF , which are schematically listed in Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The goal of this appendix is to compute the combination ⟨a(1) a (γ(1) a )2 + 2a(0) a γ(1) a γ(2) a ⟩ for twist-4 operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' This data contributes to the two-loop correlator in the log2 u coefficient, as explicitly shown in (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='22), and we use them as one of the inputs in our bootstrap algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' As is clear from the expression, only tree-level and one-loop correlators are needed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Moreover, the twist-4 operators are free of operator mixing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' This fact makes it possible to extract their CFT data from just the O2 correlators alone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The angle brackets ⟨.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='⟩ will also be dropped as they are no longer necessary in this case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' order a1 F a2 F a3 F log(u)0 ⟨a(1) a ⟩ ⟨a(2) a ⟩ ⟨a(3) a ⟩ log(u)1 ⟨a(0) a γ(1) a ⟩ ⟨a(1) a γ(1) a + a(0) a γ(2)⟩ ⟨a(2) a γ(1) a + a(1) a γ(2) a + a(0) a γ(3) a ⟩ log(u)2 ⟨a(0) a (γ(1) a )2⟩ ⟨a(1)(γ(1) a )2 + 2a(0) a γ(1) a γ(2) a ⟩ log(u)3 ⟨a(0) a (γ(1) a )3⟩ Table 2: The OPE data encoded in the coefficient functions of different log u powers for correlators up to two loops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The most efficient way to extract the CFT data from explicit correlators is to use the Lorentzian inversion formula [47].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Applying it to the tree-level correlator H(1), we obtain – 43 – the following data for the twist-4 operator with spin ℓ a(0) a γ(1) a =(−ct + (−1)ℓcu)a 2Γ(ℓ + 3)2 Γ(2ℓ + 5) , (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1) a(1) a =(−ct + (−1)ℓcu)a Γ(ℓ + 3)2 Γ(2ℓ + 5) � 2ψ(0)(ℓ + 3)−2ψ(0)(2ℓ + 5)+ 2 3 + (−1)ℓ 6 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='2) Here for the color part (#)a is defined by projectors as in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='20).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Similarly, from the one-loop correlator H(2), we can extract the following combinations a(0) a (γ(1) a )2 =(dst + (−1)ℓdsu)a Γ(ℓ + 3)2 Γ(2ℓ + 5) 8 (ℓ + 1)(ℓ + 4) , (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='3) a(1) a γ(1) a + a(0) a γ(2) a =(dst + (−1)ℓdsu)a Γ(ℓ + 3)2 Γ(2ℓ + 5) � −4 � 2ℓ3 + 23ℓ2 + 65ℓ + 32 � ℓ(ℓ + 1)2(ℓ + 4)2(ℓ + 5) + 2 � 12ψ(0)(ℓ + 3) − 12ψ(0)(2ℓ + 5) + 11 � 3(ℓ + 1)(ℓ + 4) � − (1 + (−1)ℓ)(dtu)a Γ(ℓ + 3)2 Γ(2ℓ + 5) 2 � 5ℓ2 + 25ℓ + 24 � 3ℓ(ℓ + 1)(ℓ + 4)(ℓ + 5) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='4) By comparing (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1) and (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='3), we get γ(1) a = (−ct + (−1)ℓcu)a 2 (ℓ + 1)(ℓ + 4) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='5) We could also solve for a(0) a , a(1) a , γ(2) a .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' However, to compute the wanted data it is sufficient to consider the following combination a(1) a (γ(1) a )2 + 2a(0) a γ(1) a γ(2) a = 2 � a(1) a γ(1) a + a(0) a γ(2) a � � γ(1) a � − � a(1) a � � γ(1) a �2 , (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='6) and get a(1) a (γ(1) a )2 + 2a(0) a γ(1) a γ(2) a = − (1 + (−1)ℓ)(fs)a Γ(ℓ + 3)2 Γ(2ℓ + 5) 16 � 5ℓ2 + 25ℓ + 24 � 3ℓ(ℓ + 1)2(ℓ + 4)2(ℓ + 5) + (es1 + (−1)ℓes2)a Γ(ℓ + 3)2 Γ(2ℓ + 5) � −32 � 2ℓ3 + 23ℓ2 + 65ℓ + 32 � ℓ(ℓ + 1)3(ℓ + 4)3(ℓ + 5) + 8 � 12ψ(0)(ℓ + 3) − 12ψ(0)(2ℓ + 5) + 18 − (−1)ℓ� 3(ℓ + 1)2(ℓ + 4)2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='7) Note this result should only be trusted down to ℓ = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' This is because it uses the one-loop data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' The presence of the contact counterterm at one loop spoils the analyticity in spin at ℓ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' – 44 – References [1] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Freedman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Mathur, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Matusis, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rastelli, “Correlation functions in the CFT(d) / AdS(d+1) correspondence,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B546 (1999) 96–118, arXiv:hep-th/9804058 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [2] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' D’Hoker, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Freedman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Mathur, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Matusis, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rastelli, “Graviton exchange and complete four point functions in the AdS / CFT correspondence,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B562 (1999) 353–394, arXiv:hep-th/9903196 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [3] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Arutyunov and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Frolov, “Four point functions of lowest weight CPOs in N=4 SYM(4) in supergravity approximation,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' D62 (2000) 064016, arXiv:hep-th/0002170 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [4] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Arutyunov, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dolan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Osborn, and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Sokatchev, “Correlation functions and massive Kaluza-Klein modes in the AdS / CFT correspondence,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 665 (2003) 273–324, arXiv:hep-th/0212116.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [5] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Arutyunov and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Sokatchev, “On a large N degeneracy in N=4 SYM and the AdS / CFT correspondence,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B663 (2003) 163–196, arXiv:hep-th/0301058 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [6] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Arutyunov and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Frolov, “Scalar quartic couplings in type IIB supergravity on AdS(5) x S**5,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B579 (2000) 117–176, arXiv:hep-th/9912210 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [7] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rastelli and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “Mellin amplitudes for AdS5 × S5,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 118 no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 9, (2017) 091602, arXiv:1608.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='06624 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [8] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rastelli and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “How to Succeed at Holographic Correlators Without Really Trying,” JHEP 04 (2018) 014, arXiv:1710.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='05923 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [9] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “All Tree-Level Correlators for M-theory on AdS7 × S4,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 125 no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 13, (2020) 131604, arXiv:2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='06653 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [10] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “All Holographic Four-Point Functions in All Maximally Supersymmetric CFTs,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' X 11 no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 1, (2021) 011056, arXiv:2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='12505 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [11] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rastelli, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Roumpedakis, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “AdS3 × S3 Tree-Level Correlators: Hidden Six-Dimensional Conformal Symmetry,” JHEP 10 (2019) 140, arXiv:1905.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='11983 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [12] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Giusto, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Russo, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Tyukov, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Wen, “The CFT6 origin of all tree-level 4-point correlators in AdS3 × S3,” Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' C 80 no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 8, (2020) 736, arXiv:2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='08560 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [13] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Behan, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Ferrero, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “Gluon Scattering in AdS from CFT,” JHEP 06 (2021) 020, arXiv:2103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='15830 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [14] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bissi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Sinha, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “Selected topics in analytic conformal bootstrap: A guided journey,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 991 (2022) 1–89, arXiv:2202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='08475 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [15] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Aprile, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Heslop, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Paul, “Quantum Gravity from Conformal Field Theory,” JHEP 01 (2018) 035, arXiv:1706.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='02822 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [16] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bissi, “Loop Corrections to Supergravity on AdS5 × S5,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 119 no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 17, (2017) 171601, arXiv:1706.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='02388 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [17] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday, “On genus-one string amplitudes on AdS5 × S5,” JHEP 04 (2021) 005, arXiv:1812.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='11783 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [18] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Aprile, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Heslop, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Paul, “Loop corrections for Kaluza-Klein – 45 – AdS amplitudes,” JHEP 05 (2018) 056, arXiv:1711.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='03903 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [19] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Aprile, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Heslop, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Paul, “One-loop amplitudes in AdS5 × S5 supergravity from N = 4 SYM at strong coupling,” JHEP 03 (2020) 190, arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='01047 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [20] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “Simplicity of AdS Supergravity at One Loop,” JHEP 09 (2020) 008, arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='02663 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [21] O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Aharony, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bissi, and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Perlmutter, “Loops in AdS from Conformal Field Theory,” JHEP 07 (2017) 036, arXiv:1612.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='03891 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [22] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Gon¸calves, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Pereira, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “20′ Five-Point Function from AdS5 × S5 Supergravity,” JHEP 10 (2019) 247, arXiv:1906.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='05305 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [23] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Gon¸calves, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Meneghelli, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Pereira, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Boas, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “To appear,”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [24] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bissi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Fardelli, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Georgoudis, “Towards All Loop Supergravity Amplitudes on AdS5 × S5,” arXiv:2002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='04604 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [25] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bissi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Fardelli, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Georgoudis, “All loop structures in Supergravity Amplitudes on AdS5 × S5 from CFT,” arXiv:2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='12557 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [26] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Huang and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Yuan, “Graviton Scattering in AdS5 × S5 at Two Loops,” arXiv:2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='15174 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [27] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Paul, “Two-loop supergravity on AdS5 × S5 from CFT,” JHEP 08 (2022) 275, arXiv:2204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='01829 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [28] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Fayyazuddin and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Spalinski, “Large N superconformal gauge theories and supergravity orientifolds,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 535 (1998) 219–232, arXiv:hep-th/9805096.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [29] O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Aharony, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Fayyazuddin, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Maldacena, “The Large N limit of N=2, N=1 field theories from three-branes in F theory,” JHEP 07 (1998) 013, arXiv:hep-th/9806159.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [30] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Karch and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Katz, “Adding flavor to AdS / CFT,” JHEP 06 (2002) 043, arXiv:hep-th/0205236.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [31] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bern, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Carrasco, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Johansson, “Perturbative Quantum Gravity as a Double Copy of Gauge Theory,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 105 (2010) 061602, arXiv:1004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0476 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [32] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “Double Copy Relation in AdS Space,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 127 no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 14, (2021) 141601, arXiv:2106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='07651 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [33] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bissi, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “One-loop gluon amplitudes in AdS,” JHEP 02 (2022) 105, arXiv:2110.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='09861 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [34] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Caron-Huot and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='-K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Trinh, “All Tree-Level Correlators in AdS5×S5 Supergravity: Hidden Ten-Dimensional Conformal Symmetry,” JHEP 01 (2019) 196, arXiv:1809.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='09173 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [35] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Nirschl and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Osborn, “Superconformal Ward identities and their solution,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 711 (2005) 409–479, arXiv:hep-th/0407060.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [36] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dolan and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Osborn, “Conformal four point functions and the operator product expansion,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 599 (2001) 459–496, arXiv:hep-th/0011040.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [37] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Carmi and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Caron-Huot, “A Conformal Dispersion Relation: Correlations from Absorption,” JHEP 09 (2020) 009, arXiv:1910.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='12123 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [38] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Aprile, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Heslop, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Paul, “Unmixing Supergravity,” JHEP 02 – 46 – (2018) 133, arXiv:1706.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='08456 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [39] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Cutkosky, “Singularities and discontinuities of Feynman amplitudes,” J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 1 (1960) 429–433.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [40] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bern, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dixon, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dunbar, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Kosower, “One loop n point gauge theory amplitudes, unitarity and collinear limits,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 425 (1994) 217–260, arXiv:hep-ph/9403226.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [41] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bern, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dixon, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dunbar, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Kosower, “Fusing gauge theory tree amplitudes into loop amplitudes,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 435 (1995) 59–101, arXiv:hep-ph/9409265.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [42] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Glew, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Santagata, “BCJ relations in AdS5 × S3 and the double-trace spectrum of super gluons,” arXiv:2202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='09837 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [43] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dolan and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Osborn, “Conformal Partial Waves: Further Mathematical Results,” arXiv:1108.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='6194 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [44] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Chang and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lin, “Carving Out the End of the World or (Superconformal Bootstrap in Six Dimensions),” JHEP 08 (2017) 128, arXiv:1705.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='05392 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [45] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Maldacena, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Simmons-Duffin, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhiboedov, “Looking for a bulk point,” JHEP 01 (2017) 013, arXiv:1509.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='03612 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [46] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Alday and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Caron-Huot, “Gravitational S-matrix from CFT dispersion relations,” JHEP 12 (2018) 017, arXiv:1711.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='02031 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [47] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Caron-Huot, “Analyticity in Spin in Conformal Theories,” JHEP 09 (2017) 078, arXiv:1703.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='00278 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [48] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Ceplak, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Giusto, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Hughes, and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Russo, “Holographic correlators with multi-particle states,” JHEP 09 (2021) 204, arXiv:2105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='04670 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [49] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Ma and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Zhou, “Scattering bound states in AdS,” JHEP 08 (2022) 107, arXiv:2204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='13419 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [50] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Goncharov, “Multiple polylogarithms, cyclotomy and modular complexes,” Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Res.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 5 (1998) 497–516, arXiv:1105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='2076 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='AG].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [51] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Goncharov, “Multiple polylogarithms and mixed Tate motives,” arXiv:math/0103059.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [52] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Drummond and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Paul, “One-loop string corrections to AdS amplitudes from CFT,” JHEP 03 (2021) 038, arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='07632 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [53] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Duhr and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dulat, “PolyLogTools — polylogs for the masses,” JHEP 08 (2019) 135, arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='07279 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [54] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Chavez and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Duhr, “Three-mass triangle integrals and single-valued polylogarithms,” JHEP 11 (2012) 114, arXiv:1209.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='2722 [hep-ph].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [55] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Brown, “Single-valued multiple polylogarithms in one variable,” C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='Acad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='Paris I338 (2004) 527.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [56] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Dixon, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Duhr, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Pennington, “Single-valued harmonic polylogarithms and the multi-Regge limit,” JHEP 10 (2012) 074, arXiv:1207.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0186 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [57] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Bern, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rozowsky, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Yan, “Two loop four gluon amplitudes in N=4 superYang-Mills,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 401 (1997) 273–282, arXiv:hep-ph/9702424.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [58] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Kotikov, “Differential equations method: New technique for massive Feynman – 47 – diagrams calculation,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 254 (1991) 158–164.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [59] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Kotikov, “Differential equation method: The Calculation of N point Feynman diagrams,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 267 (1991) 123–127.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [Erratum: Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='B 295, 409–409 (1992)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [60] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Remiddi, “Differential equations for Feynman graph amplitudes,” Nuovo Cim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' A 110 (1997) 1435–1452, arXiv:hep-th/9711188.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [61] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Henn, “Multiloop integrals in dimensional regularization made simple,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 110 (2013) 251601, arXiv:1304.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1806 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [62] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Papadopoulos, “Simplified differential equations approach for Master Integrals,” JHEP 07 (2014) 088, arXiv:1401.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='6057 [hep-ph].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [63] O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Tarasov, “Connection between Feynman integrals having different values of the space-time dimension,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' D 54 (1996) 6479–6490, arXiv:hep-th/9606018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [64] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lee, “Space-time dimensionality D as complex variable: Calculating loop integrals using dimensional recurrence relation and analytical properties with respect to D,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 830 (2010) 474–492, arXiv:0911.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='0252 [hep-ph].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [65] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lee, “Calculating multiloop integrals using dimensional recurrence relation and D-analyticity,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Suppl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 205-206 (2010) 135–140, arXiv:1007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='2256 [hep-ph].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [66] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Jiang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Xu, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Yang, “Constructing canonical Feynman integrals with intersection theory,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' B 814 (2021) 136085, arXiv:2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='03045 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [67] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Jiang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Ma, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Xu, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Yang, “Baikov representations, intersection theory, and canonical Feynman integrals,” JHEP 07 (2022) 066, arXiv:2202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='08127 [hep-th].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' [68] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Lee, “LiteRed 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='4: a powerful tool for reduction of multiloop integrals,” J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Conf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' Ser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' 523 (2014) 012059, arXiv:1310.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content='1145 [hep-ph].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
+page_content=' – 48 –' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/19FQT4oBgHgl3EQfFDWe/content/2301.13240v1.pdf'}
diff --git a/1tE0T4oBgHgl3EQfuQHv/content/2301.02604v1.pdf b/1tE0T4oBgHgl3EQfuQHv/content/2301.02604v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..eeef23d4deea175f5390fdb282b1dcb84422cd19
--- /dev/null
+++ b/1tE0T4oBgHgl3EQfuQHv/content/2301.02604v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab19a58cf52a9a428b6d643708c5d6428fa8bb7402fd31550a4e7427d386654c
+size 2272156
diff --git a/2dFQT4oBgHgl3EQfFzW1/content/tmp_files/2301.13242v1.pdf.txt b/2dFQT4oBgHgl3EQfFzW1/content/tmp_files/2301.13242v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..54916be5fef92e29aee28a8f0e7d82e7b4da1ace
--- /dev/null
+++ b/2dFQT4oBgHgl3EQfFzW1/content/tmp_files/2301.13242v1.pdf.txt
@@ -0,0 +1,3565 @@
+Consensus based optimization with memory effects:
+random selection and applications
+Giacomo Borghi∗
+Sara Grassi†
+Lorenzo Pareschi†
+February 1, 2023
+Abstract
+In this work we extend the class of Consensus-Based Optimization (CBO) metaheuris-
+tic methods by considering memory effects and a random selection strategy. The proposed
+algorithm iteratively updates a population of particles according to a consensus dynamics
+inspired by social interactions among individuals. The consensus point is computed taking
+into account the past positions of all particles. While sharing features with the popular Parti-
+cle Swarm Optimization (PSO) method, the exploratory behavior is fundamentally different
+and allows better control over the convergence of the particle system. We discuss some im-
+plementation aspects which lead to an increased efficiency while preserving the success rate
+in the optimization process. In particular, we show how employing a random selection strat-
+egy to discard particles during the computation improves the overall performance. Several
+benchmark problems and applications to image segmentation and Neural Networks training
+are used to validate and test the proposed method. A theoretical analysis allows to recover
+convergence guarantees under mild assumptions of the objective function. This is done by
+first approximating the particles evolution with a continuous-in-time dynamics, and then by
+taking the mean-field limit of such dynamics. Convergence to a global minimizer is finally
+proved at the mean-field level.
+Keywords: consensus-based optimization, stochastic particle methods, memory effects, ran-
+dom selection, machine learning, mean-field limit
+Contents
+1
+Introduction
+2
+2
+Consensus-based optimization with memory effects
+4
+2.1
+Particles update rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+4
+2.2
+Random selection strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+5
+2.3
+Comparison with CBO and PSO . . . . . . . . . . . . . . . . . . . . . . . . . . .
+5
+∗RWTH
+Aachen
+University,
+Institute
+for
+Geometry
+and
+Applied
+Mathematics,
+Aachen,
+Germany
+(borghi@eddy.rwth-aachen.de)
+†University of Ferrara, Department of Mathematics and Computer Science & Center for Modelling Computing
+and Statistics, Ferrara, Italy (sara.grassi@unife.it, lorenzo.pareschi@unife.it)
+1
+arXiv:2301.13242v1 [math.OC] 30 Jan 2023
+
+3
+Numerical results
+7
+3.1
+Tests on benchmark problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+8
+3.2
+Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+13
+3.2.1
+Image segmentation
+. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+13
+3.2.2
+Approximating functions with NN . . . . . . . . . . . . . . . . . . . . . .
+16
+3.2.3
+Application on MNIST dataset . . . . . . . . . . . . . . . . . . . . . . . .
+17
+4
+Theoretical analysis
+19
+4.1
+Mean-field approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+19
+4.2
+Convergence in mean-field law . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+21
+4.3
+Random selection analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+23
+5
+Conclusions
+25
+A Proofs
+25
+A.1 Notation and auxiliary lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+25
+A.2 Proof of Proposition 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+27
+A.3 Proof of Theorem 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+28
+A.4 Proof of Proposition 4.2 and Theorem 4.2 . . . . . . . . . . . . . . . . . . . . . .
+30
+1
+Introduction
+Meta-heuristic algorithms are recognized as trustworthy, easy to understand and to adapt op-
+timization methods which have been widely applied to a several fields such as Machine Learn-
+ing [28], path planning [29] and image processing [45], to name a few. Starting form a set of
+possible solutions, a meta-heuristic algorithm typically updates such set iteratively by combining
+deterministic and stochastic choices, often inspired by natural phenomena. Exploration of the
+search space and exploitation of the current knowledge are the two fundamental mechanisms
+driving the algorithm iteration [46]. Examples of established meta-heuristic algorithms are given
+by Genetic Algorithm (GA) [17,42], Simulated Annealing (SA) [25], Particle Swarm Optimiza-
+tion (PSO) [24] and Differential Evolution (DE) [40]. We refer to [21] for a complete literature
+review.
+Consensus-Based Optimization (CBO) is a class of gradient-free meta-heuristic algorithms
+inspired by consensus dynamics among individuals. After its introduction [34] it has gained
+popularity among the mathematical community due to its robust mathematical framework [3,9,
+16,19]. In CBO algorithms, a population of particles concentrates around a consensus point given
+by a weighted average of the particles position. In the computation of such consensus point, more
+importance is given to those particles attaining relatively low values of the objective function.
+The exploration mechanism is introduced by randomly perturbing the particles positions at each
+iteration. Particles which are close to the consensus point are subject to small perturbations,
+while those that are far from it display a more exploratory behavior.
+In this work, following the recent analysis in [14], we study a Consensus-Based Optimization
+algorithm with Memory Effects (CBO-ME) where the consensus point is computed among the
+whole history of the particles positions and not just on the positions of the current iteration, as
+2
+
+in the original CBO method. This is done by keeping track of the best position found so far by
+each particle, and computing the consensus point among these “personal” bests. While sharing
+common elements with PSO, such as convergence to a promising point and the presence of
+personal bests, CBO-ME differs in the way the exploration mechanism is implemented. Indeed,
+in CBO-ME, as in CBO algorithms, the stochastic behavior is given by adding Gaussian noise to
+the particles dynamics and can be tuned independently on the exploitation mechanisms, leading
+to a better control over the particles convergence. Therefore, while in classical PSO methods it
+is the balance between local best and global best that governs the optimization strategy, in CBO
+methods it is the balance between exploration and exploitation mechanisms that determines the
+choice of parameters. We recall that a generalization of PSO methods that allows leveraging
+the same flexibility in searching the global minimum as in CBO algorithms has been recently
+presented in [14].
+Many real-life problems, especially those regarding Machine Learning, require to optimize a
+large number of parameters. Therefore, it essential to design fast algorithm to save computa-
+tional time and memory. This is a major weakness of swarm-based methods, which require a set
+of particles to minimize the problem, unlike gradient-based methods that can work on a single
+particle trajectory. For methods based on a collection of particles, existing algorithms can be
+improved by discarding particles whenever the system has a prominent exploitative behavior.
+This is sometimes referred as “natural selection strategy” in the DE literature [27,40] and aims
+to discard the non-promising solutions. Inspired by particle simulations techniques where it
+is important to preserve the particles distribution, we examine a “random selection strategy”
+where particles are discarded randomly based on the local consensus achieved. We will discuss
+such implementation aspects by testing CBO-ME against high-dimensional learning problems
+and theoretically analyze the impact of the random selection strategy on the system. In partic-
+ular, we prove that if the full particle system is expected to converge towards a solution to (2.1),
+so will the reduce one, provided a sufficient number of particles remains active. Note that, such
+analysis can be generalized to other particle dynamics and may be of independent interest.
+Owing to the convergence analysis of CBO algorithms [3, 9, 10, 19] and recent analysis of
+PSO [14, 20] we are able to prove convergence of the algorithm under mild assumption on the
+objective function. This is done by first approximating the algorithm with a continuous-in-
+time dynamics and secondly by giving a probabilistic description to the particles system. By
+assuming propagation of chaos [41], particles are considered to behave independently according
+to the same law. This allows to reduce the possible large system of equations to a single partial
+differential equation: the so-called mean-field model. Such model is then analyzed to recover
+convergence guarantees under precise assumption on the objective function. Developed in the
+field of statistical physics, this approach has shown be fruitful in studying particle-based meta-
+heuristic algorithms [9,10,20]. We note that convergence in mean-field law was recently proved
+in [37] in an independent work.
+The rest of the paper is organized as follows.
+Section 2 is devoted to the introduction
+of the CBO-ME algorithm with random selection and comparison with CBO methods without
+memory effects and PSO. In Section 3 validate the proposed methods against several benchmark
+problems and two Machine Learning tasks. Theoretical convergence guarantees and analysis of
+the random selection strategy are summarized in Section 4. Some final remarks are given in
+Section 5. Technical details of the theoretical analysis are given in Appendix A.
+3
+
+2
+Consensus-based optimization with memory effects
+In this section, we present the Consensus-Based Optimization algorithm with Memory Effects
+(CBO-ME) to solve problems of the form
+x∗ ∈ argmin
+x∈Rd F(x) ,
+(2.1)
+where Rd, d ∈ N is the, possibly large, search domain for the continuous function F ∈ C(Rd, R).
+We will do so by highlighting similarities and differences between classical CBO methods and
+PSO algorithms.
+2.1
+Particles update rule
+At each iteration step k and for every particle i = 1, . . . , N, we store its position xk
+i and its best
+position found so far yk
+i = argmin{F(xk
+1), . . . , F(xk
+N)}. The best positions are used to compute
+a consensus point
+¯yα,k =
+N
+�
+i=1
+ωk
+i yk
+i
+with
+ωk
+i =
+e−αF(yk
+i )
+�N
+j=1 e−αF(yk
+j )
+(2.2)
+which approximate the global best solution ¯yα,k among all particles and all times for α > 1.
+Indeed, thanks to the choice of the weights ωk
+i , we have that
+¯yα,k
+−→
+¯y∞,k := argmin{F(yk
+1), . . . , F(yk
+N)}
+(2.3)
+as α → ∞, provided that there is only one global best position among {yk
+1, . . . , yk
+N}. Such ap-
+proximation was first introduced for CBO methods [34] as it leads to more amenable theoretical
+analysis, but it also allows more flexibility. Indeed, relatively small values of α are typically
+used at the beginning of the computation to promote exploration. Large values of α, on the
+other hand, lead to better exploitation of the computed solutions and to higher accuracy. We
+note that the weights used in (2.2) correspond in statistical mechanics to the Boltzmann-Gibbs
+distribution associated with the energy F. In this context, α plays the role of the inverse of the
+system temperature T and the limit α → ∞ corresponds to T → 0.
+Once the consensus point ¯yα,k is computed, the particle positions are then updated according
+to the law
+xk+1
+i
+= xk
+i + λ
+�
+¯yα,k − xk
+i
+�
++ σ
+�
+¯yα,k − xk
+i
+�
+⊗ θk
+i
+(2.4)
+with θk
+i ∈ Rd randomly sampled from the normal distribution (θk
+i ∼ N(0, Id)) and where ⊗ is
+the component-wise product.
+The update rule is characterized by a deterministic component of strength λ ∈ (0, 1) promot-
+ing concentration around the consensus point ¯yα,k and a stochastic component of strength σ > 0
+promoting exploration of the search space. As the latter depends on the difference (¯yα,k − xk
+i ),
+the random behavior is stronger for particles which are far form the consensus point, whereas
+it is weaker for those that are close to it. Also, such exploration resemble an anisotropic diffu-
+sive behavior exploring every coordinate direction at a different rate. This approach was first
+proposed in [4] in the context of CBO methods and has been proved to suffer less from the
+curse of dimensionality with the respect to the originally proposed isotropic diffusion given by
+σ∥¯yα,k − xk
+i ∥2θk
+i with θk
+i being again a normally distributed d-dimensional vector [4].
+4
+
+2.2
+Random selection strategy
+When the particle system concentrates around the consensus point, showing a mostly exploita-
+tive behavior, we employ a particle selection strategy. Discarding particles introduces additional
+stochasticity to the system, while reducing the computational cost. Following the approach sug-
+gested in [7], we check the evolution of the system variance to decide how many particles to
+(eventually) discard.
+For a given set of particles z = {zi}i∈J, the system variance is given by
+var(z) := 1
+|J|
+�
+j∈J
+∥zj − m(z)∥2
+2
+with
+m(z) := 1
+|J|
+�
+i∈J
+zi ,
+(2.5)
+where |J| indicates the cardinality of I, that is, the number of particles in this context.
+Let Ik ⊆ {1, . . . , N} be the set of active particles at step k and Nk = |Ik|.
+To decide
+how many particles to select, we compare the variance of the particle system before the position
+update (2.4), xk = {xk}i∈Ik and after it, ˜xk+1 = {xk+1
+i
+}i∈Ik. Then, the number Nk+1 of particles
+we select for the next iteration is given by
+˜Nk+1 =
+�
+Nk
+�
+1 + µ var(˜xk+1) − var(xk+1)
+var(xk+1)
+��
+Nk+1 = min
+�
+max
+� ˜Nk+1, Nmin
+�
+, Nk
+�
+(2.6)
+⌊z⌋ being the integer part of a number z and Nmin ∈ N the smallest amount of particles we allow
+to have. Then, a subset Ik+1 ⊂ Ik, |Ik+1| = Nk+1, of particles is randomly selected to continue
+the computation. The parameter µ ∈ [0, 1] regulates the mechanism: for µ = 0 there is no
+particle discarding, while for µ = 1 the maximum number of particles is discarded if the variance
+is decreasing. As we will see in Section 3, this random selection mechanism dramatically reduces
+the computational time without affecting the algorithm performance. We will also theoretically
+analyze this aspect in Section 4.3, where we show that convergence properties are preserved.
+As stopping criterion, we keep a counter n on how many times ∥¯yα,k+1 − ¯yα,k∥2 is smaller
+than a certain tolerance δstall. If this happens for more than a given nstall number of times in a
+row, we assume the particles system found a solution and stop the computation. A maximum
+number of iteration kmax representing the computational budget is also given. The proposed
+CBO-ME is summarized in Algorithm 1.
+Remark 2.1. In the meta-heuristic literature, particles are usually discarded depending on their
+objective value, in a way that particles with high values are more likely to be discarded [27,40].
+The proposed strategy does not add a further heuristic strategy but simply cut down the algorithm
+complexity. Also, the convergence properties are in this way expected to be preserved. We note
+that, on the other hand, there is no straightforward way to generate particles and, at the same
+time, preserve the particle system distribution.
+2.3
+Comparison with CBO and PSO
+What distinguishes CBO-ME from plain CBO, see e.g [4,34], is clearly the introduction of the
+best positions {yk
+i }N
+i=1 and the fact that the consensus point is calculated among them and not
+5
+
+Algorithm 1: Consensus-Based Optimization with Memory Effects (CBO-ME)
+Input: F, N0, kmax, λ, σ, α, nstall and δstall;
+1 Inizialize N0 particle positions xi
+0, i = 1, . . . , N;
+2 y0
+i ← x0
+i for all i = 1, . . . , Nk;
+3 Compute yα,0 according to (2.2);
+4 k ← 0, n ← 0;
+5 while k < kmax and n < nstall do
+6
+for i = 1 to Nk do
+7
+θk
+i ∼ N(0, Id);
+8
+Compute xk+1
+i
+according to (2.4);
+9
+if F(xk+1
+i
+) < F(yk
+i ) then
+10
+yk+1
+i
+← xk+1
+i
+;
+11
+else
+12
+yk+1
+i
+← yk
+i ;
+13
+end
+14
+end
+15
+Compute ¯yα,k+1 according to (2.2);
+16
+if ∥¯yα,k+1 − ¯yα,k∥2 < δstall then
+17
+n ← n + 1;
+18
+else
+19
+n ← 0;
+20
+end
+21
+Compute Nk+1 according to (2.6);
+22
+if Nk+1 < Nk then
+23
+Randomly discard Nk+1 − Nk particles;
+24
+k ← k + 1;
+25 end
+26 return ¯yα,k, F(¯yα,k)
+just among the particle positions {xk
+i }N
+i=1 at that given time k. Indeed, the classical CBO update
+rule without memory effects (and with anisotropic diffusion and projection step) is given by
+xk+1
+i
+= xk
+i + λ
+�
+¯xα,k − xk
+i
+�
++ σ
+�
+¯xα,k − xk
+i
+�
+⊗ θk
+i
+(2.7)
+where ¯xα,k is defined consistently with (2.2) (by substituting yk
+i with xk
+i ). As we will see in the
+numerical tests, the use of memory effects improves the algorithm performance.
+Since alignment towards personal bests yk
+i and towards the global best ¯y∞,k are also the
+fundamental building blocks of PSO algorithms, we highlight now the main differences and
+similarities between PSO and CBO-ME. For completeness, we recall the canonical PSO method,
+see e.g. [36], using the notation of (2.4) for easier comparison
+�
+xk+1
+i
+= xk
+i + vk+1
+i
+vk+1
+i
+= wvk
+i + C1
+�
+yk
+i − xk
+i
+�
+⊗ ˆθk
+i,1 + C2
+�
+¯y∞,k − xk
+i
+�
+⊗ ˆθk
+i,2
+(2.8)
+6
+
+where vk
+i are the particles velocities, w, C1, C2 > 0 are the algorithm parameters and θk
+i,1, θk
+i,2
+are uniformly sampled from [0, 1]d (ˆθk
+i,1, ˆθk
+i,2) ∼ Unif([0, 1]d). Several variants and improvements
+have been proposed starting from the above dynamics, but a complete review is beyond the
+scope of this paper and we refer to the recent survey [47] for more references.
+We are interested in highlighting the main differences between (2.4) and (2.8) regarding
+the stochastic components: in CBO-ME deterministic and stochastic steps are de-coupled and
+tuned by two different parameters (λ and σ), while in PSO they are coupled. Indeed, in (2.8),
+deterministic and stochastic components are both controlled by the same parameter: C1 in
+the case of personal best dynamics and C2 for the global best one.
+By splitting the term
+C2
+�
+¯y∞,k − xk
+i
+� ˆθk
+i,2 into a deterministic step and a zero-mean term we obtain
+C2
+�
+¯y∞,k − xk
+i
+�
+⊗ ˆθk
+i,2 = C2
+2
+�
+¯y∞,k − xk
+i
+�
++ C2
+2
+�
+¯y∞,k − xk
+i
+�
+⊗ θk
+i,2
+(2.9)
+with θk
+i,2 = 2ˆθk
+i,2 − 1, θk
+i,2 ∼ Unif([−1, 1]d). Suggested in [14], such rewriting highlights how
+increasing the alignment strength towards the global best (by increasing C2) necessary increases
+the stochasticity of the system as well. In (2.4) and (2.7), on the other hand, one is allowed to
+tune the exploration and exploitation behaviors separately, by either changing parameters λ or
+σ.
+Clearly, CBO-ME also differs from PSO due to its first-order dynamics. Having the aim of
+resembling birds flocking, the first PSO algorithm [24] was proposed as a second-order dynamics.
+The inertia weight w, introduced later in [39], became an essential parameter to prevent early
+convergence of the swarm and to increase the global exploration behavior, especially at the
+beginning of the computation, see e.g. [31, 39] and reviews [18, 36, 47] for more references. We
+note that several other strategies have proposed to improve PSO exploration behavior, see,
+for example, [50]. As already mentioned, in CBO methods convergence and exploration are
+de-coupled and can be tuned separately. Therefore, to keep the algorithm more amenable to
+theoretical analysis, we consider a simpler first-order dynamics. We note that a CBO dynamics
+with inertia mechanism was proposed in [5].
+Similarly, we found the contribution given by the personal best alignment non-essential and
+difficult to tune. Thus, the lack of alignment towards personal best in (2.4). Replacing alignment
+towards personal best with gaussian noise was also suggested in [48] where authors proposed the
+Accelerated PSO (APSO) algorithm. Further studied in [11,49], APSO also allows to de-couple
+the stochastic component from the deterministic one and the noise is heuristically tuned to
+decrease during the computation (as in Simulated Annealing [25]). In CBO methods, the noise
+strength automatically adapts as it depends on the distance from the consensus point, which
+is also different for every particle. For completeness, we note that many other variants of PSO
+have been proposed to include the explorative behavior, see e.g. Chaotic PSO [30].
+3
+Numerical results
+Having discussed the fundamental features of the CBO dynamics with memory effects, we now
+validate Algorithm 1 and compare its performance with plain CBO and PSO algorithms. We will
+test the methods against several benchmark optimization problems and analyze the impact of the
+7
+
+Name
+Objective function F(x)
+Search space
+x∗
+F(x∗)
+Ackley
+−20 exp
+�
+−0.2
+�
+1
+d
+�d
+i=1 (xi)2
+�
+− exp
+�
+1
+d
+�d
+i=1 cos (2π(xi))
+�
++ 20 + e
+[−32, 32]d
+(0, . . . , 0)
+0
+Griewank
+1 + �d
+i=1
+(xi)2
+4000 − �d
+i=1 cos
+� xi
+i
+�
+[−600, 600]d
+(0, . . . , 0)
+0
+Rastrigin
+10d + �d
+i=1
+�
+(xi)2 − 10 cos (2π(xi))
+�
+[−5.12, 5.12]d
+(0, . . . , 0)
+0
+Rosenbrock
+1 − cos
+�
+2π
+��d
+i=1 (xi)2
+�
++ 0.1
+��d
+i=1 (xi)2
+[−5, 10]d
+(1, . . . , 1)
+0
+Salomon
+1 − cos
+�
+2π
+��d
+i=1 (xi)2
+�
++ 0.1
+��d
+i=1 (xi)2
+[−100, 100]d
+(0, . . . , 0)
+0
+Schwefel 2.20
+�d
+i=1 |xi|
+[−100, 100]d
+(0, . . . , 0)
+0
+XSY random
+�d
+i=1 ηi|xi|i,
+ηi ∼ Unif([0, 1])
+[−5, 5]d
+(0, . . . , 0)
+0
+XSY 4
+��d
+i=1 sin2(xi) − e − �d
+i=1(xi)2�
+e − �d
+i=1 sin2 √
+|xi|
+[−10, 10]d
+(0, . . . , 0)
+−1
+Table 1: Considered benchmark test functions for global optimization. For each function,
+the corresponding search space and global solution is given.
+random selection technique on the convergence speed. We also employ 1 to solve problems arising
+form applications, such as image segmentation and training of a machine learning architectures
+for function approximation and image recognition.
+3.1
+Tests on benchmark problems
+We test the proposed algorithm against different optimization problems, by considering 8 bench-
+mark objective functions, see e.g. [22], which we report in Table 1 for completeness. The search
+space dimension is set to d = 20.
+As in plain CBO methods, we expect the most important parameters are those governing
+the balance between the exploitative behavior (λ in this case) and the explorative one (σ). In
+particular, we are interested in the algorithm performance as we change the ratio between λ
+and σ. Therefore, in the first experiment we fix λ = 0.01, while considering different values of
+σ. The parameter α is adapted during the computation: starting form α0 = 10, it increases
+according to the law
+α = α0 · k · log2(k) .
+(3.1)
+Fig. 1 shows the accuracy and the objective value reached for σ ∈ [0, 2] after kmax = 104
+algorithm iterations with N = 200 particles, with no random selection. The optimal value for σ
+is clearly problem-dependent, but we note that the optimal values for the problems considered
+all fall within a relative small range (underlined in gray in Fig. 1).
+From Fig.
+1 we infer that a good value for all benchmark problems considered is given
+by σ = 0.8. Using this value, we now compare CBO-ME, with plain CBO and the standard
+PSO (with and without alignment towards personal best) for different population sizes N =
+50, 100, 200. We keep the random selection mechanism off by setting µ = 0 and use the same
+8
+
+0
+0.5
+1
+1.5
+2
+<
+10!10
+10!5
+100
+105
+1010
+ky,;k ! x$k1
+Rastrigin
+Ackley
+Griewank
+Rosenbrock
+Salomon
+Schwefel
+XSY 4
+XSY random
+(a) ∥¯yα,k − x∗∥∞
+0
+0.5
+1
+1.5
+2
+<
+10!10
+10!5
+100
+105
+1010
+F(7y,;k)
+Rastrigin
+Ackley
+Griewank
+Rosenbrock
+Salomon
+Schwefel
+XSY 4
+XSY random
+(b) F(¯yα,k)
+Figure 1: Optimization on benchmark functions using CBO-ME. Behavior of the expec-
+tation error and fitness value for different values of σ. Here λ = 0.01 and α is adaptive,
+with α0 = 10. The particle population is N = 200. Grey bands (of values [0.70, 1.05] for
+the error and [0.65, 1] for the fitness) show the range in which the minima of the different
+benchmark functions fall. The dotted line marks the visually estimate pseudo-optimal value
+σ = 0.8. Results averaged on 250 runs, are obtained with kmax = 104 iterations and without
+stopping criterion.
+previously chosen parameters when memory effects are used. For plain CBO, without memory
+effects, we set σ = 0.71 ≈
+√
+2/2. Concerning PSO, we use the solver provided by the MATLAB
+Global Optimisation Toolbox (particleswarm), changing the maximum number of iterations
+and the stall condition to the one used for CBO methods, to make the results comparable.
+The remaining parameters are kept as described in the relative documentation [33]. We set
+kmax = 104, δstall = 10−4 and consider a run successful when either
+∥¯yα,k − x∗∥∞ < 0.1
+or
+|F(¯yα,k) − F(x∗)| < 0.01 .
+(3.2)
+Table 2 reports success rate, final error given by ∥¯yα,k −x∗∥∞, mean objective function value
+and total number of iterations, averaged over 250 runs. In addition to the classic PSO method,
+where the acceleration coefficients are chosen to be equal C1 = C2 = 1.49, Table 2 also shows
+the results when only the alignment towards global best is considered in PSO (C1 = 0).
+While CBO already manages to find the global minimizer in most of the problems considered,
+we note that it fails when Rastrigin, Rosenbrock or XSY random functions are optimized. CBO-
+ME, on the other hand, is able to solve the optimization problem correctly even in these cases
+if the population size N is large enough. CBO seems to achieve greater accuracy in some cases,
+such as with Schwefel 2.20 and Salomon objectives, at the cost of more iterations. Standard
+PSO in many cases fails to solve the problem, see e.g. Rastrigin, Salomon or XSY 4 functions.
+PSO success rate is also lower among all problems, with the exception of the Schwefel 2.20
+benchmark problem. Considering only global adjustment seems to show advantages with respect
+to the classical PSO method, except in the case of Ackley where setting C1 = 0 decreases the
+success rate or, in the case of XSY 4, Salomon or Rastrigin, where convergence is not achieved
+even for C1 = 0 Consensus methods, however, seem to perform better in terms of both success
+9
+
+(a) Error: ∥¯yα,k − x∗∥∞
+(b) Fitness Value: F(¯yα,k)
+Figure 2: Optimization of Ackley function for different values of the random selection
+parameter µ, where the initial particle population is N 0 = 104. We report error (on the
+left) and fitness values (on the right) as the number of function evaluations increases.
+Parameters are set as λ = 0.01, σ = 0.8, α adaptive starting from α0 = 10 and following
+the law α = α0 · k · log2(k). Results are averaged over 250 runs.
+rate and speed up. In addition, for most problems, the population size N seems not to play a
+significant role in the algorithms performance. This further motivates the introduction of the
+random selection strategy described in the Section 2.1 in order to save computational costs.
+In the third experiment, we test the proposed random selection mechanism (2.6) for different
+values of the parameter µ. We recall that with µ = 0 we have no particles removal, while as
+µ increases, more particles are likely to be discarded when the system variance decreases. The
+initial population is set to N0 = 200, while the minimum number of particles to Nmin = 10.
+Results are reported in Tables 3 and 4 in terms of: success rate, error, objective value, weighted
+number of iterations, given by
+witer =
+kend
+�
+k=1
+Nk
+N0
+(3.3)
+and percentage of Computational Time Saved (CTS). Results show that relative large values
+of µ allow to reach fast convergence without affecting the algorithm performance. The values of
+µ considered in Table 4 as different from those in Table 3 as in our experiments, the Rastrigin
+problem allows for larger values of µ, while the Rosenbrock one seems to be more sensitive to
+the selection mechanism with respect to the other objectives. In both cases, a suitable value of
+µ reduces the computational time with almost no impact in terms of accuracy.
+Fig.s 2 and 3 show error and fitness value as a function of the number of fitness evaluation
+during the algorithm computation, for the Ackley and Rastrigin problem respectively. Several
+values of µ are considered to display how the random selection mechanism affects the convergence
+speed.
+Initial particle population is set to N0 = 104 and particles evolve for kmax = 104
+iterations. We note how convergence speed increases as µ increases.
+10
+
+CBO (σ =
+√
+2/2)
+CBO-ME (σ = 0.8)
+PSO
+PSO (C1 = 0)
+N = 50
+N = 100
+N = 200
+N = 50
+N = 100
+N = 200
+N = 50
+N = 100
+N = 200
+N = 50
+N = 100
+N = 200
+Ackley
+Rate
+99.3%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+14.6%
+38.6%
+53.3%
+4.0%
+16.0%
+39.3%
+Error
+4.11e-06
+2.03e-06
+2.55e-06
+2.39e-06
+1.73e-06
+1.74e-06
+6.97e-09
+8.56e-11
+2.16e-12
+2.30e-08
+1.90e-10
+8.96e-13
+Favg
+1.18e-04
+5.81e-05
+7.30e-05
+1.54e-04
+4.96e-05
+4.99e-05
+6.24e-09
+7.65e-11
+1.94e-12
+2.06e-08
+1.70e-10
+8.01e-13
+Iterations
+954.3
+778.2
+678.1
+997.7
+724.3
+626.9
+493.8
+420.6
+391.4
+502.0
+436.0
+390.1
+Griewank
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+46.0%
+48.6%
+55.3%
+50.0%
+58.7%
+78.0%
+Error
+2.20e-02
+2.21e-02
+2.24e-02
+2.13e-02
+2.16e-02
+2.25e-02
+7.34e-02
+1.56e-02
+9.45e-03
+1.17e-01
+1.10e-01
+8.96e-02
+Favg
+5.26e-02
+5.31e-02
+5.47e-02
+4.95e-02
+5.15e-02
+5.82e-02
+3.23e-03
+4.11e-03
+3.78-03
+3.73e-03
+3.71e-03
+2.90e-03
+Iterations
+927.5
+777.0
+682.7
+891.9
+735.0
+635.4
+436.0
+394.5
+374.5
+427.2
+370.2
+345.5
+Rastrigin
+Rate
+9.3%
+27.3%
+60.7%
+26.0%
+68.7%
+89.3%
+0.0%
+0.0%
+0.0%
+0.0%
+0.0%
+0.0%
+Error
+1.28e-04
+1.83e-04
+2.34e-04
+9.73e-05
+1.27e-04
+1.76e-04
+-
+-
+-
+-
+-
+-
+Favg
+4.51e-06
+9.03e-06
+1.46e-05
+2.54e-06
+4.31e-06
+8.28e-06
+-
+-
+-
+-
+-
+-
+Iterations
+1083.0
+933.7
+819.8
+1007.6
+922.5
+769.9
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+Rosenbrock
+Rate
+65.3%
+86.7%
+97.3%
+72.7%
+98.0%
+100.0%
+9.3%
+22.6%
+36.6%
+46.7%
+60.7%
+76.7%
+Error
+1.81e-02
+2.44e-02
+1.48e-02
+3.62e-02
+4.04e-02
+1.78e-02
+6.19e-04
+2.56e-04
+1.67e-04
+4.44e-02
+4.45e-02
+4.46e-02
+Favg
+6.13e-03
+7.57e-03
+2.40e-03
+1.26e-02
+1.42e-02
+2.65e-03
+3.80e-02
+3.76e-02
+2.56e-02
+2.56e-03
+8.95e-04
+3.71e-04
+Iterations
+5772.0
+5440.3
+5439.2
+5955.7
+4977.3
+4275.9
+4830.0
+3322.4
+2892.7
+5886.4
+3419.1
+2164.2
+Schwefel 2.20
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+5.79e-06
+8.23e-07
+2.44e-07
+8.42e-06
+1.03e-06
+2.76e-07
+8.34e-10
+1.97e-12
+4.58e-14
+1.68e-07
+3.41e-10
+8.03e-14
+Favg
+1.04e-03
+2.15e-04
+8.36e-05
+1.50e-03
+3.12e-04
+9.37e-05
+1.94e-09
+6.36e-12
+1.52e-13
+2.44e-07
+6.48e-10
+2.46e-13
+Iterations
+814.7
+691.5
+619.2
+670.8
+547.2
+467.7
+484.3
+428.0
+401.7
+593.3
+457.8
+410.3
+Salomon
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+100.0%
+0.0%
+0.0%
+0.0%
+0.0%
+0.0%
+0.0%
+Error
+3.12e-02
+2.14e-02
+1.87e-02
+5.28e-02
+4.49e-02
+3.91e-02
+-
+-
+-
+-
+-
+-
+Favg
+3.14e-01
+2.15e-01
+1.88e-01
+2.44e-01
+1.86e-01
+1.91e-01
+-
+-
+-
+-
+-
+-
+Iterations
+10000.0
+10000.0
+10000.0
+8886.4
+9296.2
+2456.5
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+XSY random
+Rate
+55.3%
+84.7%
+92.0%
+100.0%
+100.0%
+100.0%
+1.2%
+11.7%
+21.0%
+100.0%
+100.0%
+100.0%
+Error
+2.64e-02
+1.62e-02
+9.80e-03
+3.06e-02
+1.86e-02
+1.15e-02
+2.25e-01
+9.56e-02
+8.42e-02
+6.23e-02
+5.12e-02
+2.34e-02
+Favg
+6.95e-08
+3.54e-08
+2.13e-08
+2.21e-06
+4.85e-08
+3.17e-08
+3.35e-04
+2.28e-04
+1.34e-04
+8.22e-04
+4.11e-04
+3.45e-04
+Iterations
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+XSY 4
+Rate
+22.0%
+87.3%
+98.7%
+23.3%
+86.7%
+100.0%
+0.0%
+0.0%
+0.0%
+0.0%
+0.0%
+0.0%
+Error
+8.07e-01
+7.48e-01
+7.16e-01
+8.44e-01
+7.35e-01
+6.95e-01
+-
+-
+-
+-
+-
+-
+Favg
+4.79e-07
+3.78e-07
+3.46e-07
+1.58e-06
+8.56e-07
+5.43e-07
+-
+-
+-
+-
+-
+-
+Iterations
+10000.0
+10000.0
+10000.0
+9677.5
+9128.4
+8943.2
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+10000.0
+Table 2: Comparison between classical CBO, CBO-ME and standard PSO with and with-
+out alignment towards personal best on benchmark problems. The solver particleswarm
+available in the MATLAB Global Optimisation Toolbox was used for the results concerning
+the PSO method. Optimal choice of parameters, different for each method, are used for the
+CBO algorithms. Same stopping criterion and definition of success, see (3.2), were used.
+Performance metric considered: success rate (see (3.2)), error (∥¯yα,k−x∗∥∞), fitness value
+F(¯yα,k) and number of iterations. Results are averaged over 250 runs.
+11
+
+µ = 0
+µ = 0.05
+µ = 0.1
+µ = 0.2
+Ackley
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+1.84e-06
+2.16e-06
+6.54e-06
+1.34e-05
+Favg
+7.30e-05
+6.17e-05
+1.87e-04
+3.95e-04
+witer
+674.2
+505.2
+357.2
+182.1
+CTS
+-
+31.1%
+51.6 %
+73.8%
+Griewank
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+2.35e-02
+2.22e-02
+2.32e-02
+2.28e-02
+Favg
+5.82e-02
+5.20e-02
+5.72e-02
+5.70e-02
+witer
+635.4
+395.9
+204.6
+184.6
+CTS
+-
+31.8%
+58.2%
+73.3%
+Schwefel 2.20
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+2.76e-07
+9.08e-07
+8.21e-07
+2.73e-08
+Favg
+9.37e-05
+2.93e-05
+1.58e-05
+3.74e-05
+witer
+467.7
+359.8
+318.7
+172.4
+CTS
+-
+24.4%
+32.9%
+64.1%
+Salomon
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+4.11e-02
+3.35e-02
+2.74e-02
+1.75e-02
+Favg
+4.34e-01
+4.43e-01
+4.07e-01
+3.26e-01
+witer
+2456.5
+1595.8
+1289.2
+913.0
+CTS
+-
+36.7%
+49.1%
+63.7%
+XSY random
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+1.50e-02
+8.62e-02
+8.89e-02
+9.08e-02
+Favg
+5.97e-07
+1.75e-05
+5.48e-05
+1.06e-04
+witer
+10000.0
+2642.3
+1755.7
+1123.7
+CTS
+-
+73.6%
+82.4%
+88.7%
+XSY 4
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+5.30e-01
+3.78e-01
+1.35e-01
+1.37e-01
+Favg
+1.17e-05
+6.28e-06
+3.41e-06
+3.55e-06
+witer
+8943.2
+3910.4
+1890.2
+1060.1
+CTS
+-
+46.9%
+68.1%
+79.4%
+Table 3: CBO-ME algorithm with random selection of particles tested against different
+benchmark functions with different values of µ, which regulates the random selection mech-
+anism. The system is initialized with N0 = 200 particles and σ = 0.8. Performance metric
+considered: success rate (see (3.2)), error (∥¯yα,k − x∗∥∞), fitness value F(¯yα,k), weighted
+iteration (3.3), and Computational Time Saved (CTS). Results are averaged over 250 runs.
+µ = 0
+µ = 0.1
+µ = 0.2
+µ = 0.5
+Rastrigin
+Rate
+100.0%
+100.0%
+100.0%
+100.0%
+Error
+9.14e-05
+7.12e-05
+3.77e-05
+1.24e-05
+Favg
+2.23e-06
+2.19e-06
+1.98e-06
+1.27e-06
+witer
+1161.1
+719.6
+256.5
+111.2
+CTS
+-
+38.1%
+77.9%
+90.4%
+µ = 0
+µ = 0.01
+µ = 0.02
+µ = 0.05
+Rosenbrock
+Rate
+100.0%
+100.0%
+99.4%
+99.0%
+Error
+2.55e-02
+2.23e-02
+1.66e-02
+1.341e-02
+Favg
+4.20e-03
+5.23e-03
+4.10e-03
+4.24e-03
+witer
+3172.3
+852.9
+347.8
+82.5
+CTS
+-
+73.1%
+89.1%
+97.4%
+Table 4: CBO-ME algorithm with particle reduction tested against Rastrigin and Rosen-
+brock functions with an higher diffusion parameter σ = 1.1 and for different values of µ ,
+which regulates the random selection mechanism. The system is initialized with N0 = 200
+particles. Performance metric considered: success rate (see (3.2)), error (∥¯yα,k − x∗∥∞),
+fitness value F(¯yα,k), weighted iteration (3.3), and Computational Time Saved (CTS).
+12
+
+(a) Error: ∥¯yα,k − x∗∥∞
+(b) Fitness Value: F(¯yα,k)
+Figure 3: Optimization of Rastigin function for different values of the random selection
+parameter µ where the initial particle population is N0 = 104. We report error (on the
+left) and fitness values (on the right) as the number of function evaluations increases.
+Parameters are set as λ = 0.01, σ = 1.1, α adaptive starting from α0 = 10 and following
+the law α = α0 · k · log2(k). Results are averaged over 250 runs.
+3.2
+Applications
+In this section, we propose some applications of the proposed optimization algorithm. First
+we consider a image segmentation problem using multi-thresholding, then we use the CBO-
+ME to train a Neural Network (NN) architecture to approximate functions and perform image
+classification on MNIST database of handwritten digits.
+3.2.1
+Image segmentation
+To perform image segmentation, we use a threshold detection technique, namely, the multidimen-
+sional Otsu algorithm [32,44] in order to compare the results to similar optimization algorithm,
+such as the Modified PSO in [43].
+In the Otsu algorithm, every pixel of the image is assigned to one of the possible L grayscale
+values. We denote with ηi the number of pixel with gray level i, 1 ≤ i ≤ L and Npix = �L
+i=1 ηi
+the total number of pixels [32]. Then, the image is divided into object C0 with gray-level [1, . . . , l]
+and background C1 with gray-level [l + 1, . . . , L] by inserting a threshold l. The probabilities of
+class occurrence and the class mean level for the object, respectively, are given by
+ω0(l) =
+l
+�
+i=1
+pi,
+pi =
+ηi
+Npix
+µ0(l) =
+l
+�
+i=1
+ipi
+ω0(k) .
+For the background, the class occurrence probabilities and the class mean level are given by
+ω1(l) =
+L
+�
+i=l+1
+pi,
+pi =
+ηi
+Npix
+13
+
+µ1(l) =
+L
+�
+i=l+1
+ipi
+ω1(k) .
+As in [32], the best threshold l∗ is obtained when the variance formula
+f(l) = ω0(l) ω1(l) (µ0(l) − µ1(l))2
+(3.4)
+between object group and background reaches its maximum value, i.e. l∗ = argmaxlf(l). The
+problem is then reduced to a threshold problem, which we can solve with optimization methods.
+Since segmentation is a trivial one-dimensional problem, we consider an extension of Otsu’s
+technique to the multidimensional case [44] to test capabilities of method. Assuming we want to
+optimize the choice of d thresholds, we require d + 1 classes of different gray-scales (C0, . . . , Cd)
+with relative probabilities of occurrence classes defined as
+ω0(l1) =
+l1
+�
+i=1
+pi , . . . , ωd(ld) =
+L
+�
+i=ld+1
+pi,
+pi =
+ηi
+Npix
+and classes mean levels
+µ0(l1) =
+�l1
+i=1 ipi
+ω0
+, . . . , µd(ld) =
+�L
+i=ld+1 ipi
+ωd
+,
+The optimal thresholds (ˆl1, . . . , ˆld) are those that satisfy ˆl1 < · · · < ˆld and maximise
+f(l1, . . . ld) =
+d
+�
+i=1
+ωi(li)µ2
+i (li)
+(3.5)
+For the experiment, we chose d = 5 thresholds and compare the segmentation performed by
+Otsu’s method, solved with both standard PSO and CBO-ME, with segmentation obtained by
+dividing the greyscale into d + 1 uniformly spaced intervals. For PSO, we use to the default
+parameters in the particleswarm function in the MATLAB Global Optimisation Toolbox, while
+for CBO-ME we used optimal parameters found in Section 3.1 and exploit the random selection
+technique to speed up the algorithm.
+We report the results on two sample images, Fig.s 4 and 5. We fix kmax = 103 and average
+results over 250 runs. As in [2], we evaluate multi-thresholding segmentation through the Peak
+Signal to Noise Ratio (PSNR) computed as:
+PSNR = 20 · log10
+�
+255
+RMSE
+�
+where RMSE is the Root Mean-Squared Error, defined as
+RMSE =
+�
+�
+�
+�
+1
+Npix
+Nrow
+�
+i=1
+Ncol
+�
+j=1
+[I(i, j) − S(i, j)]2
+where Npix = Nrow · Ncol, I is the original image and S is the associated segmented image.
+The higher the value of PSNR is, the greater the similarity between the clustered image and
+the original image is. From Fig.s 4,5, we note that the most accurate segmentation on details
+is obtained by the CBO-ME method. This is quantitatively confirmed by the PSNR values
+reported in Table 5.
+14
+
+(a) Original
+(b) Standard segmentation
+(c) Otsu seg. (PSO)
+(d) Otsu seg. (CBO-ME)
+Figure 4: Image segmentation of darkhair woman image (256 × 256 pixels) with standard
+segmentation and Otsu segmentation solved respectively by PSO (c) and by CBO-ME (d);
+results are averaged over 250 runs, with an initial population of N0 = 103 particles.
+(a) Original
+(b) Standard segmentation
+(c) Otsu seg. (PSO)
+(d) Otsu seg. (CBO-ME)
+Figure 5: Image segmentation of lake image (256×256 pixels) with standard segmentation
+and Otsu segmentation solved respectively by PSO (c) and by CBO-ME (d); results are
+averaged over 250 runs, with an initial population of N0 = 103 particles.
+cameraman
+lake
+lena
+peppers
+woman darkhair
+Standard segmentation
+22.83
+21.72
+24.35
+27.24
+25.33
+Otsu segmentation
+(PSO)
+34.62
+32.33
+38.19
+38.03
+37.14
+Otsu segmentation
+(CBO-ME)
+37.22
+35.44
+38.72
+38.28
+39.57
+Table 5: PSNR values to evaluating the advantages of the method in optimising threshold
+values in 5 sample images known in literature. For these results, we compared the Otsu
+segmentation solved by the proposed CBO-ME method with the classical PSO method with
+equispaced thresholding segmentation. Experiments are performed with d = 5 thresholds.
+15
+
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(a) 2000 epochs
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(b) 3000 epochs
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(c) 5000 epochs
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(d) 8000 epochs
+Figure 6: Approximating smooth function u1 (3.8) using a network with n = 50 and m = 3.
+The learning rate is λ = 0.2 and we initially use N0 = 500 particles. The others parameters
+are set as λ = 1, σ = 0.8 and α adaptive starting from α0 = 10.
+3.2.2
+Approximating functions with NN
+In this section, we use the proposed CBO-ME algorithm to train a NN architecture into approx-
+imating a function u : I → R, I ⊂ R with low regularity. As in [5], we use a fully-connected NN
+with m layers
+f(x; θ) = (Lm ◦ . . . L2 ◦ L1)(x)
+(3.6)
+where each layer is given by
+Li = σ(W ix + bi)
+with σ(x) = 1/(1+exp(−x)) being the sigmoid function. We use internal layers of dimension n,
+so W 1 ∈ Rn×1, b1 ∈ R, W m ∈ R1×n, bm ∈ Rd and W i ∈ Rn×n for all i = 2, . . . , m − 1. In (3.6),
+all DNN parameters are collected in θ = {W i, bi}m
+i=1.
+As loss function which need to be minimized, we consider the L2-norm between the target
+function u and its NN approximation f(· ; θ)
+F(θ) := ∥f(· ; θ) − u∥L2(I) .
+(3.7)
+Again, similarly to [5], we test the method against the following two functions:
+u1(x) = sin(2πx) + sin(8πx2)
+(3.8)
+16
+
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(a) 2000 epochs
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(b) 3000 epochs
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(c) 5000 epochs
+-1
+-0.5
+0
+0.5
+1
+-2
+-1
+0
+1
+2
+(d) 8000 epochs
+Figure 7:
+Approximating non-smooth u2 (3.9) function using a network with n = 50,
+m = 3. The learning rate is λ = 0.2 and we use initially N0 = 500 particles. The others
+parameters are set as λ = 1, σ = 0.8 and α adaptive starting from α0 = 10.
+u2(x) =
+�
+�
+�
+�
+�
+1
+if x < − 7
+8, − 1
+8 < x < 1
+8, x > 7
+8
+−1
+if
+3
+8 < x < 5
+8, − 5
+8 < x < − 3
+8,
+0
+otherwise .
+(3.9)
+We note that u1 is smooth, while u2 is discontinuous. Parameters of the CBO-ME algorithm
+have been set to λ = 0.01, σ = 0.8, as in the previous sections. Parameter α is adapted during
+the computation as in 3.1 and random selection mechanism is used. We employ m = 3 layers
+with internal dimension n = 50. Results are displayed in Fig.s 6 and 7. We note that smooth
+function u1 is well-approximated already after 5000 epochs, while convergence is slower for the
+discontinuous step function u2.
+3.2.3
+Application on MNIST dataset
+We now employ the proposed algorithm to train a NN architecture to solve a image classification
+tasks. We will consider the MNIST dataset [26] composed of handwritten digits in grayscale with
+28 × 28 pixels. For better comparability with CBO methods without memory effects, we closely
+follow the experiment settings used in the literature [4,10,37], which we summarize below.
+We consider a 1-layer NN where input images x ∈ R28×28 are first vectorized x �→ vec(x) ∈
+R728 and then processed through a fully-connected layer with parameters θ = {W, b}, with
+17
+
+10
+20
+30
+40
+50
+Epochs
+0.4
+0.6
+0.8
+1
+Accuracy on test data
+CBO-ME
+CBO
+10
+20
+30
+40
+50
+Epochs
+0.06
+0.08
+0.1
+0.12
+0.14
+0.16
+0.18
+Loss
+CBO-ME
+CBO
+Figure 8:
+Performance during training of shallow NN (3.10) on image classification
+(MNIST dataset) with CBO-ME optimizer and plain CBO without memory effects [10].
+Training is performed by Algorithm 1 with N = 100 particles and no particle selection.
+Cross-entropy loss function (3.11) and adaptive parameters strategy (3.12) were used in
+the training.
+W ∈ R10×728, b ∈ R10. That is, the network is given by
+fSNN(x; θ) = softmax (ReLU (Wvec(x) + b) ) ,
+(3.10)
+where ReLU(z) = max{z, 0} (component-wise) and softmax(z) = (ez1, . . . , ezn)/(�
+i ezi) are
+the commonly activation functions.
+During the training, batch regularization is performed
+after ReLU is applied in order to speed up convergence. Given a training set {(xm, ℓm)}M
+m=1,
+xm ∈ R28×28, ℓm ∈ {0, 1}10 made of M image-label tuples we train the model by minimizing the
+categorical cross-entropy loss
+F(θ) = 1
+M
+M
+�
+m=1
+�
+−
+10
+�
+i=1
+ℓm
+i log(fi(xm, θ))
+�
+.
+(3.11)
+We employ a population on N = 100 particles throughout the entire computation, initially
+sampled from the standard normal distribution N(0, Id). Following the mini-batch approach
+suggested in [4], the consensus points ¯yα,k is computed only among a random subset of nN = 10
+particles, but all particles are updated at each step. The training data is divided in batches of
+nF = 60 images. The drift parameter is set to λ = 0.01, while σ and α are adapted during the
+computation after each epoch as
+σepoch = σ0/ log2 (epoch + 2)
+αepoch+1 = 2 · αepoch+1
+(3.12)
+starting form σ0 =
+√
+0.04, and α0 = 50,
+Fig.
+8 shows the results in terms of loss function (3.11) over the test data set and the
+accuracy reached in the classification task. While challenging state-of-the-art training methods
+18
+
+is beyond the scope of the experiment, we note how high-dimensional data optimization tasks
+can be solved with as little as N = 100 particles by the proposed method, obtaining results
+comparable with the literature on CBO methods [4, 10, 37]. Also, we remark that parameters
+have not been tuned extensively.
+4
+Theoretical analysis
+A strength of CBO algorithms lays on the possibility of theoretically analyze the particle system
+by relying on a mean-field approximation of the dynamics. We will illustrate in this section how
+to derive such approximation and present the main theoretical result regarding the convergence
+of the particle system towards a solution to (2.1), in case of no selection mechanism. Next,
+we will study the impact of the random selection strategy on the convergence properties of the
+algorithm. Technical details are left to Appendix A.
+4.1
+Mean-field approximation
+First, we note that a simple update rule for the personal bests yk
+i is given by
+yk+1
+i
+= yk
+i + 1
+2
+�
+xk+1
+i
+− yk
+i
+�
+S(xk+1
+i
+, yk
+i ) ,
+with
+S(x, y) = 1 + sign (F(y) − F(x)) .
+(4.1)
+As in [14], we approximate it for β ≫ 1 as
+yk+1
+i
+= yk
+i + ν
+2
+�
+xk+1
+i
+− yk
+i
+�
+Sβ(xk+1
+i
+, yk
+i ) ,
+(4.2)
+with Sβ(x, y) being a continuous approximation of S(x, y) as β → ∞. By choosing ν = 1 we
+get (4.1) with the only difference of having Sβ instead of S. As for ¯yα with respect to ¯y∞, this
+is needed to make the update rule easier to handle mathematically, but it does have an impact
+on the performance for large values of β.
+With the aim of deriving a continuous-in-time reformulation of (2.4) and (4.2), we introduce
+a single parameter ∆t > 0 which controls the step length of all involved update mechanisms.
+By performing the rescaling
+λ ← λ∆t ,
+σ ← σ
+√
+∆t ,
+ν ← ν∆t
+to get the update rules
+�
+xk+1
+i
+= xk
+i + λ∆t
+�
+¯yα,k − xk
+i
+�
++ σ
+√
+∆t
+�
+¯yα,k − xk
+i
+�
+⊗ θk
+i
+yk+1
+i
+= yk
+i + (ν∆t/2)
+�
+xk+1
+i
+− yk
+i
+�
+Sβ(xk+1
+i
+, yk
+i )
+(4.3)
+which differ form the original formulation (2.4), (4.1) only due to the use of Sβ instead of S.
+As already noted in [14], the iterative process (4.3) corresponds to an Euler-Maruyama
+scheme applied to a system of Stochastic Differential Equations (SDEs). Indeed, (4.3) corre-
+sponds to a discretization of the system
+�
+dXi
+t
+= λ
+�
+¯yα(ρN
+t ) − Xi
+t
+�
+dt + σ
+�
+¯yα(ρN
+t ) − Xi
+t
+�
+⊗ dBi
+t
+dY i
+t
+= ν(Xi
+t − Y i
+t )Sβ(Xi
+t, Y i
+t ) dt
+(4.4)
+19
+
+where, for convenience, we underlined above the dependence of the consensus point on the
+empirical distribution ρN
+t = �
+i δY i
+t (δy being the Dirac measure at y ∈ Rd) by using
+¯yα(ρ) :=
+�
+ye−αF(y)dρ(y)
+�
+e−αF(y)dρ(y) ,
+(4.5)
+defined for any Borel probability measure ρ over Rd (ρ ∈ P(Rd)). In this way, we generalized
+the definition introduced in (2.2) to any ρ ∈ P(Rd), provided the above integrals exists. In (4.4),
+the random component of the dynamics is now described by N independent Wiener processes
+(Bi
+t)t>0. As before, we supplement the system with initial conditions Xi
+0 ∼ ρ0, Y i
+0 = Xi
+0 for some
+ρ0 ∈ P(Rd).
+The continuous-in-time description (4.4) already simplifies the analytical analysis of the
+optimization algorithm, but still pays the price of a possible large number O(N) of equations.
+This issue is typically addressed by assuming that for large populations N, the particles become
+indistinguishable from one another and start behaving, in some sense, as a unique system.
+More precisely, let F N(t) ∈ P(R(2d)N) denote the joint probability distribution of N tuples
+(Xi
+t, Y i
+t ). We assume propagation of chaos [41] for large N ≫ 1, that is, we assume that the
+joint probability distribution decomposes as F N(t) = f(t)⊗N for some f(t) ∈ P(R2d). System
+(4.4) becomes independent on the index i and hence every particle moves according to the
+mono-particle process
+d ¯Xt = λ(¯yα(¯ρt) − ¯Xt) dt + σ (¯yα(¯ρt) − ¯Xt) ⊗ d ¯Bt
+d ¯Yt = ν( ¯Xt − ¯Yt)Sβ( ¯Xt, ¯Yt) dt
+(4.6)
+where ¯ρt = Law( ¯Yt).
+Assume ( ¯Xt, ¯Yt) are initially distributed according to f0 = ρ⊗2
+0 , by applying Itˆo formula we
+have that f(t) = Law( ¯Xi
+t, ¯Y i
+t ) satisfies
+∂tf + ∇x · (λ(¯yα(¯ρ) − x)f) + ∇y ·
+�
+ν(x − y)Sβ(x, y)f
+�
+=
+d
+�
+ℓ=1
+∂2
+xℓ
+�
+σ(¯yα(¯ρ) − x)2
+ℓf
+�
+(4.7)
+and initial data limt→0 f(t) = f0 in a weak sense.
+Dynamics (4.6), or, equivalently, (4.7),
+corresponds to the mean-field approximation of the particle system (4.4) as N → ∞. We remark
+that the above derivation has only been possible thanks to the approximations S ≈ Sβ and
+¯y∞ ≈ ¯yα for large α and β. Well-posedness of the system is also granted by such approximations
+(proof details are given in Appendix A.2).
+Proposition 4.1 (well-posedness of (4.6)). There exists a unique process ( ¯X, ¯Y ) ∈ C([0, T], Rd),
+T > 0 satisfying (4.4) with initial conditions ( ¯X0, ¯Y0) with ¯X0 ∼ ρ0 ∈ P4(Rd) and ¯Y0 = ¯X0.
+Being mathematically tractable, we show next that the mean-field dynamics converges to a
+global solution to (2.1) if F, Sβ.
+20
+
+4.2
+Convergence in mean-field law
+We start by enunciating the necessary assumptions to the convergence result.
+Assumption 4.1 (Assumptions on F). The objective function F ∈ C(Rd, R), satisfies:
+A1
+there exists some constant LF > 0 such that
+|F(x) − F(x′)| ≤ LF
+�
+∥x∥2 + ∥x′∥2
+�
+∥x − x′∥2,
+∀ x, x′ ∈ Rd ;
+A2
+there exists uniquely x∗ ∈ Rd solution to (2.1);
+A3
+there exist η, R0 > 0 and γ ∈ (2, ∞) such that
+F(x) − inf F ≥ η ∥x − x∗∥γ
+∞
+∀x ∈ Rd , ∥x − x∗∥∞ ≤ R0
+F(x) − inf F ≥ η Rγ
+0
+∀x ∈ Rd , ∥x − x∗∥∞ > R0 .
+A4
+F is convex in a (possibly small) neighborhood {x ∈ Rd : ∥x − x∗∥∞ ≤ R1} of x∗ for
+some R1 < R0.
+A5
+There exists cg, R2 > 0 such that
+F(x) − inf F ≥ cg∥x − x∗∥2
+2
+∀x ∈ Rd , ∥x − x∗∥2 > R2 .
+Assumption 4.2 (Assumptions on Sβ). The function Sβ ∈ C(R2d, [0, 2]), with β > 0
+A6
+has the following structure
+Sβ(x, y) = 2ψ (β(F(y) − F(x))) ,
+(4.8)
+with ψ ∈ C1(R, [0, 1]) being an increasing function with Lipschitz constant Lψ = 1.
+A7
+The value Sβ(x, y) is positive only when x is strictly better than y in terms of objective
+value F:
+Sβ(x, y)
+�
+≥ 0
+if
+F(x) < F(y)
+= 0
+else .
+Assuming uniqueness of global minimum is a typical assumption for analysis of CBO methods
+[9,10] and it is due to the definition of the consensus point ¯yα (or ¯xα in the case without memory
+mechanism). Indeed, in presence of two global minima, ¯yα may be placed between them, no
+matter how large α is. Assumption A2 ensure to avoid such situations. Furthermore, A3 also
+allows to give quantitative estimates on the difference between the global minimum and eventual
+local minima. In the literature, such property is known as conditioning [12]. Requirements A4
+and A7 ensure that if a personal best yk
+i enters such small neighborhood where F is convex, it
+will not leave it for the rest of the computation. Condition A5 (quadratic growth at infinity) is
+needed for the well-posedness of the mean-field mono-particle process (4.6), see also [3]. For an
+intuition of A3 and A4 we refer to Figure 9, where the Rastrigin function is considered.
+21
+
+x$ ! R0
+x$
+x$ + R0
+0
+20
+40
+objective
+lower bound (A3)
+convex area (A4)
+Figure 9: Assumptions 4.1 illustrated for Rastrigin function. For example, such objective
+function satisfies A3 with η = 1, γ = 1.8, R0 = 1.42 and A4 with R1 = 0.25.
+Theorem 4.1 (Convergence in mean-field law). Assume F satisfies A1–A5, Sβ satisfies A6,
+A7 for some β > 0 fixed. Let ( ¯Xt, ¯Yt)t≥0 be a solution to (4.6) for t ∈ [0, T], with initial data
+¯X0 ∼ ρ0 ∈ P4(Rd), Y0 = X0 such that x∗ ∈ supp(ρ0) .
+Fix an accuracy ε > 0. If 2λ > σ2, there exists a time T ∗ such that the expected ℓ2-error
+satisfies
+E
+�
+∥ ¯XT ∗ − x∗∥2
+2
+�
+≤ ε
+(4.9)
+provided T, α > 0 are large enough.
+We refer to Appendix A for a proof.
+Remark 4.1. The mean-field mono-particle process (4.6) aims to approximate the algorithm
+iterative dynamics (4.3) for small time steps ∆t ≪ 1 and large particle populations N ≫ 1.
+Therefore, convergence of the algorithm dynamics towards the global solution x∗ can be proven
+by coupling Theorem 4.1 with error estimates of such approximation.
+For instance, assuming that all considered dynamics take place on a bounded set D ensures
+that the error introduced by the continuous-in-time particle system will be of order ∆t thanks to
+classical results on Euler-Maruyama schemes [35]. Likewise, considering a bounded dynamics
+allows to prove that the error introduced by the mean-field approximation is of order N−1 (see
+e.g. [8, Theorem 3.1], [9, Proposition 16]). Let {(xk
+i , yk
+i )}N
+i=1 be given by (4.3), {(Xi
+t, Y i
+t )}N
+i=1 be
+a solution (4.4) and {( ¯Xi
+t, ¯Y i
+t )}N
+i=1 be N-copies of a solution to (4.6). Altogether, one obtains
+the following error decomposition for K∆t = T ∗
+E
+�
+1
+N
+N
+�
+i=1
+∥xK
+i − x∗∥2
+2
+�
+≤ C
+�
+E
+�
+1
+N
+N
+�
+i=1
+∥xK
+i − Xi
+T ∗∥2
+2
+�
++ E
+�
+1
+N
+N
+�
+i=1
+∥Xi
+T ∗ − ¯Xi
+T ∗∥2
+2
+�
++ E
+�
+1
+N
+N
+�
+i=1
+∥ ¯Xi
+T ∗ − x∗∥2
+2
+� �
+≤ CEM∆t + CMFAN−1 + ε
+22
+
+where C, CEM, CMFA are positive constant independent on N, ∆t.
+4.3
+Random selection analysis
+In this section, we analytically investigate the impact of randomly discarding particles during
+the computation. We are particularly interested in tracking the distance between a particle
+system {xk
+i , xk
+j }N0
+i=1 evolving according to (4.3) where no particles are discarded, and a second
+system {ˆxk
+i , ˆyk
+i }Ik, |Ik| = Nk where Nk − Nk+1 particles are discarded after update rule (4.3).
+Clearly, we have that Nk+1 ≤ Nk and Ik+1 ⊆ Ik ⊆ I0 = {1, . . . , N0} for all k. Similarly to the
+analysis carried out in [15,16], we restrict to the simpler dynamics where, at every step k, the
+random variables θk
+i and ˆθk
+i used to generate such systems are the same for all particles:
+θk
+i = ˆθk
+j = θk ∼ N(0, Id)
+for all
+i ∈ Ik, j ∈ I0.
+(4.10)
+To compare particle systems with a different number of particles, we rely on their represen-
+tation as empirical probability measures and the notion of 2-Wasserstein distance. For {ˆxk
+i }i∈Ik
+and {xk
+i }N0
+i=1 we consider, respectively, the following probability measures
+ρk
+Nk := 1
+Nk
+�
+i∈Ik
+δˆxk
+i
+and
+ρk
+N0 := 1
+N0
+�
+i∈I0
+δxk
+i .
+(4.11)
+Informally, the 2-Wasserstein distance W2(ρk
+Nk, ρk
+N0) quantifies the minimal effort needed to
+move the mass from distribution ρk
+Nk into ρk
+N0 (or vice versa) [38]. Let wij denote the amount
+of mass leaving particle xk
+i and going into ˆxk
+i : the cost of such movement is assumed to be given
+by wij∥xk
+i − ˆxk
+j ∥2
+2. Therefore, if we indicate the set of all admissible couplings between the two
+discrete probability measures as
+Γ(ρk
+Nk, ρk
+N0) =
+�
+�
+�w ∈ RN0×Nk :
+Nk
+�
+j=1
+wij = 1
+N0
+,
+N0
+�
+i=1
+wij = 1
+Nk
+, wij ≥ 0, ∀ i, j
+�
+�
+� ,
+(4.12)
+the 2- Wasserstein distance is defined as
+W2(ρk
+Nk, ρk
+N0) :=
+min
+w∈Γ(ρk
+Nk,ρk
+N0)
+�
+��
+i,j
+wij∥xk
+i − ˆxk
+j ∥2
+2
+�
+�
+1
+2
+(4.13)
+see, for instance, [38, Section 6.4.1].
+Before providing estimates on (4.12), let us present a more general result on the impact that
+the random selection strategy has on an arbitrary particle distribution.
+Proposition 4.2 (Stability of random selection procedure). Let z = {zi}i∈I, |I| = N be an
+ensemble of particles and {zi}j∈Isel with Isel ⊆ I, |I| = Nsel a random sub-set of such ensemble.
+Consider the associated empirical distributions µN and µNsel (defined consistently to (4.11)), it
+holds
+E
+�
+W 2
+2 (µN, µNsel)
+�
+≤ 2 var(z) N − Nsel
+N − 1
+(4.14)
+where the expectation is taken with respect to the random selection of Isel.
+23
+
+The proof is provided Appendix A.4. We note how the system variance var(z) enters the
+error estimate due to the randomness of the selection, similar to the Law of Large Number error
+for random variables. In particular, the smaller the particles variance is, the closer the reduced
+particle system will be to the original distribution. This justifies the choice of Nk+1 proposed
+in Section 2.2 where we are allowed to discard particles only if the system shows a contractive
+behavior, see (2.6).
+By iteratively applying Proposition 4.2 and by using suitable stability estimates of dynamics
+(4.3), we are able to bound the error introduced by the random selection procedure as follows.
+Proof details are a given in Appendix A.4.
+Theorem 4.2. Let {xk
+i , yk
+i }N0
+i=1 be constructed according to (4.3) were particles are not discarded,
+and {ˆxk
+i , ˆyk
+i }Ik, |Ik| = Nk where Nk−Nk+1 particles are discarded after update rule (4.3). Assume
+(4.10) is satisfied and consider the probability measures (4.11). If {xk
+i , yk
+i }N0
+i=1, {ˆxk
+i , ˆyk
+i }i∈Ik ⊂
+BM(0) at all step k for some M > 0, it holds
+E
+�
+W 2
+2
+�
+ρk
+Nk, ρk
+N0
+��
+≤ C
+max
+h=1,...,k var
+�
+˜zh� N0 − Nk
+Nk − 1
+(4.15)
+where C = C(∆t, λ, σ, ν, β, α, k, LF, M) and ˜zh = {(ˆxh
+i , ˆyh
+i )}i∈Ih−1 describes the particle system
+just before the random selection procedure at step h ≤ k. The expectation is taken with respect
+to the sampling of {θh}k
+h=1 and with respect to the selection procedure.
+We can directly apply the above result to relate the expected ℓ2-errors of the two particle
+system, which we define as
+Err(k) = E
+�
+� 1
+N0
+�
+i∈I0
+∥xk
+i − x∗∥2
+2
+�
+� ,
+Err(k) = E
+�
+� 1
+Nk
+�
+i∈Ik
+∥ˆxk
+i − x∗∥2
+2
+�
+� ,
+that is, the discrete counterpart of the mean-field error E[∥ ¯Xi
+t − x∗∥2
+2] studied in Theorem 4.1.
+By definition of the Wasserstein-2 distance, we have
+Err(k) = E
+�
+W 2
+2 (ρk
+N0, δx∗)
+�
+for any solution x∗ to (2.1), and the same holds of Errsel(k). We then apply inequality
+W 2
+2 (ρk
+Nk, δx∗) ≤ 2
+�
+W 2
+2 (ρk
+Nk, ρk
+N0) + W 2
+2 (ρk
+N0, δx∗)
+�
+to obtain the following estimate.
+Corollary 4.1. Under the assumptions of Theorem 4.2, at all steps k, it holds
+Errsel(k) ≤ 2
+�
+Err(k) + C
+max
+h=1,...,k var(˜zh) N0 − Nk
+Nk − 1
+�
+.
+(4.16)
+Before concluding the section, let us report some remarks concerning the theoretical results
+just presented.
+24
+
+Remark 4.2.
+• Proof of Theorem 4.2 can be adapted to any other particle system with random selection,
+provided that the update rule is stable with respect to the 2-Wasserstein distance. In the
+proposed method, such stability was proved thanks to the approximation of the global best
+¯y∞,k with ¯yα,k for α ≫ 1 (see (2.2)) and S(x, y) with Sβ(x, y) for β ≫ 1 in the personal
+best update (4.2).
+• Quantitative estimates on the variance decay can be used, if available, to improve the error
+bound in Theorem 4.2, see also proof in Appendix A.4.
+• The error introduced by a sub-sampling technique in a Monte Carlo integral approximation
+is expected to be of order
+2 var(z)
+�
+1
+N − 1 −
+1
+Nsel − 1
+�
+= 2 var(z)
+N − Nsel
+(N − 1)(Nsel − 1) ,
+(4.17)
+see e.g. [23]. Therefore, an additional factor of order 1/(Nsel − 1) seems to be missing
+in Proposition 4.2. We remark, though, that Proposition 4.2 does not concern the Monte
+Carlo approximation of an integral quantity, but rather consider the 2-Wasserstein distance
+between discrete measures. Numerical simulations suggest that estimates of order (4.17)
+do not hold on in this case, see Fig.10.
+5
+Conclusions
+In this work, we studied a Consensus-Based Optimization algorithm with Memory Effects (CBO-
+ME) and random selection for single objective optimization problems of the form (2.1). While
+sharing common features with Particle Swarm Optimization (PSO) methods, CBO-ME differs
+on the way the particle system explore the search space. Its structure provides greater flexi-
+bility in balancing the exploration and exploitation processes. In particular, we implemented
+and analytically investigates a random selection strategy which allows to reduce the algorithm
+computational complexity, without affecting convergence properties and overall accuracy. This
+analysis is entirely general and, in perspective, applicable to other particle swarm-based opti-
+mization methods as well. The convergence analysis to the global minimum is carried out by
+relying on a mean-field approximation of the particle system and error estimates are given un-
+der mild assumptions on the objective function. We compared CBO-ME against CBO without
+memory effects and PSO against several benchmark problem and showed how the introduction
+of memory effects and random selection improves the algorithm performance. Applications to
+image segmentation and machine learning problems are finally reported.
+A
+Proofs
+A.1
+Notation and auxiliary lemmas
+We will use the following notation. For any a ∈ R, |a| indicates the absolute value. For a given
+vector b ∈ Rd, ∥b∥p indicates its p-norm, p ∈ [1, ∞]; (b)ℓ its ℓ-th component; while diag(b) ∈ Rd×d
+25
+
+0
+20
+40
+60
+80
+100
+# particle selected
+!
+Nsel
+"
+0
+0.5
+1
+1.5
+2
+N = 100; d = 3
+squared Wasserstein dist.
+estimate (4.18)
+estimate (4.21)
+0
+20
+40
+60
+80
+100
+# particle selected
+!
+Nsel
+"
+0
+1
+2
+3
+4
+5
+6
+7
+N = 100; d = 10
+squared Wasserstein dist.
+estimate (4.18)
+estimate (4.21)
+Figure 10: Numerical validation of Proposition 4.2 with different dimensions d = 3, 10.
+N = 100 points are randomly, uniformly sampled over [0, 1]d to construct the empirical
+distribution µN and Nsel ∈ [2, N − 1] are discarded to obtain µNsel. The experiment is
+repeated 500 times for all Nsel to obtain an approximation of E
+�
+W 2
+2 (µN, µNsel)
+�
+(blue line).
+In red, estimate provided by Proposition 4.2 (RHS of (4.14)), in yellow the one given
+equation (4.17). Wasserstein distances are computed with the ot.emd function provided by
+the Python Optimal Transport library [6].
+is the diagonal matrix with elements of b on the main diagonal. Let a, b ∈ Rd, ⟨a, b⟩ denotes
+the scalar product in Rd. For a given closed convex set A ⊂ Rd, N(A, x), T (A, x) denote the
+normal and the tangential cone at x ∈ A respectively. The ball or radius r centered at x ∈ Rd
+is indicated with Br(x) = {x ∈ Rd | ∥x∥2 ≤ r}. All considered stochastic processes are assumed
+to take their realizations over the common probability space (Ω, ¯F, P). P(Rd) is the set of Borel
+probability measures over Rd and Pq(Rd) = {µ ∈ P(Rd) |
+�
+∥x∥q
+2dµ < ∞} which we equip with
+the Wasserstein distance Wq, q ≥ 1, see [38]. For a random variable X, X ∼ µ, µ ∈ P(Rd)
+indicates a sampling procedure such that P(X ∈ A) = µ(A) for any Borel set A ⊂ Rd. With
+Unif(A) ∈ P(Rd) we denote the uniform probability measure over a bounded Borel set A.
+Throughout the computations, C will denote an arbitrary positive constant, whose value may
+vary from line to line. Dependence on relevant parameters or variables, will be underlined.
+Lemma A.1 ( [3, Lemma 3.2]). Let F satisfy Assumption 4.1 (in particular the locally Lipschitz
+assumption A1) and ρ1, ρ2 ∈ P4(Rd) with
+�
+∥x∥4
+2dρ1 ,
+�
+∥x∥4
+2dρ2 ≤ M .
+Then, the following stability estimate holds
+∥¯yα(ρ1) − ¯yα(ρ2)∥2 ≤ C W2(ρ1, ρ2)
+for a constant C = C(α, LF, M).
+26
+
+Lemma A.2. Under Assumptions A1 and A6, for any x1, x2, y1, y2 ∈ BM(0) and β > 0, it
+holds
+∥(x1 − y1)Sβ(x1, y1) − (x2 − y2)Sβ(x2, y2)∥2 ≤ C (∥x1 − y1∥2 + ∥x2 − y2∥2)
+where C = C(β, LF, M).
+Proof. Thanks to the Lipschitz continuity of ψ, F and the choice of ψ (Assumptions A1 and
+A6), it holds
+|Sβ(x1, y1) − Sβ(x2, y2)| = |2ψ (β(F(y1) − F(x1)) − 2ψ(β(F(y2) − F(x2)) |
+≤ 2β |F(y1) − F(x1) − F(y2) + F(x2)|
+≤ 2βLF (∥x1 − x2∥2 + ∥y1 − y2∥2) .
+Next, we have
+∥(x1 − y1)Sβ(x1, y1) − (x2 − y2)Sβ(x2, y2)∥2 ≤ ∥(x1 − y1)Sβ(x1, y1) − (x2 − y2)Sβ(x1, y1)∥2
++ (x2 − y2)Sβ(x1, y1) − (x2 − y2)Sβ(x2, y2)∥2
+≤ ∥(x1 − x2 + y2 − y1)Sβ(x1, y1)∥2
++ ∥(x2 − y2)
+�
+Sβ(x1, y1) − Sβ(x2, y2)
+�
+∥2
+≤ 2 (∥x1 − x2∥2 + ∥y1 − y2∥2)
++ 2M|Sβ(x1, y1) − Sβ(x2, y2)|
+≤ C (∥x1 − x2∥2 + ∥y1 − y2∥2)
+with C = C(β, LF, M), where we used the first estimate to conclude.
+A.2
+Proof of Proposition 4.1
+Proof of Proposition 4.1. The proof is based on the Leray–Schauder fixed point theorem [13,
+Chapter 11], and we follow closely the proof steps of [3].
+Step 1. For any ξ ∈ C([0, T], Rd) there exists a unique process ( ˆXt, ˆYt) ∈ C([0, T], Rd)
+satisfying
+d ˆXt = λ(ξ(t) − ˆXt) dt + σ(ξ(t) − ˆXt) ⊗ d ˆBt
+d ˆYt = ν( ˆXt − ˆYt)Sβ( ˆXt, ˆYt) dt
+with Law( ˆX0) = Law( ˆY0) = ρ0 ∈ Rd, by the Lipschitz continuity of the coefficients.
+As a
+consequence, we have that f(t) := Law( ˆXt, ˆYt) satisfies
+d
+dt
+�
+φ df(t) =
+� �
+−λ⟨∇xφ, ξ(t) − x⟩ +
+�
+ℓ=1
+∂2φ
+∂x2
+ℓ
+(ξt) − y)2
+ℓ − νSβ⟨∇yφ, y − x⟩
+�
+df(t)
+for all φ ∈ C2
+b (R2d). Therefore, let ¯ρ(t) = Law( ˆYt), we can set T ξ := ¯yα(¯ρ(·)) ∈ C([0, T], Rd) to
+define
+T : C([0, T], Rd) → C([0, T], Rd).
+27
+
+Step 2. We prove now compactness of T . Thanks to ρ0 ∈ P4(Rd) and standard results for
+SDEs (see [1, Chapter 7]) we have boundedness of the forth moments
+E
+�
+∥ ˆXt∥4
+2 + ∥ ˆYt∥4
+2
+�
+≤ c1
+�
+1 + E[∥ ˆX0∥4
+2 + ∥ ˆY0∥4
+2]ec2t�
+for some c1, c2 > 0. Therefore, we can apply Lemma A.1 to obtain for any 0 < s < t < T,
+∥¯yα(¯ρ(t)) − ¯yα(¯ρ(s))∥2 ≤ CW2 (¯ρ(t), ¯ρ(s)) ≤ ˜C|t − s|1/2
+for some constants C, ˜C > 0, from which H¨older continuity of t �→ ¯yα(¯ρ(t) follows. Therefore,
+by
+T (C([0, T], Rd)) ⊂ C0, 1
+2 ([0, T], Rd) �→ C([0, T], Rd)
+we get compactness of T .
+Step 3. Consider ξ ∈ C([0, T], Rd) satisfying ξ = τT ξ, for τ ∈ [0, 1]. Thanks to [3][Lemma
+3.3] and boundedness of second moments, we obtain compactness of the set
+{ξ ∈ C([0, T], Rd) : ξ = τT ξ, τ ∈ [0, 1]}
+and by Leray–Schauder fixed point theorem there exists a fixed point for the mapping T and
+hence a solution to (4.6).
+Step 4. Assume now there exist two solutions, ( ¯X1
+t , ¯Y 1
+t ) and ( ¯X2
+t , ¯Y 2
+t ) to (4.6) with same
+Brownian process ¯Bt and initial conditions. Let ¯ρℓ = Law( ¯Y ℓ
+t ), ℓ = 1, 2, we have
+∥ ¯X1
+t − ¯X2
+t ∥2
+2 =
+� t
+0
+� ¯X1
+s − ¯X2
+s , ¯yα(¯ρ1(s)) − ¯yα(¯ρ2(s)) − ¯X1
+s + ¯X2
+s
+�
+dt
++
+� t
+0
+�
+diag
+�
+¯yα(¯ρ1(s)) − ¯X1
+s
+�
+− diag
+�
+¯yα(¯ρ2(s)) − ¯X2
+s
+��
+d ¯Bs .
+(A.1)
+We note that all terms can be estimated by means of W 2
+2 (¯ρ1(s), ¯ρ2(s)) and ∥ ¯X1
+s − ¯X2
+s ∥2
+2. Similarly,
+∥ ¯Y 1
+t − ¯Y 2
+t ∥2
+2 can be bounded in terms ∥ ¯X1
+s − ¯X2
+s ∥2
+2 thanks to the Lipschitz continuity of Sβ and
+Lemma A.2. Therefore, for some constant C > 0
+∥ ¯X1
+t − ¯X2
+t ∥2
+2 + ∥ ¯Y 1
+t − ¯Y 2
+t ∥2
+2 ≤ C
+� t
+0
+�
+∥ ¯X1
+s − ¯X2
+s ∥2
+2 + ∥ ¯Y 1
+s − ¯Y 2
+s ∥2
+2 + W 2
+2 (¯ρ1(s), ¯ρ2(s))
+�
+ds
+from which, together with (A.1), follows for some ˜C > 0
+E
+�
+∥ ¯X1
+t − ¯X2
+t ∥2
+2 + ∥ ¯Y 1
+t − ¯Y 2
+t ∥2
+2
+�
+≤ E
+�
+∥ ¯X1
+0 − ¯X2
+0∥2
+2 + ∥ ¯Y 1
+0 − ¯Y 2
+0 ∥2
+2
+�
+e
+˜C t
+by Gr¨onwall’s inequality. Since E
+�
+∥ ¯X1
+0 − ¯X2
+0∥2
+2 + ∥ ¯Y 1
+0 − ¯Y 2
+0 ∥2
+2
+�
+= 0, we proved uniqueness.
+A.3
+Proof of Theorem 4.1
+Having proved there exists a solution ( ¯Xt, ¯Yt)t∈[0,T] to the mean-field process (4.6) we are here
+interested in studying the expected ℓ2-error given by
+E∥ ¯Xt − x∗∥2
+2
+where x∗ is the unique solution to the minimization problem (2.1), see Assumption 4.1. We do
+so by means of the following quantitative version of the Laplace principle.
+28
+
+Proposition A.1 (quantitative Laplace principle [10, Proposition 1]). Let ρ ∈ P(Rd) be such
+that x∗ ∈ supp(ρ) and fix α > 0. For any r > 0, define Fr = supx∈B∗r F(x) − F(x∗) with
+B∗
+r := {x | ∥x − x∗∥∞ ≤ r} .
+Then, under Assumption 4.1, for any r ∈ (0, R0] and q > 0 such that q + Fr ≤ F∞ = ηRγ
+0,
+it holds
+∥yα(ρ) − x∗∥2 ≤
+√
+d(q + Fr)γ
+η
++
+√
+d exp(−αq)
+ρ(B∗r)
+�
+∥x − x∗∥2 dρ(x).
+(A.2)
+We remark that RHS of (A.2) can be made arbitrary small by taking large values of α and
+small values of q, r. To apply Proposition A.1 to all ¯ρ(t) = Law( ¯Yt), we need though to provide
+lower bounds on ¯ρ(t)(B∗
+r) for any small radius r and times t ∈ [0, T].
+Lemma A.3. Let ¯ρ(t) = Law( ¯Yt), with ¯Yt evolving according to (4.6) and limt→0 ¯ρ(t) = ρ0 with
+x∗ ∈ supp(ρ0). Under Assumptions 4.1 and 4.2 , it holds ¯ρ(t)(B∗
+r) ≥ mr > 0, for all t ∈ [0, T]
+and for all r ≤ R0.
+Proof. Let δ = η min{R1, r}γ, we start by proving that the mass in the set
+Lδ = {x ∈ Rd | F(x) ≤ inf F + δ}
+is non-decreasing. We note that for this choice of δ, Lδ is convex due to Assumption 4.1. Consider
+now (Ω, ¯F, P) to be the common probability space over which the considered processes take
+their realization and define Ωδ = {ω : ¯Y0(ω) ∈ Lδ}. By Assumption 4.2, Sβ( ¯Xt(ω), ¯Yt(ω)) = 0
+whenever ¯Xt(ω) /∈ Lδ. Therefore, it holds
+�
+( ¯Xt(ω) − ¯Yt(ω))Sβ( ¯Xt(ω), ¯Yt(ω)) , n( ¯Yt(ω))
+� �
+= 0
+if ¯Xt(ω) /∈ Lδ
+≤ 0
+if ¯Xt(ω) ∈ Lδ
+for
+¯Yt(ω) ∈ ∂Lδ
+for any n( ¯Yt(ω)) ∈ N(Lδ, x) from which follows that ¯Yt(ω) solves
+¯Yt(ω) = ¯Y0(ω) +
+� t
+0
+ΠT (Lδ, ¯Ys(ω))
+�
+( ¯Xs(ω) − ¯Ys(ω))Sβ( ¯Xs(ω), ¯Ys(ω))
+�
+ds
+for all ω ∈ Ωδ. As a consequence, if ¯Y0(ω) ∈ Lδ, ¯Yt(ω) ∈ Lδ for all t ≥ 0 and so
+¯ρ(t)(B∗
+r) = P( ¯Yt ∈ Lδ) ≥ P( ¯Y0 ∈ Lδ) =: mr
+for all t ≥ 0. We conclude by noting that mr > 0 since x∗ ∈ supp(ρ0).
+Next, we study the evolution of the error E∥ ¯Xt − x∗∥2
+2 and, in particular, we try to bound it
+in terms of ∥¯yα(¯ρ(s)) − x∗∥2 and E∥ ¯Xt − x∗∥2 itself for s ∈ [0, t].
+29
+
+Proposition A.2.
+[10, Lemma 1] Let ( ¯Xt, ¯Yt) ∈ C([0, T], R2d) be the solution to (4.6) with
+initial datum ¯X0 ∼ ρ0, ¯Y0 = ¯X0 for some time horizon T > 0. For all t ∈ [0, T], it holds
+E∥ ¯Xt − x∗∥2
+2 ≤
+� t
+0
+�
+− (2λ − σ2)E∥ ¯Xs − x∗∥2
+2 +
+√
+2(λ + σ2)E∥ ¯Xs − x∗∥2∥¯yα(¯ρ(s)) − x∗∥2
++ σ2
+2 ∥¯yα(¯ρ(s)) − x∗∥2
+2
+�
+ds
+(A.3)
+where ¯ρ(t) = Law( ¯Yt).
+Proof of Theorem 4.1. The above result, together with Lemma A.3, leads to the convergence
+in mean-field law of the dynamics towards the solution to (2.1). The proof can be carried out
+exactly as in [10, Theorem 12].
+A.4
+Proof of Proposition 4.2 and Theorem 4.2
+We start by collecting a preliminary result.
+Lemma A.4. Let {xk
+1,i, yk
+1,i}N1
+i=1 and {xk
+2,j, yk
+2,j}N2
+j=1 be two particle populations generated through
+update rules (4.3) with θk
+1,i = θk
+2,j = θk for all i, j and k ∈ Z+. At any iteration step k and for
+any couple of indexes (i, j), it holds
+E
+�
+∥xk+1
+1,i
+− xk+1
+2,j ∥2
+2 + ∥yk+1
+1,i
+− yk+1
+2,j ∥2
+2
+�
+≤
+CE
+�
+∥xk
+1,i − xk
+2,j∥2
+2 + ∥yk
+1,i − yk
+2,j∥2
+2 + ∥¯yα(¯ρk
+1) − ¯yα(¯ρk
+2)∥2
+2
+�
+where C = C(∆t, λ, σ, ν, β) is a positive constant and ¯ρk
+1, ¯ρk
+2 are the empiricial distributions
+associated with {yk
+1,i}N1
+i=1 and {yk
+2,j}N2
+j=1 respectively.
+Proof. For all k ∈ Z+ and i, j
+E∥xk+1
+1,i
+− xk+1
+2,j ∥2
+2 ≤ E
+���xk
+1,i + λ∆t
+�
+¯yα(¯ρk
+1) − xk
+1,i
+�
++ σ
+√
+∆t
+�
+¯yα(¯ρk
+1) − xk
+1,i
+�
+⊗ θk
+1,i
+−
+�
+xk
+2,j + λ∆t
+�
+¯yα(¯ρk
+2) − xk
+2,j
+�
++ σ
+√
+∆t
+�
+¯yα(¯ρk
+2) − xk
+2,j
+�
+⊗ θk
+2,j
+� ���
+2
+2
+≤ 2E
+���
+�
+1 − λ∆t − σ
+√
+∆t diag(θk)
+�
+(xk
+1,i − xk
+2,j)
+���
+2
+2
++ 2E
+���
+�
+λ∆t + σ
+√
+∆t diag(θk)
+� �
+¯yα(¯ρk
+1) − ¯yα(¯ρk
+2)
+����
+2
+2
+≤ 2(1 + σ2∆t)E∥xk
+1,i − xk
+2,j∥2
+2
++ 2(λ2∆t2 + σ2∆t)E∥¯yα(¯ρk
+1) − ¯yα(¯ρk
+2)∥2
+2 ,
+(A.4)
+where we also used that E[(θk)2
+ℓ] = 1 for all ℓ = 1, . . . , d. We now bound ∥yk+1
+1,i
+− yk+1
+2,j ∥2
+2 as
+E∥yk+1
+1,i
+− yk+1
+2,j ∥2
+2 ≤ E
+���yk
+1,i + (ν∆t/2)
+�
+xk+1
+i,1
+− yk
+1,i
+�
+Sβ(xk+1
+1,i , yk
+1,i)
+30
+
+−
+�
+yk
+2,j + (ν∆t/2)
+�
+xk+1
+2,j − yk
+2,j
+�
+Sβ(xk+1
+2,j , yk
+2,j)
+� ���
+2
+2
+≤ CE
+�
+∥xk+1
+i,1
+− xk+1
+j,2 ∥2
+2 + ∥yk
+i,1 − yk
+j,2∥2
+2
+�
+(A.5)
+where we used Lemma A.2 and C = C(∆t, β, ν). By combining (A.4) and (A.5) we get the
+desired estimate.
+Next, we show how the particle update rule (4.3) is stable with respect to the 2-Wasserstein
+distance.
+Proposition A.3 (Stability of update rule (4.3)). Let {xk
+1,i, yk
+1,i}N1
+i=1, {xk
+2,j, yk
+2,j}N2
+j=1 ⊂ BM(0),
+for some M > 0, be two particle populations generated through the update rules (4.3) with
+θk
+1,i = θk
+2,j = θk for all i, j and k ∈ Z+. Let µk
+1, µk
+2 ∈ P(R2d) the empirical probability measures
+defined as
+µk
+1 := 1
+N1
+N1
+�
+i=1
+δ(xk
+1,i,yk
+1,i) ,
+µk
+2 := 1
+N2
+N2
+�
+j=1
+δ(xk
+2,j,yk
+2,j) ,
+it holds
+E
+�
+W 2
+2 (µk+1
+1
+, µk+1
+2
+)
+�
+≤ C1 E
+�
+W 2
+2 (µk
+1, µk
+2)
+�
+,
+where C1 = C1(∆, λ, σ, ν, α, β, LF, M) is positive constant.
+Proof. Let Eθk[·] denote the expectation taken with respect to the sampling of θk only and
+w ∈ RN1×N2 be the optimal coupling between µk
+1, µk
+2, see (4.12) and (4.13). Being w a sub-
+optimal coupling for µk+1
+1
+, µk+1
+2
+, it holds
+Eθk[W 2
+2 (µk+1
+1
+, µk+1
+2
+)] ≤ Eθk
+�
+i,j
+wij
+�
+∥xk+1
+1,i
+− xk+1
+2,j ∥2
+2 + ∥yk+1
+1,i
+− yk+1
+2,j ∥2
+2
+�
+≤ C
+�
+i,j
+wij
+�
+∥xk
+1,i − xk
+2,j∥2
+2 + ∥yk
+1,i − yk
+2,j∥2
+2
+�
++ ∥¯yα(¯ρk
+1) − ¯yα(¯ρk
+2)∥2
+2
+where we used the linearity of the expectation, estimates given by Lemma A.4 and, to take the
+last term out of the sum, the fact that �
+ij wij = 1.
+To estimate the distance between the two consensus points, we use Lemma A.1 and note
+that the coupling w is sub-optimal for ¯ρk
+1, ¯ρk
+2. By Lemma A.1, it follows
+∥¯yα(¯ρk
+1) − ¯yα(¯ρk
+2)∥2
+2 ≤ CW 2
+2 (¯ρk
+1, ¯ρk
+2) ≤ C
+�
+i,j
+wij∥yk
+1,i − yk
+2,j∥2 .
+Therefore,
+Eθk[W 2
+2 (µk+1
+1
+, µk+1
+2
+)] ≤ C1
+�
+i,j
+wij
+�
+∥xk
+1,i − xk
+2,j∥2
+2 + ∥yk
+1,i − yk
+2,j∥2
+2
+�
+= C1 W 2
+2 (µk
+1, µk
+2) ,
+thanks to the optimality of w, with C1 = C1(∆, λ, σ, ν, α, β, LF, M) being a positive constant.
+One can conclude by taking the expectation of the above inequality with respect to the remaining
+sampling processes.
+31
+
+We now quantify the impact of the particle discarding step.
+Proof of Proposition 4.2. For notational simplicity, let us introduce zi = (xi, yi) ∈ R2d.
+As
+in (4.13), the 2-Wasserstein distance is given by an optimal coupling between the full particle
+system {zi}i∈I and the reduced one {zj}j∈Isel. We consider the following transportation of mass
+from µN to µNsel: if particle i has not been discarded, all its mass remains in xi, otherwise the
+mass is uniformly distributed among the selected particles to generate an admissible coupling
+w ∈ RN×Nsel. This means that w is given by
+wij =
+�
+�
+�
+�
+�
+1/N
+if j = i, i ∈ Isel
+1/(N · Nsel)
+if i ∈ I \ Isel, j ∈ Isel
+0
+else .
+(A.6)
+We note that such coupling w satisfies the coupling conditions
+�
+j∈Isel
+wij = 1
+N
+�
+i∈I
+wij =
+1
+Nsel
+,
+∀ i ∈ I, j ∈ Isel
+(A.7)
+and that this choice will be in general sub-optimal. Therefore, it holds
+W 2
+2 (µN, µNsel) ≤
+�
+i∈I, j∈Isel
+wij∥zi − zj∥2
+2
+= 1
+N
+�
+i∈Isel
+∥zi − zi∥2
+2 +
+1
+N · Nsel
+�
+i∈I\Isel, j∈Isel
+∥zi − zj∥2
+2
+=
+1
+N · Nsel
+�
+i,j∈I
+∥zi − zj∥2
+2 1i∈I\Isel 1j∈Isel
+where 1i∈I = 1 if i ∈ I and zero otherwise.
+Now, the probability of having i ∈ I \ Isel is given by (N − Nsel)/N, while the probability of
+having j ∈ Isel (condition i ∈ I \ Isel) is given by Nsel/(N − 1). Hence, we have
+E
+�
+1i∈I\Isel 1j∈Isel
+�
+= P [i ∈ I \ Isel, j ∈ Isel] = (N − Nsel)Nsel
+N(N − 1)
+from which follows
+E
+�
+W 2
+2 (µN, µNsel)
+�
+≤
+1
+N · Nsel
+�
+i,j∈I
+∥zi − zj∥2
+2 E
+�
+1i∈I\Isel 1j∈Isel
+�
+=
+1
+N · Nsel
+· (N − Nsel)Nsel
+N(N − 1)
+�
+i,j∈I
+∥zi − zj∥2
+2 .
+The desired estimates can finally be obtained by noting that the variance can be computed as
+var(z) = 1/(2N2) �
+i,j∈I ∥zi − zj∥2
+2, see definition (2.5).
+32
+
+Finally, we are ready to provide a proof of Theorem 4.2.
+Proof of Theorem 4.2. Let {(xk
+i , yk
+i )}i∈Ik, |Ik| = Nk be the sequence of particles generated by
+iteration (4.3) where additionally Nk+1 − Nk particles are discarded after each step k ≥ 0. We
+denote with µk
+Nk ∈ P(R2d) the empirical measure associated with such particle system given by
+µk
+Nk = 1
+Nk
+�
+i∈Ik
+δ(xk
+i ,yk
+i ) .
+We also introduce the measures µk
+N0, k ≥ 0 corresponding to a particle system generated with
+the same initial conditions µ0
+N0 but where no particle reduction occurs. Consistently, we define
+µh
+Nk, h > k to represent the particle system generated starting from µk
+Nk, after h − k iterations,
+with no random selection. The relation between such measures is summarized in the following
+diagram
+µ0
+N0
+→
+µ1
+N0
+→
+µ2
+N0
+→
+. . .
+→
+µk
+N0
+...
+���
+µ1
+N1
+→
+µ2
+N1
+→
+. . .
+→
+µk
+N1
+...
+���
+µ2
+N2
+→
+. . .
+→
+µk
+N2
+...
+...
+...
+µk
+Nk
+...
+(A.8)
+where → indicates an iteration step (4.3) while ��� a particle reduction procedure. Therefore,
+we are interested in studying the distance between the main diagonal of such diagram µk
+Nk, cor-
+responding to the system with particle reduction, and the first line µk
+N0 where particle reduction
+is never performed.
+We note that the 2-Wasserstein distance between subsequent rows can be estimated thanks
+to Proposition A.3 and Proposition 4.2. Let ˜zh+1 denote the set of particles associated with
+the probability measure µh+1
+Nh , that is, the particle systems before the selection procedure (up-
+per diagonal elements in scheme (A.8)). By first applying Proposition A.3 and, subsequently,
+Proposition 4.2 to ˜zh+1, we obtain that for some constant C > 0
+E
+�
+W 2
+2 (µk
+Nk, µk
+N0)
+�
+≤ C
+k−1
+�
+h=0
+E
+�
+W 2
+2
+�
+µk
+Nh, µk
+Nℓ+1
+��
+≤ C
+k−1
+�
+h=0
+Ck−h+1
+1
+E
+�
+W 2
+2
+�
+µh+1
+Nh , µh+1
+Nh+1
+��
+≤ 2 C
+k−1
+�
+h=0
+Ck−h+1
+1
+var
+�
+˜zh+1� Nh − Nh+1
+Nh − 1
+33
+
+≤ C2
+max
+h=1,...,k var
+�
+˜zh�
+1
+Nk − 1
+k−1
+�
+h=0
+Nh − Nh+1
+= C2
+max
+h=1,...,k var
+�
+˜zh� N0 − Nk
+Nk − 1
+with C2 = C2(∆t, λ, σ, ν, β, α, k, M). Finally, the desired estimate follows after noting that
+W 2
+2 (ρk
+Nk, ρk
+N0) ≤ W 2
+2 (µk
+Nk, µk
+N0)
+since ∥xk
+i − xk
+j ∥2
+2 ≤ ∥(xk
+i , yk
+i ) − (xk
+j , yk
+j )∥2
+2 for all couples of particles (i, j).
+Acknowledgments
+This work has been written within the activities of GNCS group of INdAM (National Institute
+of High Mathematics). L.P. acknowledges the partial support of MIUR-PRIN Project 2017,
+No. 2017KKJP4X “Innovative numerical methods for evolutionary partial differential equations
+and applications”. The work of G.B. is funded by the Deutsche Forschungsgemeinschaft (DFG,
+German Research Foundation) through 320021702/GRK2326 “Energy, Entropy, and Dissipative
+Dynamics (EDDy)” and SFB 1481 “Sparsity and Singular Structures”. S.G. acknowledges the
+support of the ESF PhD Grant “Mathematical and statistical methods for machine learning in
+biomedical and socio-sanitary applications”.
+References
+[1] L. Arnold. Stochastic Differential Equations: Theory and Applications. Wiley Interscience,
+first edition, 1974. Structure-preserving algorithms for ordinary differential equations.
+[2] S. Arora, J. Acharya, A. Verma, and P. K. Panigrahi. Multilevel thresholding for image
+segmentation through a fast statistical recursive algorithm. Pattern Recognition Letters,
+29(2):119–125, 2008.
+[3] J. A. Carrillo, Y.-P. Choi, C. Totzeck, and O. Tse. An analytical framework for consensus-
+based global optimization method.
+Math. Models Methods Appl. Sci., 28(6):1037–1066,
+2018.
+[4] J. A. Carrillo, S. Jin, L. Li, and Y. Zhu. A consensus-based global optimization method for
+high dimensional machine learning problems. ESAIM: COCV, 27:S5, 2021.
+[5] J. Chen, S. Jin, and L. Lyu. A consensus-based global optimization method with adaptive
+momentum estimation. Communications in Computational Physics, 31(4):1296–1316, 2022.
+[6] R. Flamary, N. Courty, A. Gramfort, M. Z. Alaya, A. Boisbunon, S. Chambon, L. Chapel,
+A. Corenflos, K. Fatras, N. Fournier, L. Gautheron, N. T. Gayraud, H. Janati, A. Rakotoma-
+monjy, I. Redko, A. Rolet, A. Schutz, V. Seguy, D. J. Sutherland, R. Tavenard, A. Tong,
+and T. Vayer. POT: Python Optimal Transport. Journal of Machine Learning Research,
+22(78):1–8, 2021.
+34
+
+[7] M. Fornasier, H. Huang, L. Pareschi, and P. S¨unnen. Consensus-based optimization on
+the sphere: Convergence to global minimizers and machine learning. J. Machine Learning
+Research, 22(237):1–55, 2021.
+[8] M. Fornasier, H. Huang, L. Pareschi, and P. S¨unnen. Consensus-based optimization on
+hypersurfaces: Well-posedness and mean-field limit. Mathematical Models and Methods in
+Applied Sciences, 30(14):2725–2751, 2020.
+[9] M. Fornasier, T. Klock, and K. Riedl. Consensus-based optimization methods converge
+globally. arXiv:2103.15130, 2021.
+[10] M. Fornasier, T. Klock, and K. Riedl. Convergence of anisotropic consensus-based opti-
+mization in mean-field law. In J. L. Jim´enez Laredo, J. I. Hidalgo, and K. O. Babaagba,
+editors, Applications of Evolutionary Computation, pages 738–754, Cham, 2022. Springer
+International Publishing.
+[11] A. H. Gandomi, G. J. Yun, X.-S. Yang, and S. Talatahari. Chaos-enhanced accelerated
+particle swarm optimization. Communications in Nonlinear Science and Numerical Simu-
+lation, 18(2):327–340, 2013.
+[12] G. Garrigos, L. Rosasco, and S. Villa. Convergence of the forward-backward algorithm:
+beyond the worst-case with the help of geometry. Mathematical Programming, June 2022.
+[13] D. Gilbarg and N. Trudinger.
+Elliptic Partial Differential Equations of Second Order.
+Classics in Mathematics. Springer Berlin Heidelberg, 2001.
+[14] S. Grassi and L. Pareschi.
+From particle swarm optimization to consensus based opti-
+mization: Stochastic modeling and mean-field limit. Mathematical Models and Methods in
+Applied Sciences, 31(08):1625–1657, 2021.
+[15] S.-Y. Ha, S. Jin, and D. Kim. Convergence of a first-order consensus-based global optimiza-
+tion algorithm. Mathematical Models and Methods in Applied Sciences, 30(12):2417–2444,
+2020.
+[16] S.-Y. Ha, S. Jin, and D. Kim. Convergence and error estimates for time-discrete consensus-
+based optimization algorithms. Numerische Mathematik, 147(2):255–282, Feb 2021.
+[17] J. H. Holland. Adaptation in natural and artificial systems: an introductory analysis with
+applications to biology, control, and artificial intelligence. MIT press, 1992.
+[18] E. H. Houssein, A. G. Gad, K. Hussain, and P. N. Suganthan. Major advances in particle
+swarm optimization: Theory, analysis, and application. Swarm and Evolutionary Compu-
+tation, 63:100868, 2021.
+[19] H. Huang and J. Qiu. On the mean-field limit for the consensus-based optimization. Math-
+ematical Methods in the Applied Sciences, 45(12):7814–7831, 2022.
+[20] H. Huang, J. Qiu, and K. Riedl. On the global convergence of particle swarm optimization
+methods. arXiv:2201.12460, 2022.
+35
+
+[21] K. Hussain, M. N. Mohd Salleh, S. Cheng, and Y. Shi. Metaheuristic research: a compre-
+hensive survey. Artificial Intelligence Review, 52(4):2191–2233, Dec 2019.
+[22] M. Jamil and X.-S. Yang. A literature survey of benchmark functions for global optimization
+problems. Int. Journal of Mathematical Modelling and Numerical Optimisation, 2(4):150–
+194, 2013.
+[23] S. Jin, L. Li, and J.-G. Liu. Random Batch Methods (RBM) for interacting particle systems.
+Journal of Computational Physics, 400:108877, 2020.
+[24] J. Kennedy and R. Eberhart. Particle swarm optimization. In Proceedings of ICNN’95 -
+International Conference on Neural Networks, volume 4, pages 1942–1948 vol.4, 1995.
+[25] S. Kirkpatrick, C. D. Gelatt Jr, and M. P. Vecchi. Optimization by simulated annealing.
+science, 220(4598):671–680, 1983.
+[26] Y. LeCun and C. Cortes. MNIST handwritten digit database. 2010.
+[27] C. Li, Y. Liu, A. Zhou, L. Kang, and H. Wang. A fast particle swarm optimization algorithm
+with cauchy mutation and natural selection strategy. In L. Kang, Y. Liu, and S. Zeng,
+editors, Advances in Computation and Intelligence, pages 334–343, Berlin, Heidelberg, 2007.
+Springer Berlin Heidelberg.
+[28] J. Li, X. Dong, S. Ruan, and L. Shi. A parallel integrated learning technique of improved
+particle swarm optimization and bp neural network and its application. Scientific Reports,
+12(1):19325, Nov 2022.
+[29] Y. Liang and L. Wang. Applying genetic algorithm and ant colony optimization algorithm
+into marine investigation path planning model. Soft Computing, 24(11):8199–8210, Jun
+2020.
+[30] B. Liu, L. Wang, Y.-H. Jin, F. Tang, and D.-X. Huang. Improved particle swarm optimiza-
+tion combined with chaos. Chaos, Solitons & Fractals, 25(5):1261–1271, 2005.
+[31] A. Nickabadi, M. M. Ebadzadeh, and R. Safabakhsh. A novel particle swarm optimization
+algorithm with adaptive inertia weight. Applied Soft Computing, 11(4):3658–3670, 2011.
+[32] N. Otsu. A threshold selection method from gray-level histograms. IEEE transactions on
+systems, man, and cybernetics, 9(1):62–66, 1979.
+[33] M. E. H. Pedersen. Good parameters for particle swarm optimization. Hvass Lab., Copen-
+hagen, Denmark, Tech. Rep. HL1001, pages 1551–3203, 2010.
+[34] R. Pinnau, C. Totzeck, O. Tse, and S. Martin. A consensus-based model for global opti-
+mization and its mean-field limit. Math. Models Methods Appl. Sci., 27(1):183–204, 2017.
+[35] E. Platen. An introduction to numerical methods for stochastic differential equations. Acta
+Numerica, 8:197–246, 1999.
+36
+
+[36] R. Poli, J. Kennedy, and T. Blackwell. Particle swarm optimization. Swarm intelligence,
+1(1):33–57, 2007.
+[37] K. Riedl. Leveraging memory effects and gradient information in consensus-based optimiza-
+tion: On global convergence in mean-field law. arXiv.2211.12184, 2022.
+[38] F. Santambrogio. Optimal Transport for Applied Mathematicians. Birkh¨auser, 2015.
+[39] Y. Shi and R. Eberhart. A modified particle swarm optimizer. In 1998 IEEE international
+conference on evolutionary computation proceedings. IEEE world congress on computational
+intelligence (Cat. No. 98TH8360), pages 69–73. IEEE, 1998.
+[40] R. Storn and K. Price. Differential evolution–a simple and efficient heuristic for global
+optimization over continuous spaces. Journal of global optimization, 11(4):341–359, 1997.
+[41] A.-S. Sznitman. Topics in propagation of chaos. In P.-L. Hennequin, editor, Ecole d’Et´e de
+Probabilit´es de Saint-Flour XIX — 1989, pages 165–251, Berlin, Heidelberg, 1991. Springer
+Berlin Heidelberg.
+[42] K.-S. Tang, K.-F. Man, S. Kwong, and Q. He. Genetic algorithms and their applications.
+IEEE signal processing magazine, 13(6):22–37, 1996.
+[43] D. Tian and Z. Shi. MPSO: Modified particle swarm optimization and its applications.
+Swarm and evolutionary computation, 41:49–68, 2018.
+[44] C.-Y. Tsai, T.-Y. Liu, and W.-C. Chen. A novel histogram-based multi-threshold searching
+algorithm for multilevel colour thresholding. International Journal of Advanced Robotic
+Systems, 9(5):223, 2012.
+[45] R. Tuli, H. N. Soneji, and P. Churi. Pixadapt: A novel approach to adaptive image encryp-
+tion. Chaos, Solitons & Fractals, 164:112628, 2022.
+[46] M. ˇCrepinˇsek, S.-H. Liu, and M. Mernik.
+Exploration and exploitation in evolutionary
+algorithms: A survey. ACM Comput. Surv., 45(3), jul 2013.
+[47] D. Wang, D. Tan, and L. Liu. Particle swarm optimization algorithm: an overview. Soft
+computing, 22(2):387–408, 2018.
+[48] X.-S. Yang. Nature-inspired optimization algorithms. Academic Press, 2020.
+[49] X.-S. Yang, S. Deb, and S. Fong. Accelerated particle swarm optimization and support
+vector machine for business optimization and applications. In S. Fong, editor, Networked
+Digital Technologies, pages 53–66, Berlin, Heidelberg, 2011. Springer Berlin Heidelberg.
+[50] Y. Zhang and X. Kong. A particle swarm optimization algorithm with empirical balance
+strategy. Chaos, Solitons & Fractals: X, 10:100089, 2023.
+37
+
diff --git a/2dFQT4oBgHgl3EQfFzW1/content/tmp_files/load_file.txt b/2dFQT4oBgHgl3EQfFzW1/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..77beccbeda275d0d3f3e6a4c48b1cc2ed9fc40e8
--- /dev/null
+++ b/2dFQT4oBgHgl3EQfFzW1/content/tmp_files/load_file.txt
@@ -0,0 +1,2352 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf,len=2351
+page_content='Consensus based optimization with memory effects: random selection and applications Giacomo Borghi∗ Sara Grassi† Lorenzo Pareschi† February 1, 2023 Abstract In this work we extend the class of Consensus-Based Optimization (CBO) metaheuris- tic methods by considering memory effects and a random selection strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The proposed algorithm iteratively updates a population of particles according to a consensus dynamics inspired by social interactions among individuals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The consensus point is computed taking into account the past positions of all particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' While sharing features with the popular Parti- cle Swarm Optimization (PSO) method, the exploratory behavior is fundamentally different and allows better control over the convergence of the particle system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We discuss some im- plementation aspects which lead to an increased efficiency while preserving the success rate in the optimization process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In particular, we show how employing a random selection strat- egy to discard particles during the computation improves the overall performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Several benchmark problems and applications to image segmentation and Neural Networks training are used to validate and test the proposed method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A theoretical analysis allows to recover convergence guarantees under mild assumptions of the objective function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This is done by first approximating the particles evolution with a continuous-in-time dynamics, and then by taking the mean-field limit of such dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Convergence to a global minimizer is finally proved at the mean-field level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Keywords: consensus-based optimization, stochastic particle methods, memory effects, ran- dom selection, machine learning, mean-field limit Contents 1 Introduction 2 2 Consensus-based optimization with memory effects 4 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Particles update rule .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 4 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Random selection strategy .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Comparison with CBO and PSO .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 5 ∗RWTH Aachen University, Institute for Geometry and Applied Mathematics, Aachen, Germany (borghi@eddy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='rwth-aachen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='de) †University of Ferrara, Department of Mathematics and Computer Science & Center for Modelling Computing and Statistics, Ferrara, Italy (sara.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='grassi@unife.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='it, lorenzo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='pareschi@unife.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='it) 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13242v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='OC] 30 Jan 2023 3 Numerical results 7 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Tests on benchmark problems .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 8 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Applications .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 13 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Image segmentation .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 13 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Approximating functions with NN .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 16 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Application on MNIST dataset .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 17 4 Theoretical analysis 19 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Mean-field approximation .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 19 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Convergence in mean-field law .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 21 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Random selection analysis .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 23 5 Conclusions 25 A Proofs 25 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Notation and auxiliary lemmas .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 25 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 27 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 28 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 and Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 30 1 Introduction Meta-heuristic algorithms are recognized as trustworthy, easy to understand and to adapt op- timization methods which have been widely applied to a several fields such as Machine Learn- ing [28], path planning [29] and image processing [45], to name a few.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Starting form a set of possible solutions, a meta-heuristic algorithm typically updates such set iteratively by combining deterministic and stochastic choices, often inspired by natural phenomena.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Exploration of the search space and exploitation of the current knowledge are the two fundamental mechanisms driving the algorithm iteration [46].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Examples of established meta-heuristic algorithms are given by Genetic Algorithm (GA) [17,42], Simulated Annealing (SA) [25], Particle Swarm Optimiza- tion (PSO) [24] and Differential Evolution (DE) [40].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We refer to [21] for a complete literature review.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consensus-Based Optimization (CBO) is a class of gradient-free meta-heuristic algorithms inspired by consensus dynamics among individuals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' After its introduction [34] it has gained popularity among the mathematical community due to its robust mathematical framework [3,9, 16,19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In CBO algorithms, a population of particles concentrates around a consensus point given by a weighted average of the particles position.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In the computation of such consensus point, more importance is given to those particles attaining relatively low values of the objective function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The exploration mechanism is introduced by randomly perturbing the particles positions at each iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Particles which are close to the consensus point are subject to small perturbations, while those that are far from it display a more exploratory behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In this work, following the recent analysis in [14], we study a Consensus-Based Optimization algorithm with Memory Effects (CBO-ME) where the consensus point is computed among the whole history of the particles positions and not just on the positions of the current iteration, as 2 in the original CBO method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This is done by keeping track of the best position found so far by each particle, and computing the consensus point among these “personal” bests.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' While sharing common elements with PSO, such as convergence to a promising point and the presence of personal bests, CBO-ME differs in the way the exploration mechanism is implemented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, in CBO-ME, as in CBO algorithms, the stochastic behavior is given by adding Gaussian noise to the particles dynamics and can be tuned independently on the exploitation mechanisms, leading to a better control over the particles convergence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, while in classical PSO methods it is the balance between local best and global best that governs the optimization strategy, in CBO methods it is the balance between exploration and exploitation mechanisms that determines the choice of parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We recall that a generalization of PSO methods that allows leveraging the same flexibility in searching the global minimum as in CBO algorithms has been recently presented in [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Many real-life problems, especially those regarding Machine Learning, require to optimize a large number of parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, it essential to design fast algorithm to save computa- tional time and memory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This is a major weakness of swarm-based methods, which require a set of particles to minimize the problem, unlike gradient-based methods that can work on a single particle trajectory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For methods based on a collection of particles, existing algorithms can be improved by discarding particles whenever the system has a prominent exploitative behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This is sometimes referred as “natural selection strategy” in the DE literature [27,40] and aims to discard the non-promising solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Inspired by particle simulations techniques where it is important to preserve the particles distribution, we examine a “random selection strategy” where particles are discarded randomly based on the local consensus achieved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We will discuss such implementation aspects by testing CBO-ME against high-dimensional learning problems and theoretically analyze the impact of the random selection strategy on the system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In partic- ular, we prove that if the full particle system is expected to converge towards a solution to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1), so will the reduce one, provided a sufficient number of particles remains active.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Note that, such analysis can be generalized to other particle dynamics and may be of independent interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Owing to the convergence analysis of CBO algorithms [3, 9, 10, 19] and recent analysis of PSO [14, 20] we are able to prove convergence of the algorithm under mild assumption on the objective function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This is done by first approximating the algorithm with a continuous-in- time dynamics and secondly by giving a probabilistic description to the particles system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By assuming propagation of chaos [41], particles are considered to behave independently according to the same law.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This allows to reduce the possible large system of equations to a single partial differential equation: the so-called mean-field model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Such model is then analyzed to recover convergence guarantees under precise assumption on the objective function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Developed in the field of statistical physics, this approach has shown be fruitful in studying particle-based meta- heuristic algorithms [9,10,20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that convergence in mean-field law was recently proved in [37] in an independent work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The rest of the paper is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Section 2 is devoted to the introduction of the CBO-ME algorithm with random selection and comparison with CBO methods without memory effects and PSO.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In Section 3 validate the proposed methods against several benchmark problems and two Machine Learning tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Theoretical convergence guarantees and analysis of the random selection strategy are summarized in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Some final remarks are given in Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Technical details of the theoretical analysis are given in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3 2 Consensus-based optimization with memory effects In this section, we present the Consensus-Based Optimization algorithm with Memory Effects (CBO-ME) to solve problems of the form x∗ ∈ argmin x∈Rd F(x) , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) where Rd, d ∈ N is the, possibly large, search domain for the continuous function F ∈ C(Rd, R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We will do so by highlighting similarities and differences between classical CBO methods and PSO algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Particles update rule At each iteration step k and for every particle i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , N, we store its position xk i and its best position found so far yk i = argmin{F(xk 1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , F(xk N)}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The best positions are used to compute a consensus point ¯yα,k = N � i=1 ωk i yk i with ωk i = e−αF(yk i ) �N j=1 e−αF(yk j ) (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) which approximate the global best solution ¯yα,k among all particles and all times for α > 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, thanks to the choice of the weights ωk i , we have that ¯yα,k −→ ¯y∞,k := argmin{F(yk 1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , F(yk N)} (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) as α → ∞, provided that there is only one global best position among {yk 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , yk N}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Such ap- proximation was first introduced for CBO methods [34] as it leads to more amenable theoretical analysis, but it also allows more flexibility.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, relatively small values of α are typically used at the beginning of the computation to promote exploration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Large values of α, on the other hand, lead to better exploitation of the computed solutions and to higher accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that the weights used in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) correspond in statistical mechanics to the Boltzmann-Gibbs distribution associated with the energy F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In this context, α plays the role of the inverse of the system temperature T and the limit α → ∞ corresponds to T → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Once the consensus point ¯yα,k is computed, the particle positions are then updated according to the law xk+1 i = xk i + λ � ¯yα,k − xk i � + σ � ¯yα,k − xk i � ⊗ θk i (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) with θk i ∈ Rd randomly sampled from the normal distribution (θk i ∼ N(0, Id)) and where ⊗ is the component-wise product.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The update rule is characterized by a deterministic component of strength λ ∈ (0, 1) promot- ing concentration around the consensus point ¯yα,k and a stochastic component of strength σ > 0 promoting exploration of the search space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As the latter depends on the difference (¯yα,k − xk i ), the random behavior is stronger for particles which are far form the consensus point, whereas it is weaker for those that are close to it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Also, such exploration resemble an anisotropic diffu- sive behavior exploring every coordinate direction at a different rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This approach was first proposed in [4] in the context of CBO methods and has been proved to suffer less from the curse of dimensionality with the respect to the originally proposed isotropic diffusion given by σ∥¯yα,k − xk i ∥2θk i with θk i being again a normally distributed d-dimensional vector [4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 4 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Random selection strategy When the particle system concentrates around the consensus point, showing a mostly exploita- tive behavior, we employ a particle selection strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Discarding particles introduces additional stochasticity to the system, while reducing the computational cost.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Following the approach sug- gested in [7], we check the evolution of the system variance to decide how many particles to (eventually) discard.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For a given set of particles z = {zi}i∈J, the system variance is given by var(z) := 1 |J| � j∈J ∥zj − m(z)∥2 2 with m(z) := 1 |J| � i∈J zi , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5) where |J| indicates the cardinality of I, that is, the number of particles in this context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let Ik ⊆ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , N} be the set of active particles at step k and Nk = |Ik|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' To decide how many particles to select, we compare the variance of the particle system before the position update (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4), xk = {xk}i∈Ik and after it, ˜xk+1 = {xk+1 i }i∈Ik.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Then, the number Nk+1 of particles we select for the next iteration is given by ˜Nk+1 = � Nk � 1 + µ var(˜xk+1) − var(xk+1) var(xk+1) �� Nk+1 = min � max � ˜Nk+1, Nmin � , Nk � (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) ⌊z⌋ being the integer part of a number z and Nmin ∈ N the smallest amount of particles we allow to have.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Then, a subset Ik+1 ⊂ Ik, |Ik+1| = Nk+1, of particles is randomly selected to continue the computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The parameter µ ∈ [0, 1] regulates the mechanism: for µ = 0 there is no particle discarding, while for µ = 1 the maximum number of particles is discarded if the variance is decreasing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As we will see in Section 3, this random selection mechanism dramatically reduces the computational time without affecting the algorithm performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We will also theoretically analyze this aspect in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3, where we show that convergence properties are preserved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As stopping criterion, we keep a counter n on how many times ∥¯yα,k+1 − ¯yα,k∥2 is smaller than a certain tolerance δstall.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' If this happens for more than a given nstall number of times in a row, we assume the particles system found a solution and stop the computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A maximum number of iteration kmax representing the computational budget is also given.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The proposed CBO-ME is summarized in Algorithm 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In the meta-heuristic literature, particles are usually discarded depending on their objective value, in a way that particles with high values are more likely to be discarded [27,40].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The proposed strategy does not add a further heuristic strategy but simply cut down the algorithm complexity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Also, the convergence properties are in this way expected to be preserved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that, on the other hand, there is no straightforward way to generate particles and, at the same time, preserve the particle system distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Comparison with CBO and PSO What distinguishes CBO-ME from plain CBO, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g [4,34], is clearly the introduction of the best positions {yk i }N i=1 and the fact that the consensus point is calculated among them and not 5 Algorithm 1: Consensus-Based Optimization with Memory Effects (CBO-ME) Input: F, N0, kmax, λ, σ, α, nstall and δstall;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 1 Inizialize N0 particle positions xi 0, i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , N;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 2 y0 i ← x0 i for all i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , Nk;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3 Compute yα,0 according to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 4 k ← 0, n ← 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 5 while k < kmax and n < nstall do 6 for i = 1 to Nk do 7 θk i ∼ N(0, Id);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 8 Compute xk+1 i according to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 9 if F(xk+1 i ) < F(yk i ) then 10 yk+1 i ← xk+1 i ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 11 else 12 yk+1 i ← yk i ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 13 end 14 end 15 Compute ¯yα,k+1 according to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 16 if ∥¯yα,k+1 − ¯yα,k∥2 < δstall then 17 n ← n + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 18 else 19 n ← 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 20 end 21 Compute Nk+1 according to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 22 if Nk+1 < Nk then 23 Randomly discard Nk+1 − Nk particles;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 24 k ← k + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 25 end 26 return ¯yα,k, F(¯yα,k) just among the particle positions {xk i }N i=1 at that given time k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, the classical CBO update rule without memory effects (and with anisotropic diffusion and projection step) is given by xk+1 i = xk i + λ � ¯xα,k − xk i � + σ � ¯xα,k − xk i � ⊗ θk i (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7) where ¯xα,k is defined consistently with (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) (by substituting yk i with xk i ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As we will see in the numerical tests, the use of memory effects improves the algorithm performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Since alignment towards personal bests yk i and towards the global best ¯y∞,k are also the fundamental building blocks of PSO algorithms, we highlight now the main differences and similarities between PSO and CBO-ME.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For completeness, we recall the canonical PSO method, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [36], using the notation of (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) for easier comparison � xk+1 i = xk i + vk+1 i vk+1 i = wvk i + C1 � yk i − xk i � ⊗ ˆθk i,1 + C2 � ¯y∞,k − xk i � ⊗ ˆθk i,2 (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) 6 where vk i are the particles velocities, w, C1, C2 > 0 are the algorithm parameters and θk i,1, θk i,2 are uniformly sampled from [0, 1]d (ˆθk i,1, ˆθk i,2) ∼ Unif([0, 1]d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Several variants and improvements have been proposed starting from the above dynamics, but a complete review is beyond the scope of this paper and we refer to the recent survey [47] for more references.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We are interested in highlighting the main differences between (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) and (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) regarding the stochastic components: in CBO-ME deterministic and stochastic steps are de-coupled and tuned by two different parameters (λ and σ), while in PSO they are coupled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8), deterministic and stochastic components are both controlled by the same parameter: C1 in the case of personal best dynamics and C2 for the global best one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By splitting the term C2 � ¯y∞,k − xk i � ˆθk i,2 into a deterministic step and a zero-mean term we obtain C2 � ¯y∞,k − xk i � ⊗ ˆθk i,2 = C2 2 � ¯y∞,k − xk i � + C2 2 � ¯y∞,k − xk i � ⊗ θk i,2 (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9) with θk i,2 = 2ˆθk i,2 − 1, θk i,2 ∼ Unif([−1, 1]d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Suggested in [14], such rewriting highlights how increasing the alignment strength towards the global best (by increasing C2) necessary increases the stochasticity of the system as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) and (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7), on the other hand, one is allowed to tune the exploration and exploitation behaviors separately, by either changing parameters λ or σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Clearly, CBO-ME also differs from PSO due to its first-order dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Having the aim of resembling birds flocking, the first PSO algorithm [24] was proposed as a second-order dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The inertia weight w, introduced later in [39], became an essential parameter to prevent early convergence of the swarm and to increase the global exploration behavior, especially at the beginning of the computation, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [31, 39] and reviews [18, 36, 47] for more references.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that several other strategies have proposed to improve PSO exploration behavior, see, for example, [50].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As already mentioned, in CBO methods convergence and exploration are de-coupled and can be tuned separately.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, to keep the algorithm more amenable to theoretical analysis, we consider a simpler first-order dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that a CBO dynamics with inertia mechanism was proposed in [5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Similarly, we found the contribution given by the personal best alignment non-essential and difficult to tune.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Thus, the lack of alignment towards personal best in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Replacing alignment towards personal best with gaussian noise was also suggested in [48] where authors proposed the Accelerated PSO (APSO) algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Further studied in [11,49], APSO also allows to de-couple the stochastic component from the deterministic one and the noise is heuristically tuned to decrease during the computation (as in Simulated Annealing [25]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In CBO methods, the noise strength automatically adapts as it depends on the distance from the consensus point, which is also different for every particle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For completeness, we note that many other variants of PSO have been proposed to include the explorative behavior, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chaotic PSO [30].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3 Numerical results Having discussed the fundamental features of the CBO dynamics with memory effects, we now validate Algorithm 1 and compare its performance with plain CBO and PSO algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We will test the methods against several benchmark optimization problems and analyze the impact of the 7 Name Objective function F(x) Search space x∗ F(x∗) Ackley −20 exp � −0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 � 1 d �d i=1 (xi)2 � − exp � 1 d �d i=1 cos (2π(xi)) � + 20 + e [−32, 32]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) 0 Griewank 1 + �d i=1 (xi)2 4000 − �d i=1 cos � xi i � [−600, 600]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) 0 Rastrigin 10d + �d i=1 � (xi)2 − 10 cos (2π(xi)) � [−5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12, 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) 0 Rosenbrock 1 − cos � 2π ��d i=1 (xi)2 � + 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 ��d i=1 (xi)2 [−5, 10]d (1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 1) 0 Salomon 1 − cos � 2π ��d i=1 (xi)2 � + 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 ��d i=1 (xi)2 [−100, 100]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) 0 Schwefel 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20 �d i=1 |xi| [−100, 100]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) 0 XSY random �d i=1 ηi|xi|i, ηi ∼ Unif([0, 1]) [−5, 5]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) 0 XSY 4 ��d i=1 sin2(xi) − e − �d i=1(xi)2� e − �d i=1 sin2 √ |xi| [−10, 10]d (0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , 0) −1 Table 1: Considered benchmark test functions for global optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For each function, the corresponding search space and global solution is given.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' random selection technique on the convergence speed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We also employ 1 to solve problems arising form applications, such as image segmentation and training of a machine learning architectures for function approximation and image recognition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Tests on benchmark problems We test the proposed algorithm against different optimization problems, by considering 8 bench- mark objective functions, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [22], which we report in Table 1 for completeness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The search space dimension is set to d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As in plain CBO methods, we expect the most important parameters are those governing the balance between the exploitative behavior (λ in this case) and the explorative one (σ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In particular, we are interested in the algorithm performance as we change the ratio between λ and σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, in the first experiment we fix λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01, while considering different values of σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The parameter α is adapted during the computation: starting form α0 = 10, it increases according to the law α = α0 · k · log2(k) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 1 shows the accuracy and the objective value reached for σ ∈ [0, 2] after kmax = 104 algorithm iterations with N = 200 particles, with no random selection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The optimal value for σ is clearly problem-dependent, but we note that the optimal values for the problems considered all fall within a relative small range (underlined in gray in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' From Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 1 we infer that a good value for all benchmark problems considered is given by σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Using this value, we now compare CBO-ME, with plain CBO and the standard PSO (with and without alignment towards personal best) for different population sizes N = 50, 100, 200.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We keep the random selection mechanism off by setting µ = 0 and use the same 8 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 2 < 10!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10 10!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 100 105 1010 ky,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='k !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' x$k1 Rastrigin Ackley Griewank Rosenbrock Salomon Schwefel XSY 4 XSY random (a) ∥¯yα,k − x∗∥∞ 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 2 < 10!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10 10!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 100 105 1010 F(7y,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='k) Rastrigin Ackley Griewank Rosenbrock Salomon Schwefel XSY 4 XSY random (b) F(¯yα,k) Figure 1: Optimization on benchmark functions using CBO-ME.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Behavior of the expec- tation error and fitness value for different values of σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Here λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01 and α is adaptive, with α0 = 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The particle population is N = 200.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Grey bands (of values [0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='70, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='05] for the error and [0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='65, 1] for the fitness) show the range in which the minima of the different benchmark functions fall.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The dotted line marks the visually estimate pseudo-optimal value σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results averaged on 250 runs, are obtained with kmax = 104 iterations and without stopping criterion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' previously chosen parameters when memory effects are used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For plain CBO, without memory effects, we set σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='71 ≈ √ 2/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Concerning PSO, we use the solver provided by the MATLAB Global Optimisation Toolbox (particleswarm), changing the maximum number of iterations and the stall condition to the one used for CBO methods, to make the results comparable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The remaining parameters are kept as described in the relative documentation [33].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We set kmax = 104, δstall = 10−4 and consider a run successful when either ∥¯yα,k − x∗∥∞ < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 or |F(¯yα,k) − F(x∗)| < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) Table 2 reports success rate, final error given by ∥¯yα,k −x∗∥∞, mean objective function value and total number of iterations, averaged over 250 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In addition to the classic PSO method, where the acceleration coefficients are chosen to be equal C1 = C2 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='49, Table 2 also shows the results when only the alignment towards global best is considered in PSO (C1 = 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' While CBO already manages to find the global minimizer in most of the problems considered, we note that it fails when Rastrigin, Rosenbrock or XSY random functions are optimized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' CBO- ME, on the other hand, is able to solve the optimization problem correctly even in these cases if the population size N is large enough.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' CBO seems to achieve greater accuracy in some cases, such as with Schwefel 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20 and Salomon objectives, at the cost of more iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Standard PSO in many cases fails to solve the problem, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Rastrigin, Salomon or XSY 4 functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' PSO success rate is also lower among all problems, with the exception of the Schwefel 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20 benchmark problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Considering only global adjustment seems to show advantages with respect to the classical PSO method,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' except in the case of Ackley where setting C1 = 0 decreases the success rate or,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' in the case of XSY 4,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Salomon or Rastrigin,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' where convergence is not achieved even for C1 = 0 Consensus methods,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' however,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' seem to perform better in terms of both success 9 (a) Error: ∥¯yα,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='k − x∗∥∞ (b) Fitness Value: F(¯yα,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='k) Figure 2: Optimization of Ackley function for different values of the random selection parameter µ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' where the initial particle population is N 0 = 104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We report error (on the left) and fitness values (on the right) as the number of function evaluations increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Parameters are set as λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01, σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8, α adaptive starting from α0 = 10 and following the law α = α0 · k · log2(k).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results are averaged over 250 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' rate and speed up.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In addition, for most problems, the population size N seems not to play a significant role in the algorithms performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This further motivates the introduction of the random selection strategy described in the Section 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 in order to save computational costs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In the third experiment, we test the proposed random selection mechanism (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) for different values of the parameter µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We recall that with µ = 0 we have no particles removal, while as µ increases, more particles are likely to be discarded when the system variance decreases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The initial population is set to N0 = 200, while the minimum number of particles to Nmin = 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results are reported in Tables 3 and 4 in terms of: success rate, error, objective value, weighted number of iterations, given by witer = kend � k=1 Nk N0 (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) and percentage of Computational Time Saved (CTS).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results show that relative large values of µ allow to reach fast convergence without affecting the algorithm performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The values of µ considered in Table 4 as different from those in Table 3 as in our experiments, the Rastrigin problem allows for larger values of µ, while the Rosenbrock one seems to be more sensitive to the selection mechanism with respect to the other objectives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In both cases, a suitable value of µ reduces the computational time with almost no impact in terms of accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='s 2 and 3 show error and fitness value as a function of the number of fitness evaluation during the algorithm computation, for the Ackley and Rastrigin problem respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Several values of µ are considered to display how the random selection mechanism affects the convergence speed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Initial particle population is set to N0 = 104 and particles evolve for kmax = 104 iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note how convergence speed increases as µ increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 10 CBO (σ = √ 2/2) CBO-ME (σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) PSO PSO (C1 = 0) N = 50 N = 100 N = 200 N = 50 N = 100 N = 200 N = 50 N = 100 N = 200 N = 50 N = 100 N = 200 Ackley Rate 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6% 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6% 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% Error 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11e-06 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='03e-06 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='55e-06 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='39e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='73e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='74e-06 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='97e-09 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-11 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='16e-12 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='30e-08 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='90e-10 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='96e-13 Favg 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='18e-04 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='81e-05 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='30e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='54e-04 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='96e-05 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='99e-05 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='24e-09 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='65e-11 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='94e-12 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='06e-08 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='70e-10 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01e-13 Iterations 954.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 778.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 678.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 997.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 724.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 626.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9 493.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 420.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 391.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 436.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 390.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Griewank Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6% 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='21e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='24e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='16e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='25e-02 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-02 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='45e-03 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17e-01 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10e-01 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='96e-02 Favg 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='26e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='31e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='47e-02 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='95e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='15e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='82e-02 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='23e-03 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11e-03 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='78-03 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='73e-03 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='71e-03 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='90e-03 Iterations 927.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 777.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 682.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 891.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9 735.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 635.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 436.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 394.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 374.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 427.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 370.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 345.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 Rastrigin Rate 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28e-04 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='83e-04 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-04 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='73e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='27e-04 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='76e-04 Favg 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='51e-06 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='03e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='46e-05 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='54e-06 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='31e-06 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28e-06 Iterations 1083.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 933.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 819.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 1007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 922.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 769.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 Rosenbrock Rate 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6% 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6% 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% Error 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='81e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='48e-02 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='62e-02 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='04e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='78e-02 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='19e-04 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-04 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='67e-04 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44e-02 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='45e-02 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='46e-02 Favg 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13e-03 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='57e-03 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='40e-03 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='26e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='42e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='65e-03 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='80e-02 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='76e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-03 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='95e-04 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='71e-04 Iterations 5772.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 5440.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 5439.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 5955.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 4977.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 4275.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9 4830.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 3322.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 2892.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 5886.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 3419.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 2164.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Schwefel 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20 Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='79e-06 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='23e-07 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44e-07 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='42e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='03e-06 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='76e-07 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-10 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='97e-12 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='58e-14 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='68e-07 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='41e-10 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='03e-14 Favg 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='04e-03 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='15e-04 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='36e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='50e-03 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12e-04 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='37e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='94e-09 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='36e-12 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='52e-13 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44e-07 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='48e-10 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='46e-13 Iterations 814.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 691.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 619.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 670.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 547.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 467.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 484.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 428.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 401.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 593.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 457.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 410.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Salomon Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='87e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28e-02 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='49e-02 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='91e-02 Favg 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14e-01 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='15e-01 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='88e-01 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44e-01 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='86e-01 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='91e-01 Iterations 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 8886.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 9296.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 2456.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 XSY random Rate 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2% 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='64e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='62e-02 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='80e-03 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='06e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='86e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='15e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='25e-01 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-02 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='42e-02 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='23e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-02 Favg 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='95e-08 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='54e-08 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13e-08 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='21e-06 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='85e-08 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17e-08 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='35e-04 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28e-04 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-04 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='22e-04 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11e-04 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='45e-04 Iterations 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 XSY 4 Rate 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='07e-01 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='48e-01 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='16e-01 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44e-01 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='35e-01 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='95e-01 Favg 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='79e-07 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='78e-07 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='46e-07 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='58e-06 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='56e-07 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='43e-07 Iterations 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 9677.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 9128.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 8943.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 Table 2: Comparison between classical CBO, CBO-ME and standard PSO with and with- out alignment towards personal best on benchmark problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The solver particleswarm available in the MATLAB Global Optimisation Toolbox was used for the results concerning the PSO method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Optimal choice of parameters, different for each method, are used for the CBO algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Same stopping criterion and definition of success, see (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2), were used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Performance metric considered: success rate (see (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2)), error (∥¯yα,k−x∗∥∞), fitness value F(¯yα,k) and number of iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results are averaged over 250 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 11 µ = 0 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='05 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Ackley Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='84e-06 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='16e-06 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='54e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-05 Favg 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='30e-05 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='87e-04 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='95e-04 witer 674.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 505.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 357.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 182.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 CTS 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% 51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 % 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8% Griewank Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='35e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='22e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='32e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28e-02 Favg 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='82e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='72e-02 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='70e-02 witer 635.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 395.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9 204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 184.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 CTS 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8% 58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2% 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3% Schwefel 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20 Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='76e-07 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='08e-07 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='21e-07 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='73e-08 Favg 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='37e-05 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='93e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='58e-05 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='74e-05 witer 467.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 359.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 318.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 172.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 CTS 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4% 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9% 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% Salomon Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11e-02 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='35e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='74e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='75e-02 Favg 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='34e-01 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='43e-01 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='07e-01 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='26e-01 witer 2456.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1595.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 1289.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 913.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 CTS 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% 63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% XSY random Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='50e-02 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='62e-02 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='89e-02 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='08e-02 Favg 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='97e-07 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='75e-05 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='48e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='06e-04 witer 10000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0 2642.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 1755.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 1123.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7 CTS 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6% 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4% 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7% XSY 4 Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='30e-01 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='78e-01 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='35e-01 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='37e-01 Favg 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17e-05 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28e-06 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='41e-06 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='55e-06 witer 8943.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 3910.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 1890.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 1060.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 CTS 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9% 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4% Table 3: CBO-ME algorithm with random selection of particles tested against different benchmark functions with different values of µ, which regulates the random selection mech- anism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The system is initialized with N0 = 200 particles and σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Performance metric considered: success rate (see (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2)), error (∥¯yα,k − x∗∥∞), fitness value F(¯yα,k), weighted iteration (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3), and Computational Time Saved (CTS).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results are averaged over 250 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' µ = 0 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 Rastrigin Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14e-05 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12e-05 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='77e-05 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='24e-05 Favg 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='23e-06 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='19e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='98e-06 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='27e-06 witer 1161.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 719.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 256.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 CTS 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9% 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4% µ = 0 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='02 µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='05 Rosenbrock Rate 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4% 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='0% Error 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='55e-02 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='23e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='66e-02 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='341e-02 Favg 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='20e-03 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='23e-03 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10e-03 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='24e-03 witer 3172.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 852.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9 347.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 CTS 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1% 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4% Table 4: CBO-ME algorithm with particle reduction tested against Rastrigin and Rosen- brock functions with an higher diffusion parameter σ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 and for different values of µ , which regulates the random selection mechanism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The system is initialized with N0 = 200 particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Performance metric considered: success rate (see (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2)), error (∥¯yα,k − x∗∥∞), fitness value F(¯yα,k), weighted iteration (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3), and Computational Time Saved (CTS).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 12 (a) Error: ∥¯yα,k − x∗∥∞ (b) Fitness Value: F(¯yα,k) Figure 3: Optimization of Rastigin function for different values of the random selection parameter µ where the initial particle population is N0 = 104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We report error (on the left) and fitness values (on the right) as the number of function evaluations increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Parameters are set as λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01, σ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1, α adaptive starting from α0 = 10 and following the law α = α0 · k · log2(k).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results are averaged over 250 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Applications In this section, we propose some applications of the proposed optimization algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' First we consider a image segmentation problem using multi-thresholding, then we use the CBO- ME to train a Neural Network (NN) architecture to approximate functions and perform image classification on MNIST database of handwritten digits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Image segmentation To perform image segmentation, we use a threshold detection technique, namely, the multidimen- sional Otsu algorithm [32,44] in order to compare the results to similar optimization algorithm, such as the Modified PSO in [43].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In the Otsu algorithm, every pixel of the image is assigned to one of the possible L grayscale values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We denote with ηi the number of pixel with gray level i, 1 ≤ i ≤ L and Npix = �L i=1 ηi the total number of pixels [32].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Then, the image is divided into object C0 with gray-level [1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , l] and background C1 with gray-level [l + 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , L] by inserting a threshold l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The probabilities of class occurrence and the class mean level for the object, respectively, are given by ω0(l) = l � i=1 pi, pi = ηi Npix µ0(l) = l � i=1 ipi ω0(k) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For the background, the class occurrence probabilities and the class mean level are given by ω1(l) = L � i=l+1 pi, pi = ηi Npix 13 µ1(l) = L � i=l+1 ipi ω1(k) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As in [32], the best threshold l∗ is obtained when the variance formula f(l) = ω0(l) ω1(l) (µ0(l) − µ1(l))2 (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) between object group and background reaches its maximum value, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' l∗ = argmaxlf(l).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The problem is then reduced to a threshold problem, which we can solve with optimization methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Since segmentation is a trivial one-dimensional problem, we consider an extension of Otsu’s technique to the multidimensional case [44] to test capabilities of method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assuming we want to optimize the choice of d thresholds, we require d + 1 classes of different gray-scales (C0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , Cd) with relative probabilities of occurrence classes defined as ω0(l1) = l1 � i=1 pi , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , ωd(ld) = L � i=ld+1 pi, pi = ηi Npix and classes mean levels µ0(l1) = �l1 i=1 ipi ω0 , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , µd(ld) = �L i=ld+1 ipi ωd , The optimal thresholds (ˆl1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , ˆld) are those that satisfy ˆl1 < · · · < ˆld and maximise f(l1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' ld) = d � i=1 ωi(li)µ2 i (li) (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5) For the experiment, we chose d = 5 thresholds and compare the segmentation performed by Otsu’s method, solved with both standard PSO and CBO-ME, with segmentation obtained by dividing the greyscale into d + 1 uniformly spaced intervals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For PSO, we use to the default parameters in the particleswarm function in the MATLAB Global Optimisation Toolbox, while for CBO-ME we used optimal parameters found in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 and exploit the random selection technique to speed up the algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We report the results on two sample images, Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='s 4 and 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We fix kmax = 103 and average results over 250 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As in [2], we evaluate multi-thresholding segmentation through the Peak Signal to Noise Ratio (PSNR) computed as: PSNR = 20 · log10 � 255 RMSE � where RMSE is the Root Mean-Squared Error, defined as RMSE = � � � � 1 Npix Nrow � i=1 Ncol � j=1 [I(i, j) − S(i, j)]2 where Npix = Nrow · Ncol, I is the original image and S is the associated segmented image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The higher the value of PSNR is, the greater the similarity between the clustered image and the original image is.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' From Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='s 4,5, we note that the most accurate segmentation on details is obtained by the CBO-ME method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This is quantitatively confirmed by the PSNR values reported in Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 14 (a) Original (b) Standard segmentation (c) Otsu seg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (PSO) (d) Otsu seg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (CBO-ME) Figure 4: Image segmentation of darkhair woman image (256 × 256 pixels) with standard segmentation and Otsu segmentation solved respectively by PSO (c) and by CBO-ME (d);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' results are averaged over 250 runs, with an initial population of N0 = 103 particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (a) Original (b) Standard segmentation (c) Otsu seg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (PSO) (d) Otsu seg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (CBO-ME) Figure 5: Image segmentation of lake image (256×256 pixels) with standard segmentation and Otsu segmentation solved respectively by PSO (c) and by CBO-ME (d);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' results are averaged over 250 runs, with an initial population of N0 = 103 particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' cameraman lake lena peppers woman darkhair Standard segmentation 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='83 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='72 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='35 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='24 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='33 Otsu segmentation (PSO) 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='62 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='33 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='19 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='03 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14 Otsu segmentation (CBO-ME) 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='22 35.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='44 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='72 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='28 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='57 Table 5: PSNR values to evaluating the advantages of the method in optimising threshold values in 5 sample images known in literature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For these results, we compared the Otsu segmentation solved by the proposed CBO-ME method with the classical PSO method with equispaced thresholding segmentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Experiments are performed with d = 5 thresholds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 15 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (a) 2000 epochs 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (b) 3000 epochs 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (c) 5000 epochs 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (d) 8000 epochs Figure 6: Approximating smooth function u1 (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) using a network with n = 50 and m = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The learning rate is λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 and we initially use N0 = 500 particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The others parameters are set as λ = 1, σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 and α adaptive starting from α0 = 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Approximating functions with NN In this section, we use the proposed CBO-ME algorithm to train a NN architecture into approx- imating a function u : I → R, I ⊂ R with low regularity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As in [5], we use a fully-connected NN with m layers f(x;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' θ) = (Lm ◦ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' L2 ◦ L1)(x) (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) where each layer is given by Li = σ(W ix + bi) with σ(x) = 1/(1+exp(−x)) being the sigmoid function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We use internal layers of dimension n, so W 1 ∈ Rn×1, b1 ∈ R, W m ∈ R1×n, bm ∈ Rd and W i ∈ Rn×n for all i = 2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , m − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6), all DNN parameters are collected in θ = {W i, bi}m i=1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As loss function which need to be minimized, we consider the L2-norm between the target function u and its NN approximation f(· ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' θ) F(θ) := ∥f(· ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' θ) − u∥L2(I) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7) Again, similarly to [5], we test the method against the following two functions: u1(x) = sin(2πx) + sin(8πx2) (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) 16 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (a) 2000 epochs 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (b) 3000 epochs 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (c) 5000 epochs 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 2 1 0 1 2 (d) 8000 epochs Figure 7: Approximating non-smooth u2 (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9) function using a network with n = 50, m = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The learning rate is λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 and we use initially N0 = 500 particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The others parameters are set as λ = 1, σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 and α adaptive starting from α0 = 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' u2(x) = � � � � � 1 if x < − 7 8, − 1 8 < x < 1 8, x > 7 8 −1 if 3 8 < x < 5 8, − 5 8 < x < − 3 8, 0 otherwise .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9) We note that u1 is smooth, while u2 is discontinuous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Parameters of the CBO-ME algorithm have been set to λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01, σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8, as in the previous sections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Parameter α is adapted during the computation as in 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 and random selection mechanism is used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We employ m = 3 layers with internal dimension n = 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Results are displayed in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='s 6 and 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that smooth function u1 is well-approximated already after 5000 epochs, while convergence is slower for the discontinuous step function u2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Application on MNIST dataset We now employ the proposed algorithm to train a NN architecture to solve a image classification tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We will consider the MNIST dataset [26] composed of handwritten digits in grayscale with 28 × 28 pixels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For better comparability with CBO methods without memory effects, we closely follow the experiment settings used in the literature [4,10,37], which we summarize below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We consider a 1-layer NN where input images x ∈ R28×28 are first vectorized x �→ vec(x) ∈ R728 and then processed through a fully-connected layer with parameters θ = {W, b}, with 17 10 20 30 40 50 Epochs 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8 1 Accuracy on test data CBO-ME CBO 10 20 30 40 50 Epochs 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='16 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='18 Loss CBO-ME CBO Figure 8: Performance during training of shallow NN (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10) on image classification (MNIST dataset) with CBO-ME optimizer and plain CBO without memory effects [10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Training is performed by Algorithm 1 with N = 100 particles and no particle selection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Cross-entropy loss function (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11) and adaptive parameters strategy (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12) were used in the training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' W ∈ R10×728, b ∈ R10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' That is, the network is given by fSNN(x;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' θ) = softmax (ReLU (Wvec(x) + b) ) , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10) where ReLU(z) = max{z, 0} (component-wise) and softmax(z) = (ez1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , ezn)/(� i ezi) are the commonly activation functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' During the training, batch regularization is performed after ReLU is applied in order to speed up convergence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Given a training set {(xm, ℓm)}M m=1, xm ∈ R28×28, ℓm ∈ {0, 1}10 made of M image-label tuples we train the model by minimizing the categorical cross-entropy loss F(θ) = 1 M M � m=1 � − 10 � i=1 ℓm i log(fi(xm, θ)) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11) We employ a population on N = 100 particles throughout the entire computation, initially sampled from the standard normal distribution N(0, Id).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Following the mini-batch approach suggested in [4], the consensus points ¯yα,k is computed only among a random subset of nN = 10 particles, but all particles are updated at each step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The training data is divided in batches of nF = 60 images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The drift parameter is set to λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='01, while σ and α are adapted during the computation after each epoch as σepoch = σ0/ log2 (epoch + 2) αepoch+1 = 2 · αepoch+1 (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12) starting form σ0 = √ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='04, and α0 = 50, Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 8 shows the results in terms of loss function (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11) over the test data set and the accuracy reached in the classification task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' While challenging state-of-the-art training methods 18 is beyond the scope of the experiment, we note how high-dimensional data optimization tasks can be solved with as little as N = 100 particles by the proposed method, obtaining results comparable with the literature on CBO methods [4, 10, 37].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Also, we remark that parameters have not been tuned extensively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 4 Theoretical analysis A strength of CBO algorithms lays on the possibility of theoretically analyze the particle system by relying on a mean-field approximation of the dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We will illustrate in this section how to derive such approximation and present the main theoretical result regarding the convergence of the particle system towards a solution to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1), in case of no selection mechanism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Next, we will study the impact of the random selection strategy on the convergence properties of the algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Technical details are left to Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Mean-field approximation First, we note that a simple update rule for the personal bests yk i is given by yk+1 i = yk i + 1 2 � xk+1 i − yk i � S(xk+1 i , yk i ) , with S(x, y) = 1 + sign (F(y) − F(x)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) As in [14], we approximate it for β ≫ 1 as yk+1 i = yk i + ν 2 � xk+1 i − yk i � Sβ(xk+1 i , yk i ) , (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) with Sβ(x, y) being a continuous approximation of S(x, y) as β → ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By choosing ν = 1 we get (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) with the only difference of having Sβ instead of S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As for ¯yα with respect to ¯y∞, this is needed to make the update rule easier to handle mathematically, but it does have an impact on the performance for large values of β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' With the aim of deriving a continuous-in-time reformulation of (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) and (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2), we introduce a single parameter ∆t > 0 which controls the step length of all involved update mechanisms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By performing the rescaling λ ← λ∆t , σ ← σ √ ∆t , ν ← ν∆t to get the update rules � xk+1 i = xk i + λ∆t � ¯yα,k − xk i � + σ √ ∆t � ¯yα,k − xk i � ⊗ θk i yk+1 i = yk i + (ν∆t/2) � xk+1 i − yk i � Sβ(xk+1 i , yk i ) (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) which differ form the original formulation (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4), (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) only due to the use of Sβ instead of S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As already noted in [14], the iterative process (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) corresponds to an Euler-Maruyama scheme applied to a system of Stochastic Differential Equations (SDEs).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) corre- sponds to a discretization of the system � dXi t = λ � ¯yα(ρN t ) − Xi t � dt + σ � ¯yα(ρN t ) − Xi t � ⊗ dBi t dY i t = ν(Xi t − Y i t )Sβ(Xi t, Y i t ) dt (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) 19 where, for convenience, we underlined above the dependence of the consensus point on the empirical distribution ρN t = � i δY i t (δy being the Dirac measure at y ∈ Rd) by using ¯yα(ρ) := � ye−αF(y)dρ(y) � e−αF(y)dρ(y) , (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5) defined for any Borel probability measure ρ over Rd (ρ ∈ P(Rd)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In this way, we generalized the definition introduced in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) to any ρ ∈ P(Rd), provided the above integrals exists.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4), the random component of the dynamics is now described by N independent Wiener processes (Bi t)t>0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As before, we supplement the system with initial conditions Xi 0 ∼ ρ0, Y i 0 = Xi 0 for some ρ0 ∈ P(Rd).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The continuous-in-time description (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) already simplifies the analytical analysis of the optimization algorithm, but still pays the price of a possible large number O(N) of equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This issue is typically addressed by assuming that for large populations N, the particles become indistinguishable from one another and start behaving, in some sense, as a unique system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' More precisely, let F N(t) ∈ P(R(2d)N) denote the joint probability distribution of N tuples (Xi t, Y i t ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We assume propagation of chaos [41] for large N ≫ 1, that is, we assume that the joint probability distribution decomposes as F N(t) = f(t)⊗N for some f(t) ∈ P(R2d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' System (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) becomes independent on the index i and hence every particle moves according to the mono-particle process d ¯Xt = λ(¯yα(¯ρt) − ¯Xt) dt + σ (¯yα(¯ρt) − ¯Xt) ⊗ d ¯Bt d ¯Yt = ν( ¯Xt − ¯Yt)Sβ( ¯Xt, ¯Yt) dt (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) where ¯ρt = Law( ¯Yt).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assume ( ¯Xt, ¯Yt) are initially distributed according to f0 = ρ⊗2 0 , by applying Itˆo formula we have that f(t) = Law( ¯Xi t, ¯Y i t ) satisfies ∂tf + ∇x · (λ(¯yα(¯ρ) − x)f) + ∇y · � ν(x − y)Sβ(x, y)f � = d � ℓ=1 ∂2 xℓ � σ(¯yα(¯ρ) − x)2 ℓf � (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7) and initial data limt→0 f(t) = f0 in a weak sense.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Dynamics (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6), or, equivalently, (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7), corresponds to the mean-field approximation of the particle system (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) as N → ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We remark that the above derivation has only been possible thanks to the approximations S ≈ Sβ and ¯y∞ ≈ ¯yα for large α and β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Well-posedness of the system is also granted by such approximations (proof details are given in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 (well-posedness of (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' There exists a unique process ( ¯X, ¯Y ) ∈ C([0, T], Rd), T > 0 satisfying (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) with initial conditions ( ¯X0, ¯Y0) with ¯X0 ∼ ρ0 ∈ P4(Rd) and ¯Y0 = ¯X0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Being mathematically tractable, we show next that the mean-field dynamics converges to a global solution to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) if F, Sβ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 20 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Convergence in mean-field law We start by enunciating the necessary assumptions to the convergence result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 (Assumptions on F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The objective function F ∈ C(Rd, R), satisfies: A1 there exists some constant LF > 0 such that |F(x) − F(x′)| ≤ LF � ∥x∥2 + ∥x′∥2 � ∥x − x′∥2, ∀ x, x′ ∈ Rd ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A2 there exists uniquely x∗ ∈ Rd solution to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A3 there exist η, R0 > 0 and γ ∈ (2, ∞) such that F(x) − inf F ≥ η ∥x − x∗∥γ ∞ ∀x ∈ Rd , ∥x − x∗∥∞ ≤ R0 F(x) − inf F ≥ η Rγ 0 ∀x ∈ Rd , ∥x − x∗∥∞ > R0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A4 F is convex in a (possibly small) neighborhood {x ∈ Rd : ∥x − x∗∥∞ ≤ R1} of x∗ for some R1 < R0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A5 There exists cg, R2 > 0 such that F(x) − inf F ≥ cg∥x − x∗∥2 2 ∀x ∈ Rd , ∥x − x∗∥2 > R2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 (Assumptions on Sβ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The function Sβ ∈ C(R2d, [0, 2]), with β > 0 A6 has the following structure Sβ(x, y) = 2ψ (β(F(y) − F(x))) , (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) with ψ ∈ C1(R, [0, 1]) being an increasing function with Lipschitz constant Lψ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A7 The value Sβ(x, y) is positive only when x is strictly better than y in terms of objective value F: Sβ(x, y) � ≥ 0 if F(x) < F(y) = 0 else .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assuming uniqueness of global minimum is a typical assumption for analysis of CBO methods [9,10] and it is due to the definition of the consensus point ¯yα (or ¯xα in the case without memory mechanism).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Indeed, in presence of two global minima, ¯yα may be placed between them, no matter how large α is.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assumption A2 ensure to avoid such situations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Furthermore, A3 also allows to give quantitative estimates on the difference between the global minimum and eventual local minima.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In the literature, such property is known as conditioning [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Requirements A4 and A7 ensure that if a personal best yk i enters such small neighborhood where F is convex, it will not leave it for the rest of the computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Condition A5 (quadratic growth at infinity) is needed for the well-posedness of the mean-field mono-particle process (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6), see also [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For an intuition of A3 and A4 we refer to Figure 9, where the Rastrigin function is considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 21 x$ !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' R0 x$ x$ + R0 0 20 40 objective lower bound (A3) convex area (A4) Figure 9: Assumptions 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 illustrated for Rastrigin function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For example, such objective function satisfies A3 with η = 1, γ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8, R0 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='42 and A4 with R1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 (Convergence in mean-field law).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assume F satisfies A1–A5, Sβ satisfies A6, A7 for some β > 0 fixed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let ( ¯Xt, ¯Yt)t≥0 be a solution to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) for t ∈ [0, T], with initial data ¯X0 ∼ ρ0 ∈ P4(Rd), Y0 = X0 such that x∗ ∈ supp(ρ0) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fix an accuracy ε > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' If 2λ > σ2, there exists a time T ∗ such that the expected ℓ2-error satisfies E � ∥ ¯XT ∗ − x∗∥2 2 � ≤ ε (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='9) provided T, α > 0 are large enough.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We refer to Appendix A for a proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The mean-field mono-particle process (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) aims to approximate the algorithm iterative dynamics (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) for small time steps ∆t ≪ 1 and large particle populations N ≫ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, convergence of the algorithm dynamics towards the global solution x∗ can be proven by coupling Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 with error estimates of such approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For instance, assuming that all considered dynamics take place on a bounded set D ensures that the error introduced by the continuous-in-time particle system will be of order ∆t thanks to classical results on Euler-Maruyama schemes [35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Likewise, considering a bounded dynamics allows to prove that the error introduced by the mean-field approximation is of order N−1 (see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [8, Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1], [9, Proposition 16]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let {(xk i , yk i )}N i=1 be given by (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3), {(Xi t, Y i t )}N i=1 be a solution (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) and {( ¯Xi t, ¯Y i t )}N i=1 be N-copies of a solution to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Altogether, one obtains the following error decomposition for K∆t = T ∗ E � 1 N N � i=1 ∥xK i − x∗∥2 2 � ≤ C � E � 1 N N � i=1 ∥xK i − Xi T ∗∥2 2 � + E � 1 N N � i=1 ∥Xi T ∗ − ¯Xi T ∗∥2 2 � + E � 1 N N � i=1 ∥ ¯Xi T ∗ − x∗∥2 2 � � ≤ CEM∆t + CMFAN−1 + ε 22 where C, CEM, CMFA are positive constant independent on N, ∆t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Random selection analysis In this section, we analytically investigate the impact of randomly discarding particles during the computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We are particularly interested in tracking the distance between a particle system {xk i , xk j }N0 i=1 evolving according to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) where no particles are discarded, and a second system {ˆxk i , ˆyk i }Ik, |Ik| = Nk where Nk − Nk+1 particles are discarded after update rule (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Clearly, we have that Nk+1 ≤ Nk and Ik+1 ⊆ Ik ⊆ I0 = {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , N0} for all k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Similarly to the analysis carried out in [15,16], we restrict to the simpler dynamics where, at every step k, the random variables θk i and ˆθk i used to generate such systems are the same for all particles: θk i = ˆθk j = θk ∼ N(0, Id) for all i ∈ Ik, j ∈ I0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10) To compare particle systems with a different number of particles, we rely on their represen- tation as empirical probability measures and the notion of 2-Wasserstein distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For {ˆxk i }i∈Ik and {xk i }N0 i=1 we consider, respectively, the following probability measures ρk Nk := 1 Nk � i∈Ik δˆxk i and ρk N0 := 1 N0 � i∈I0 δxk i .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11) Informally, the 2-Wasserstein distance W2(ρk Nk, ρk N0) quantifies the minimal effort needed to move the mass from distribution ρk Nk into ρk N0 (or vice versa) [38].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let wij denote the amount of mass leaving particle xk i and going into ˆxk i : the cost of such movement is assumed to be given by wij∥xk i − ˆxk j ∥2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, if we indicate the set of all admissible couplings between the two discrete probability measures as Γ(ρk Nk, ρk N0) = � � �w ∈ RN0×Nk : Nk � j=1 wij = 1 N0 , N0 � i=1 wij = 1 Nk , wij ≥ 0, ∀ i, j � � � , (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12) the 2- Wasserstein distance is defined as W2(ρk Nk, ρk N0) := min w∈Γ(ρk Nk,ρk N0) � �� i,j wij∥xk i − ˆxk j ∥2 2 � � 1 2 (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13) see, for instance, [38, Section 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Before providing estimates on (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12), let us present a more general result on the impact that the random selection strategy has on an arbitrary particle distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 (Stability of random selection procedure).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let z = {zi}i∈I, |I| = N be an ensemble of particles and {zi}j∈Isel with Isel ⊆ I, |I| = Nsel a random sub-set of such ensemble.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consider the associated empirical distributions µN and µNsel (defined consistently to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11)), it holds E � W 2 2 (µN, µNsel) � ≤ 2 var(z) N − Nsel N − 1 (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14) where the expectation is taken with respect to the random selection of Isel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 23 The proof is provided Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note how the system variance var(z) enters the error estimate due to the randomness of the selection, similar to the Law of Large Number error for random variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In particular, the smaller the particles variance is, the closer the reduced particle system will be to the original distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This justifies the choice of Nk+1 proposed in Section 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 where we are allowed to discard particles only if the system shows a contractive behavior, see (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By iteratively applying Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 and by using suitable stability estimates of dynamics (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3), we are able to bound the error introduced by the random selection procedure as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof details are a given in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let {xk i , yk i }N0 i=1 be constructed according to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) were particles are not discarded, and {ˆxk i , ˆyk i }Ik, |Ik| = Nk where Nk−Nk+1 particles are discarded after update rule (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assume (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10) is satisfied and consider the probability measures (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' If {xk i , yk i }N0 i=1, {ˆxk i , ˆyk i }i∈Ik ⊂ BM(0) at all step k for some M > 0, it holds E � W 2 2 � ρk Nk, ρk N0 �� ≤ C max h=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=',k var � ˜zh� N0 − Nk Nk − 1 (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='15) where C = C(∆t, λ, σ, ν, β, α, k, LF, M) and ˜zh = {(ˆxh i , ˆyh i )}i∈Ih−1 describes the particle system just before the random selection procedure at step h ≤ k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The expectation is taken with respect to the sampling of {θh}k h=1 and with respect to the selection procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We can directly apply the above result to relate the expected ℓ2-errors of the two particle system, which we define as Err(k) = E � � 1 N0 � i∈I0 ∥xk i − x∗∥2 2 � � , Err(k) = E � � 1 Nk � i∈Ik ∥ˆxk i − x∗∥2 2 � � , that is, the discrete counterpart of the mean-field error E[∥ ¯Xi t − x∗∥2 2] studied in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By definition of the Wasserstein-2 distance, we have Err(k) = E � W 2 2 (ρk N0, δx∗) � for any solution x∗ to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1), and the same holds of Errsel(k).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We then apply inequality W 2 2 (ρk Nk, δx∗) ≤ 2 � W 2 2 (ρk Nk, ρk N0) + W 2 2 (ρk N0, δx∗) � to obtain the following estimate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Corollary 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Under the assumptions of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2, at all steps k, it holds Errsel(k) ≤ 2 � Err(k) + C max h=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=',k var(˜zh) N0 − Nk Nk − 1 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='16) Before concluding the section, let us report some remarks concerning the theoretical results just presented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 24 Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 can be adapted to any other particle system with random selection, provided that the update rule is stable with respect to the 2-Wasserstein distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In the proposed method, such stability was proved thanks to the approximation of the global best ¯y∞,k with ¯yα,k for α ≫ 1 (see (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2)) and S(x, y) with Sβ(x, y) for β ≫ 1 in the personal best update (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Quantitative estimates on the variance decay can be used, if available, to improve the error bound in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2, see also proof in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The error introduced by a sub-sampling technique in a Monte Carlo integral approximation is expected to be of order 2 var(z) � 1 N − 1 − 1 Nsel − 1 � = 2 var(z) N − Nsel (N − 1)(Nsel − 1) , (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17) see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [23].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, an additional factor of order 1/(Nsel − 1) seems to be missing in Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We remark, though, that Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 does not concern the Monte Carlo approximation of an integral quantity, but rather consider the 2-Wasserstein distance between discrete measures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Numerical simulations suggest that estimates of order (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17) do not hold on in this case, see Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 5 Conclusions In this work, we studied a Consensus-Based Optimization algorithm with Memory Effects (CBO- ME) and random selection for single objective optimization problems of the form (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' While sharing common features with Particle Swarm Optimization (PSO) methods, CBO-ME differs on the way the particle system explore the search space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Its structure provides greater flexi- bility in balancing the exploration and exploitation processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In particular, we implemented and analytically investigates a random selection strategy which allows to reduce the algorithm computational complexity, without affecting convergence properties and overall accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This analysis is entirely general and, in perspective, applicable to other particle swarm-based opti- mization methods as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The convergence analysis to the global minimum is carried out by relying on a mean-field approximation of the particle system and error estimates are given un- der mild assumptions on the objective function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We compared CBO-ME against CBO without memory effects and PSO against several benchmark problem and showed how the introduction of memory effects and random selection improves the algorithm performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Applications to image segmentation and machine learning problems are finally reported.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A Proofs A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Notation and auxiliary lemmas We will use the following notation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For any a ∈ R, |a| indicates the absolute value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For a given vector b ∈ Rd, ∥b∥p indicates its p-norm, p ∈ [1, ∞];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (b)ℓ its ℓ-th component;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' while diag(b) ∈ Rd×d 25 0 20 40 60 80 100 # particle selected !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Nsel " 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5 2 N = 100;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' d = 3 squared Wasserstein dist.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' estimate (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='18) estimate (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='21) 0 20 40 60 80 100 # particle selected !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Nsel " 0 1 2 3 4 5 6 7 N = 100;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' d = 10 squared Wasserstein dist.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' estimate (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='18) estimate (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='21) Figure 10: Numerical validation of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 with different dimensions d = 3, 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' N = 100 points are randomly, uniformly sampled over [0, 1]d to construct the empirical distribution µN and Nsel ∈ [2, N − 1] are discarded to obtain µNsel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The experiment is repeated 500 times for all Nsel to obtain an approximation of E � W 2 2 (µN, µNsel) � (blue line).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In red, estimate provided by Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 (RHS of (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='14)), in yellow the one given equation (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='17).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Wasserstein distances are computed with the ot.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='emd function provided by the Python Optimal Transport library [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' is the diagonal matrix with elements of b on the main diagonal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let a, b ∈ Rd, ⟨a, b⟩ denotes the scalar product in Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For a given closed convex set A ⊂ Rd, N(A, x), T (A, x) denote the normal and the tangential cone at x ∈ A respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The ball or radius r centered at x ∈ Rd is indicated with Br(x) = {x ∈ Rd | ∥x∥2 ≤ r}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' All considered stochastic processes are assumed to take their realizations over the common probability space (Ω, ¯F, P).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' P(Rd) is the set of Borel probability measures over Rd and Pq(Rd) = {µ ∈ P(Rd) | � ∥x∥q 2dµ < ∞} which we equip with the Wasserstein distance Wq, q ≥ 1, see [38].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For a random variable X, X ∼ µ, µ ∈ P(Rd) indicates a sampling procedure such that P(X ∈ A) = µ(A) for any Borel set A ⊂ Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' With Unif(A) ∈ P(Rd) we denote the uniform probability measure over a bounded Borel set A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Throughout the computations, C will denote an arbitrary positive constant, whose value may vary from line to line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Dependence on relevant parameters or variables, will be underlined.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 ( [3, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let F satisfy Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 (in particular the locally Lipschitz assumption A1) and ρ1, ρ2 ∈ P4(Rd) with � ∥x∥4 2dρ1 , � ∥x∥4 2dρ2 ≤ M .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Then, the following stability estimate holds ∥¯yα(ρ1) − ¯yα(ρ2)∥2 ≤ C W2(ρ1, ρ2) for a constant C = C(α, LF, M).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 26 Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Under Assumptions A1 and A6, for any x1, x2, y1, y2 ∈ BM(0) and β > 0, it holds ∥(x1 − y1)Sβ(x1, y1) − (x2 − y2)Sβ(x2, y2)∥2 ≤ C (∥x1 − y1∥2 + ∥x2 − y2∥2) where C = C(β, LF, M).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Thanks to the Lipschitz continuity of ψ, F and the choice of ψ (Assumptions A1 and A6), it holds |Sβ(x1, y1) − Sβ(x2, y2)| = |2ψ (β(F(y1) − F(x1)) − 2ψ(β(F(y2) − F(x2)) | ≤ 2β |F(y1) − F(x1) − F(y2) + F(x2)| ≤ 2βLF (∥x1 − x2∥2 + ∥y1 − y2∥2) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Next, we have ∥(x1 − y1)Sβ(x1, y1) − (x2 − y2)Sβ(x2, y2)∥2 ≤ ∥(x1 − y1)Sβ(x1, y1) − (x2 − y2)Sβ(x1, y1)∥2 + (x2 − y2)Sβ(x1, y1) − (x2 − y2)Sβ(x2, y2)∥2 ≤ ∥(x1 − x2 + y2 − y1)Sβ(x1, y1)∥2 + ∥(x2 − y2) � Sβ(x1, y1) − Sβ(x2, y2) � ∥2 ≤ 2 (∥x1 − x2∥2 + ∥y1 − y2∥2) + 2M|Sβ(x1, y1) − Sβ(x2, y2)| ≤ C (∥x1 − x2∥2 + ∥y1 − y2∥2) with C = C(β, LF, M), where we used the first estimate to conclude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The proof is based on the Leray–Schauder fixed point theorem [13, Chapter 11], and we follow closely the proof steps of [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Step 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For any ξ ∈ C([0, T], Rd) there exists a unique process ( ˆXt, ˆYt) ∈ C([0, T], Rd) satisfying d ˆXt = λ(ξ(t) − ˆXt) dt + σ(ξ(t) − ˆXt) ⊗ d ˆBt d ˆYt = ν( ˆXt − ˆYt)Sβ( ˆXt, ˆYt) dt with Law( ˆX0) = Law( ˆY0) = ρ0 ∈ Rd, by the Lipschitz continuity of the coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As a consequence, we have that f(t) := Law( ˆXt, ˆYt) satisfies d dt � φ df(t) = � � −λ⟨∇xφ, ξ(t) − x⟩ + � ℓ=1 ∂2φ ∂x2 ℓ (ξt) − y)2 ℓ − νSβ⟨∇yφ, y − x⟩ � df(t) for all φ ∈ C2 b (R2d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, let ¯ρ(t) = Law( ˆYt), we can set T ξ := ¯yα(¯ρ(·)) ∈ C([0, T], Rd) to define T : C([0, T], Rd) → C([0, T], Rd).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 27 Step 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We prove now compactness of T .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Thanks to ρ0 ∈ P4(Rd) and standard results for SDEs (see [1, Chapter 7]) we have boundedness of the forth moments E � ∥ ˆXt∥4 2 + ∥ ˆYt∥4 2 � ≤ c1 � 1 + E[∥ ˆX0∥4 2 + ∥ ˆY0∥4 2]ec2t� for some c1, c2 > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, we can apply Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 to obtain for any 0 < s < t < T, ∥¯yα(¯ρ(t)) − ¯yα(¯ρ(s))∥2 ≤ CW2 (¯ρ(t), ¯ρ(s)) ≤ ˜C|t − s|1/2 for some constants C, ˜C > 0, from which H¨older continuity of t �→ ¯yα(¯ρ(t) follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, by T (C([0, T], Rd)) ⊂ C0, 1 2 ([0, T], Rd) �→ C([0, T], Rd) we get compactness of T .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Step 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consider ξ ∈ C([0, T], Rd) satisfying ξ = τT ξ, for τ ∈ [0, 1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Thanks to [3][Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3] and boundedness of second moments, we obtain compactness of the set {ξ ∈ C([0, T], Rd) : ξ = τT ξ, τ ∈ [0, 1]} and by Leray–Schauder fixed point theorem there exists a fixed point for the mapping T and hence a solution to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Step 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Assume now there exist two solutions, ( ¯X1 t , ¯Y 1 t ) and ( ¯X2 t , ¯Y 2 t ) to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) with same Brownian process ¯Bt and initial conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let ¯ρℓ = Law( ¯Y ℓ t ), ℓ = 1, 2, we have ∥ ¯X1 t − ¯X2 t ∥2 2 = � t 0 � ¯X1 s − ¯X2 s , ¯yα(¯ρ1(s)) − ¯yα(¯ρ2(s)) − ¯X1 s + ¯X2 s � dt + � t 0 � diag � ¯yα(¯ρ1(s)) − ¯X1 s � − diag � ¯yα(¯ρ2(s)) − ¯X2 s �� d ¯Bs .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1) We note that all terms can be estimated by means of W 2 2 (¯ρ1(s), ¯ρ2(s)) and ∥ ¯X1 s − ¯X2 s ∥2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Similarly, ∥ ¯Y 1 t − ¯Y 2 t ∥2 2 can be bounded in terms ∥ ¯X1 s − ¯X2 s ∥2 2 thanks to the Lipschitz continuity of Sβ and Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, for some constant C > 0 ∥ ¯X1 t − ¯X2 t ∥2 2 + ∥ ¯Y 1 t − ¯Y 2 t ∥2 2 ≤ C � t 0 � ∥ ¯X1 s − ¯X2 s ∥2 2 + ∥ ¯Y 1 s − ¯Y 2 s ∥2 2 + W 2 2 (¯ρ1(s), ¯ρ2(s)) � ds from which, together with (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1), follows for some ˜C > 0 E � ∥ ¯X1 t − ¯X2 t ∥2 2 + ∥ ¯Y 1 t − ¯Y 2 t ∥2 2 � ≤ E � ∥ ¯X1 0 − ¯X2 0∥2 2 + ∥ ¯Y 1 0 − ¯Y 2 0 ∥2 2 � e ˜C t by Gr¨onwall’s inequality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Since E � ∥ ¯X1 0 − ¯X2 0∥2 2 + ∥ ¯Y 1 0 − ¯Y 2 0 ∥2 2 � = 0, we proved uniqueness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 Proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 Having proved there exists a solution ( ¯Xt, ¯Yt)t∈[0,T] to the mean-field process (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) we are here interested in studying the expected ℓ2-error given by E∥ ¯Xt − x∗∥2 2 where x∗ is the unique solution to the minimization problem (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1), see Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We do so by means of the following quantitative version of the Laplace principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 28 Proposition A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 (quantitative Laplace principle [10, Proposition 1]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let ρ ∈ P(Rd) be such that x∗ ∈ supp(ρ) and fix α > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For any r > 0, define Fr = supx∈B∗r F(x) − F(x∗) with B∗ r := {x | ∥x − x∗∥∞ ≤ r} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Then, under Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1, for any r ∈ (0, R0] and q > 0 such that q + Fr ≤ F∞ = ηRγ 0, it holds ∥yα(ρ) − x∗∥2 ≤ √ d(q + Fr)γ η + √ d exp(−αq) ρ(B∗r) � ∥x − x∗∥2 dρ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) We remark that RHS of (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2) can be made arbitrary small by taking large values of α and small values of q, r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' To apply Proposition A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 to all ¯ρ(t) = Law( ¯Yt), we need though to provide lower bounds on ¯ρ(t)(B∗ r) for any small radius r and times t ∈ [0, T].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let ¯ρ(t) = Law( ¯Yt), with ¯Yt evolving according to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) and limt→0 ¯ρ(t) = ρ0 with x∗ ∈ supp(ρ0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Under Assumptions 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 , it holds ¯ρ(t)(B∗ r) ≥ mr > 0, for all t ∈ [0, T] and for all r ≤ R0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let δ = η min{R1, r}γ, we start by proving that the mass in the set Lδ = {x ∈ Rd | F(x) ≤ inf F + δ} is non-decreasing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that for this choice of δ, Lδ is convex due to Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consider now (Ω, ¯F, P) to be the common probability space over which the considered processes take their realization and define Ωδ = {ω : ¯Y0(ω) ∈ Lδ}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2, Sβ( ¯Xt(ω), ¯Yt(ω)) = 0 whenever ¯Xt(ω) /∈ Lδ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, it holds � ( ¯Xt(ω) − ¯Yt(ω))Sβ( ¯Xt(ω), ¯Yt(ω)) , n( ¯Yt(ω)) � � = 0 if ¯Xt(ω) /∈ Lδ ≤ 0 if ¯Xt(ω) ∈ Lδ for ¯Yt(ω) ∈ ∂Lδ for any n( ¯Yt(ω)) ∈ N(Lδ, x) from which follows that ¯Yt(ω) solves ¯Yt(ω) = ¯Y0(ω) + � t 0 ΠT (Lδ, ¯Ys(ω)) � ( ¯Xs(ω) − ¯Ys(ω))Sβ( ¯Xs(ω), ¯Ys(ω)) � ds for all ω ∈ Ωδ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As a consequence, if ¯Y0(ω) ∈ Lδ, ¯Yt(ω) ∈ Lδ for all t ≥ 0 and so ¯ρ(t)(B∗ r) = P( ¯Yt ∈ Lδ) ≥ P( ¯Y0 ∈ Lδ) =: mr for all t ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We conclude by noting that mr > 0 since x∗ ∈ supp(ρ0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Next, we study the evolution of the error E∥ ¯Xt − x∗∥2 2 and, in particular, we try to bound it in terms of ∥¯yα(¯ρ(s)) − x∗∥2 and E∥ ¯Xt − x∗∥2 itself for s ∈ [0, t].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 29 Proposition A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [10, Lemma 1] Let ( ¯Xt, ¯Yt) ∈ C([0, T], R2d) be the solution to (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) with initial datum ¯X0 ∼ ρ0, ¯Y0 = ¯X0 for some time horizon T > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For all t ∈ [0, T], it holds E∥ ¯Xt − x∗∥2 2 ≤ � t 0 � − (2λ − σ2)E∥ ¯Xs − x∗∥2 2 + √ 2(λ + σ2)E∥ ¯Xs − x∗∥2∥¯yα(¯ρ(s)) − x∗∥2 + σ2 2 ∥¯yα(¯ρ(s)) − x∗∥2 2 � ds (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) where ¯ρ(t) = Law( ¯Yt).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The above result, together with Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3, leads to the convergence in mean-field law of the dynamics towards the solution to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The proof can be carried out exactly as in [10, Theorem 12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 and Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 We start by collecting a preliminary result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let {xk 1,i, yk 1,i}N1 i=1 and {xk 2,j, yk 2,j}N2 j=1 be two particle populations generated through update rules (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) with θk 1,i = θk 2,j = θk for all i, j and k ∈ Z+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' At any iteration step k and for any couple of indexes (i, j), it holds E � ∥xk+1 1,i − xk+1 2,j ∥2 2 + ∥yk+1 1,i − yk+1 2,j ∥2 2 � ≤ CE � ∥xk 1,i − xk 2,j∥2 2 + ∥yk 1,i − yk 2,j∥2 2 + ∥¯yα(¯ρk 1) − ¯yα(¯ρk 2)∥2 2 � where C = C(∆t, λ, σ, ν, β) is a positive constant and ¯ρk 1, ¯ρk 2 are the empiricial distributions associated with {yk 1,i}N1 i=1 and {yk 2,j}N2 j=1 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For all k ∈ Z+ and i,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' j E∥xk+1 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i − xk+1 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j ∥2 2 ≤ E ���xk 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i + λ∆t � ¯yα(¯ρk 1) − xk 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i � + σ √ ∆t � ¯yα(¯ρk 1) − xk 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i � ⊗ θk 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i − � xk 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j + λ∆t � ¯yα(¯ρk 2) − xk 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j � + σ √ ∆t � ¯yα(¯ρk 2) − xk 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j � ⊗ θk 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j � ��� 2 2 ≤ 2E ��� � 1 − λ∆t − σ √ ∆t diag(θk) � (xk 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i − xk 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j) ��� 2 2 + 2E ��� � λ∆t + σ √ ∆t diag(θk) � � ¯yα(¯ρk 1) − ¯yα(¯ρk 2) ���� 2 2 ≤ 2(1 + σ2∆t)E∥xk 1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='i − xk 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='j∥2 2 + 2(λ2∆t2 + σ2∆t)E∥¯yα(¯ρk 1) − ¯yα(¯ρk 2)∥2 2 ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) where we also used that E[(θk)2 ℓ] = 1 for all ℓ = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' , d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We now bound ∥yk+1 1,i − yk+1 2,j ∥2 2 as E∥yk+1 1,i − yk+1 2,j ∥2 2 ≤ E ���yk 1,i + (ν∆t/2) � xk+1 i,1 − yk 1,i � Sβ(xk+1 1,i , yk 1,i) 30 − � yk 2,j + (ν∆t/2) � xk+1 2,j − yk 2,j � Sβ(xk+1 2,j , yk 2,j) � ��� 2 2 ≤ CE � ∥xk+1 i,1 − xk+1 j,2 ∥2 2 + ∥yk i,1 − yk j,2∥2 2 � (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5) where we used Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 and C = C(∆t, β, ν).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By combining (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4) and (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5) we get the desired estimate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Next, we show how the particle update rule (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) is stable with respect to the 2-Wasserstein distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proposition A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 (Stability of update rule (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let {xk 1,i, yk 1,i}N1 i=1, {xk 2,j, yk 2,j}N2 j=1 ⊂ BM(0), for some M > 0, be two particle populations generated through the update rules (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) with θk 1,i = θk 2,j = θk for all i, j and k ∈ Z+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let µk 1, µk 2 ∈ P(R2d) the empirical probability measures defined as µk 1 := 1 N1 N1 � i=1 δ(xk 1,i,yk 1,i) , µk 2 := 1 N2 N2 � j=1 δ(xk 2,j,yk 2,j) , it holds E � W 2 2 (µk+1 1 , µk+1 2 ) � ≤ C1 E � W 2 2 (µk 1, µk 2) � , where C1 = C1(∆, λ, σ, ν, α, β, LF, M) is positive constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let Eθk[·] denote the expectation taken with respect to the sampling of θk only and w ∈ RN1×N2 be the optimal coupling between µk 1, µk 2, see (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12) and (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Being w a sub- optimal coupling for µk+1 1 , µk+1 2 , it holds Eθk[W 2 2 (µk+1 1 , µk+1 2 )] ≤ Eθk � i,j wij � ∥xk+1 1,i − xk+1 2,j ∥2 2 + ∥yk+1 1,i − yk+1 2,j ∥2 2 � ≤ C � i,j wij � ∥xk 1,i − xk 2,j∥2 2 + ∥yk 1,i − yk 2,j∥2 2 � + ∥¯yα(¯ρk 1) − ¯yα(¯ρk 2)∥2 2 where we used the linearity of the expectation, estimates given by Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4 and, to take the last term out of the sum, the fact that � ij wij = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' To estimate the distance between the two consensus points, we use Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1 and note that the coupling w is sub-optimal for ¯ρk 1, ¯ρk 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By Lemma A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='1, it follows ∥¯yα(¯ρk 1) − ¯yα(¯ρk 2)∥2 2 ≤ CW 2 2 (¯ρk 1, ¯ρk 2) ≤ C � i,j wij∥yk 1,i − yk 2,j∥2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, Eθk[W 2 2 (µk+1 1 , µk+1 2 )] ≤ C1 � i,j wij � ∥xk 1,i − xk 2,j∥2 2 + ∥yk 1,i − yk 2,j∥2 2 � = C1 W 2 2 (µk 1, µk 2) , thanks to the optimality of w, with C1 = C1(∆, λ, σ, ν, α, β, LF, M) being a positive constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' One can conclude by taking the expectation of the above inequality with respect to the remaining sampling processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 31 We now quantify the impact of the particle discarding step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' For notational simplicity, let us introduce zi = (xi, yi) ∈ R2d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' As in (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='13), the 2-Wasserstein distance is given by an optimal coupling between the full particle system {zi}i∈I and the reduced one {zj}j∈Isel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We consider the following transportation of mass from µN to µNsel: if particle i has not been discarded, all its mass remains in xi, otherwise the mass is uniformly distributed among the selected particles to generate an admissible coupling w ∈ RN×Nsel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' This means that w is given by wij = � � � � � 1/N if j = i, i ∈ Isel 1/(N · Nsel) if i ∈ I \\ Isel, j ∈ Isel 0 else .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='6) We note that such coupling w satisfies the coupling conditions � j∈Isel wij = 1 N � i∈I wij = 1 Nsel , ∀ i ∈ I, j ∈ Isel (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='7) and that this choice will be in general sub-optimal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, it holds W 2 2 (µN, µNsel) ≤ � i∈I, j∈Isel wij∥zi − zj∥2 2 = 1 N � i∈Isel ∥zi − zi∥2 2 + 1 N · Nsel � i∈I\\Isel, j∈Isel ∥zi − zj∥2 2 = 1 N · Nsel � i,j∈I ∥zi − zj∥2 2 1i∈I\\Isel 1j∈Isel where 1i∈I = 1 if i ∈ I and zero otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Now, the probability of having i ∈ I \\ Isel is given by (N − Nsel)/N, while the probability of having j ∈ Isel (condition i ∈ I \\ Isel) is given by Nsel/(N − 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Hence, we have E � 1i∈I\\Isel 1j∈Isel � = P [i ∈ I \\ Isel, j ∈ Isel] = (N − Nsel)Nsel N(N − 1) from which follows E � W 2 2 (µN, µNsel) � ≤ 1 N · Nsel � i,j∈I ∥zi − zj∥2 2 E � 1i∈I\\Isel 1j∈Isel � = 1 N · Nsel (N − Nsel)Nsel N(N − 1) � i,j∈I ∥zi − zj∥2 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The desired estimates can finally be obtained by noting that the variance can be computed as var(z) = 1/(2N2) � i,j∈I ∥zi − zj∥2 2, see definition (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 32 Finally, we are ready to provide a proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let {(xk i , yk i )}i∈Ik, |Ik| = Nk be the sequence of particles generated by iteration (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) where additionally Nk+1 − Nk particles are discarded after each step k ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We denote with µk Nk ∈ P(R2d) the empirical measure associated with such particle system given by µk Nk = 1 Nk � i∈Ik δ(xk i ,yk i ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We also introduce the measures µk N0, k ≥ 0 corresponding to a particle system generated with the same initial conditions µ0 N0 but where no particle reduction occurs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consistently, we define µh Nk, h > k to represent the particle system generated starting from µk Nk, after h − k iterations, with no random selection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The relation between such measures is summarized in the following diagram µ0 N0 → µ1 N0 → µ2 N0 → .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' → µk N0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' ��� µ1 N1 → µ2 N1 → .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' → µk N1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' ��� µ2 N2 → .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' → µk N2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' µk Nk .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8) where → indicates an iteration step (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3) while ��� a particle reduction procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Therefore, we are interested in studying the distance between the main diagonal of such diagram µk Nk, cor- responding to the system with particle reduction, and the first line µk N0 where particle reduction is never performed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' We note that the 2-Wasserstein distance between subsequent rows can be estimated thanks to Proposition A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 and Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Let ˜zh+1 denote the set of particles associated with the probability measure µh+1 Nh , that is, the particle systems before the selection procedure (up- per diagonal elements in scheme (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='8)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' By first applying Proposition A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='3 and, subsequently, Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2 to ˜zh+1, we obtain that for some constant C > 0 E � W 2 2 (µk Nk, µk N0) � ≤ C k−1 � h=0 E � W 2 2 � µk Nh, µk Nℓ+1 �� ≤ C k−1 � h=0 Ck−h+1 1 E � W 2 2 � µh+1 Nh , µh+1 Nh+1 �� ≤ 2 C k−1 � h=0 Ck−h+1 1 var � ˜zh+1� Nh − Nh+1 Nh − 1 33 ≤ C2 max h=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=',k var � ˜zh� 1 Nk − 1 k−1 � h=0 Nh − Nh+1 = C2 max h=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=',k var � ˜zh� N0 − Nk Nk − 1 with C2 = C2(∆t, λ, σ, ν, β, α, k, M).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Finally, the desired estimate follows after noting that W 2 2 (ρk Nk, ρk N0) ≤ W 2 2 (µk Nk, µk N0) since ∥xk i − xk j ∥2 2 ≤ ∥(xk i , yk i ) − (xk j , yk j )∥2 2 for all couples of particles (i, j).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Acknowledgments This work has been written within the activities of GNCS group of INdAM (National Institute of High Mathematics).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' acknowledges the partial support of MIUR-PRIN Project 2017, No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 2017KKJP4X “Innovative numerical methods for evolutionary partial differential equations and applications”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' The work of G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through 320021702/GRK2326 “Energy, Entropy, and Dissipative Dynamics (EDDy)” and SFB 1481 “Sparsity and Singular Structures”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' acknowledges the support of the ESF PhD Grant “Mathematical and statistical methods for machine learning in biomedical and socio-sanitary applications”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' References [1] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Arnold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Stochastic Differential Equations: Theory and Applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Wiley Interscience, first edition, 1974.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Structure-preserving algorithms for ordinary differential equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [2] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Arora, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Acharya, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Verma, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Panigrahi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Multilevel thresholding for image segmentation through a fast statistical recursive algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pattern Recognition Letters, 29(2):119–125, 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [3] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Carrillo, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Choi, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Totzeck, and O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tse.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' An analytical framework for consensus- based global optimization method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Models Methods Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=', 28(6):1037–1066, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [4] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Carrillo, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jin, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Li, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Zhu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A consensus-based global optimization method for high dimensional machine learning problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' ESAIM: COCV, 27:S5, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [5] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jin, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Lyu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A consensus-based global optimization method with adaptive momentum estimation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Communications in Computational Physics, 31(4):1296–1316, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [6] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Flamary, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Courty, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gramfort, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Alaya, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Boisbunon, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chambon, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chapel, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Corenflos, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fatras, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fournier, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gautheron, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gayraud, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Janati, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Rakotoma- monjy, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Redko, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Rolet, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Schutz, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Seguy, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Sutherland, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tavenard, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tong, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Vayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' POT: Python Optimal Transport.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Journal of Machine Learning Research, 22(78):1–8, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 34 [7] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fornasier, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Huang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pareschi, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' S¨unnen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consensus-based optimization on the sphere: Convergence to global minimizers and machine learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Machine Learning Research, 22(237):1–55, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [8] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fornasier, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Huang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pareschi, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' S¨unnen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consensus-based optimization on hypersurfaces: Well-posedness and mean-field limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Mathematical Models and Methods in Applied Sciences, 30(14):2725–2751, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [9] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fornasier, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Klock, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Riedl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Consensus-based optimization methods converge globally.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' arXiv:2103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='15130, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [10] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fornasier, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Klock, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Riedl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Convergence of anisotropic consensus-based opti- mization in mean-field law.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jim´enez Laredo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Hidalgo, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Babaagba, editors, Applications of Evolutionary Computation, pages 738–754, Cham, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Springer International Publishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [11] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gandomi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Yun, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Yang, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Talatahari.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chaos-enhanced accelerated particle swarm optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Communications in Nonlinear Science and Numerical Simu- lation, 18(2):327–340, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [12] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Garrigos, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Rosasco, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Villa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Convergence of the forward-backward algorithm: beyond the worst-case with the help of geometry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Mathematical Programming, June 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [13] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gilbarg and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Trudinger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Elliptic Partial Differential Equations of Second Order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Classics in Mathematics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Springer Berlin Heidelberg, 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [14] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Grassi and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pareschi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' From particle swarm optimization to consensus based opti- mization: Stochastic modeling and mean-field limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Mathematical Models and Methods in Applied Sciences, 31(08):1625–1657, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [15] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Ha, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jin, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Convergence of a first-order consensus-based global optimiza- tion algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Mathematical Models and Methods in Applied Sciences, 30(12):2417–2444, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [16] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Ha, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jin, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Convergence and error estimates for time-discrete consensus- based optimization algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Numerische Mathematik, 147(2):255–282, Feb 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [17] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Holland.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Adaptation in natural and artificial systems: an introductory analysis with applications to biology, control, and artificial intelligence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' MIT press, 1992.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [18] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Houssein, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gad, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Hussain, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Suganthan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Major advances in particle swarm optimization: Theory, analysis, and application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Swarm and Evolutionary Compu- tation, 63:100868, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [19] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Huang and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Qiu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' On the mean-field limit for the consensus-based optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Math- ematical Methods in the Applied Sciences, 45(12):7814–7831, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [20] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Huang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Qiu, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Riedl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' On the global convergence of particle swarm optimization methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' arXiv:2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12460, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 35 [21] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Hussain, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Mohd Salleh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Cheng, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Shi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Metaheuristic research: a compre- hensive survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Artificial Intelligence Review, 52(4):2191–2233, Dec 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [22] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jamil and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Yang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A literature survey of benchmark functions for global optimization problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Int.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Journal of Mathematical Modelling and Numerical Optimisation, 2(4):150– 194, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [23] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jin, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Li, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Random Batch Methods (RBM) for interacting particle systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Journal of Computational Physics, 400:108877, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [24] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kennedy and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Eberhart.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Particle swarm optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In Proceedings of ICNN’95 - International Conference on Neural Networks, volume 4, pages 1942–1948 vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='4, 1995.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [25] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kirkpatrick, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Gelatt Jr, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Vecchi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Optimization by simulated annealing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' science, 220(4598):671–680, 1983.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [26] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' LeCun and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Cortes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' MNIST handwritten digit database.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [27] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Zhou, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kang, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Wang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A fast particle swarm optimization algorithm with cauchy mutation and natural selection strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Zeng, editors, Advances in Computation and Intelligence, pages 334–343, Berlin, Heidelberg, 2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Springer Berlin Heidelberg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [28] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Dong, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Ruan, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Shi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A parallel integrated learning technique of improved particle swarm optimization and bp neural network and its application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Scientific Reports, 12(1):19325, Nov 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [29] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liang and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Wang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Applying genetic algorithm and ant colony optimization algorithm into marine investigation path planning model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Soft Computing, 24(11):8199–8210, Jun 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [30] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Jin, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tang, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Huang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Improved particle swarm optimiza- tion combined with chaos.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chaos, Solitons & Fractals, 25(5):1261–1271, 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [31] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Nickabadi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Ebadzadeh, and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Safabakhsh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A novel particle swarm optimization algorithm with adaptive inertia weight.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Applied Soft Computing, 11(4):3658–3670, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [32] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Otsu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A threshold selection method from gray-level histograms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' IEEE transactions on systems, man, and cybernetics, 9(1):62–66, 1979.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [33] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pedersen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Good parameters for particle swarm optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Hvass Lab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=', Copen- hagen, Denmark, Tech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Rep.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' HL1001, pages 1551–3203, 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [34] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pinnau, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Totzeck, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tse, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Martin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A consensus-based model for global opti- mization and its mean-field limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Models Methods Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=', 27(1):183–204, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [35] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Platen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' An introduction to numerical methods for stochastic differential equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Acta Numerica, 8:197–246, 1999.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 36 [36] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Poli, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kennedy, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Blackwell.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Particle swarm optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Swarm intelligence, 1(1):33–57, 2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [37] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Riedl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Leveraging memory effects and gradient information in consensus-based optimiza- tion: On global convergence in mean-field law.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='2211.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='12184, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [38] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Santambrogio.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Optimal Transport for Applied Mathematicians.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Birkh¨auser, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [39] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Shi and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Eberhart.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A modified particle swarm optimizer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In 1998 IEEE international conference on evolutionary computation proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' IEEE world congress on computational intelligence (Cat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 98TH8360), pages 69–73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' IEEE, 1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [40] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Storn and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Price.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Journal of global optimization, 11(4):341–359, 1997.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [41] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Sznitman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Topics in propagation of chaos.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Hennequin, editor, Ecole d’Et´e de Probabilit´es de Saint-Flour XIX — 1989, pages 165–251, Berlin, Heidelberg, 1991.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Springer Berlin Heidelberg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [42] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Man, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kwong, and Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' He.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Genetic algorithms and their applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' IEEE signal processing magazine, 13(6):22–37, 1996.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [43] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tian and Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Shi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' MPSO: Modified particle swarm optimization and its applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Swarm and evolutionary computation, 41:49–68, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [44] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tsai, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu, and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A novel histogram-based multi-threshold searching algorithm for multilevel colour thresholding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' International Journal of Advanced Robotic Systems, 9(5):223, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [45] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tuli, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Soneji, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Churi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Pixadapt: A novel approach to adaptive image encryp- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chaos, Solitons & Fractals, 164:112628, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [46] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' ˇCrepinˇsek, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Mernik.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Exploration and exploitation in evolutionary algorithms: A survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' ACM Comput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Surv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=', 45(3), jul 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [47] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Wang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Tan, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Liu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Particle swarm optimization algorithm: an overview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Soft computing, 22(2):387–408, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [48] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Yang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Nature-inspired optimization algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Academic Press, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [49] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Yang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Deb, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fong.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Accelerated particle swarm optimization and support vector machine for business optimization and applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' In S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Fong, editor, Networked Digital Technologies, pages 53–66, Berlin, Heidelberg, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Springer Berlin Heidelberg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' [50] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Zhang and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Kong.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' A particle swarm optimization algorithm with empirical balance strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' Chaos, Solitons & Fractals: X, 10:100089, 2023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
+page_content=' 37' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dFQT4oBgHgl3EQfFzW1/content/2301.13242v1.pdf'}
diff --git a/4tE4T4oBgHgl3EQfBAt_/content/2301.04848v1.pdf b/4tE4T4oBgHgl3EQfBAt_/content/2301.04848v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..29b8e83897e296145e1d6870d91f61447254603f
--- /dev/null
+++ b/4tE4T4oBgHgl3EQfBAt_/content/2301.04848v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d90e0d0ec6ffc8a588bfede14d7074a701576ab0cdd32bf4ce3feca5f0c5600b
+size 327460
diff --git a/4tE4T4oBgHgl3EQfBAt_/vector_store/index.faiss b/4tE4T4oBgHgl3EQfBAt_/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..de166c39899144b48aa84fc190257e742b7332d8
--- /dev/null
+++ b/4tE4T4oBgHgl3EQfBAt_/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8e86af0e363f1a7266bf91176b26d442006102ee793b170208fb7c84f6225f2
+size 3342381
diff --git a/4tE4T4oBgHgl3EQfBAt_/vector_store/index.pkl b/4tE4T4oBgHgl3EQfBAt_/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..808c22ca1afcfb6028ff91bc6eb6c0dc7c8b10ce
--- /dev/null
+++ b/4tE4T4oBgHgl3EQfBAt_/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:525a1815d1c206677dfd4644f0a3bad1777b03c95dca8e62b4cf2b8e65a2e38b
+size 131450
diff --git a/5NE5T4oBgHgl3EQfPA41/vector_store/index.faiss b/5NE5T4oBgHgl3EQfPA41/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..7641301318d65be5b1b9780b6e389cb0cad470e6
--- /dev/null
+++ b/5NE5T4oBgHgl3EQfPA41/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbeb7d2ee01f964382c16982bc08a5d830bf00637fb8a28757144824e30519df
+size 5898285
diff --git a/5dE3T4oBgHgl3EQfpQo9/content/2301.04640v1.pdf b/5dE3T4oBgHgl3EQfpQo9/content/2301.04640v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..e692000851edd5f296dcad3430e40f955b6f1983
--- /dev/null
+++ b/5dE3T4oBgHgl3EQfpQo9/content/2301.04640v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a60dc765ccadd082daa5717be29f578f49c20e3a199df6ccfc1868ad30920da9
+size 276756
diff --git a/5dE3T4oBgHgl3EQfpQo9/vector_store/index.faiss b/5dE3T4oBgHgl3EQfpQo9/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..2fa0f8037fdc1e6d2830b484da7ca056fd84f74d
--- /dev/null
+++ b/5dE3T4oBgHgl3EQfpQo9/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57f364e4eef8a296aa89d839d540bedd6719d8ca9b500fcfa947ad8d29cad451
+size 1900589
diff --git a/5dE3T4oBgHgl3EQfpQo9/vector_store/index.pkl b/5dE3T4oBgHgl3EQfpQo9/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..44229b6974bdd24ca07ae9405b2927332eb45a91
--- /dev/null
+++ b/5dE3T4oBgHgl3EQfpQo9/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:284b67b07932f827ad2d805a37cab801366845866ec80e897df3974cf0348689
+size 65747
diff --git a/5tE3T4oBgHgl3EQfQwkt/content/tmp_files/2301.04415v1.pdf.txt b/5tE3T4oBgHgl3EQfQwkt/content/tmp_files/2301.04415v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3754232ef49f2b29adfc30dde29a683d60cca5bc
--- /dev/null
+++ b/5tE3T4oBgHgl3EQfQwkt/content/tmp_files/2301.04415v1.pdf.txt
@@ -0,0 +1,2048 @@
+arXiv:2301.04415v1 [hep-th] 11 Jan 2023
+Charges for Hypertranslations and Hyperrotations
+Chethan Krishnana and Jude Pereirab
+aCentre for High Energy Physics, Indian Institute of Science,
+C.V. Raman Road, Bangalore 560012, India.
+Email: chethan.krishnan.physics@gmail.com
+b Department of Physics, Arizona State University,
+Tempe, Arizona 85287-1504, USA.
+Email:
+jude.pereira@asu.edu
+Hypertranslations and hyperrotations are asymptotic symmetries of flat space, on top of the
+familiar supertranslations and superrotations. They were discovered in arXiv:2205.01422 by working
+in the Special Double Null (SDN) gauge, where I + and I − are approached along null directions. It
+was observed there that while the hair degrees of freedom associated to these diffeomorphisms show
+up in the covariant surfaces charges, the diffeomorphisms themselves do not. This made their status
+intermediate in some ways between global symmetries and trivial gauge transformations, making
+interpretation ambiguous.
+In this paper, we revisit the fall-offs considered in arXiv:2205.01422
+which were strictly subleading to Minkowski in conventional double null coordinates. We identify
+a new class of fall-offs where this assumption is relaxed, but whose charges nonetheless remain
+finite. Remarkably, the leading behavior is still Riemann flat, indicating that these are soft modes.
+With this more refined definition of asymptotic flatness, we show that leading hypertranslations
+and leading hyperrotations explicitly show up in the charges.
+This makes them genuine global
+symmetries of asymptotically flat Einstein gravity in the SDN gauge.
+We write down the new
+algebra of asymptotic Killing vectors that subsumes the BMS algebra.
+Introduction:
+When the cosmological constant Λ is
+negative, it is widely believed that the radial direction of
+the resulting anti-de Sitter (AdS) spacetime is holograph-
+ically emergent [1]. When Λ is positive, less is known,
+but there are many suggestions in the literature that the
+timelike direction of de Sitter (dS) space may have a holo-
+graphic origin [2]. These observations make one suspect
+that it may be useful to view the holographic direction
+for flat space, which has Λ = 0, as a null coordinate.
+Historically however, the null boundary of flat space is
+typically approached along a spacelike direction, eg., in
+the famous Bondi gauge [3].
+With quite different motivations, various aspects of flat
+space were explored from a holographic perspective in [4–
+6]. Along the way, it was realized that a natural gauge
+for asymptotically flat space is the Special Double Null
+(SDN) gauge [7], defined by
+guu = 0, gvv = 0, guA = gvA.
+(1)
+Here u and v are null coordinates and I + and I − are
+covered (generically) by two separate patches around v →
+∞ and u → −∞ respectively. The holographic directions
+are v and −u in these patches.
+The notion of a double null coordinate system has been
+explored in various contexts in the literature before, see
+eg.
+[8, 9].
+But usually in these settings, not enough
+constraints are imposed to fix all the coordinate free-
+dom; they are therefore not genuine gauge choices. In
+fact in the context of mathematical relativity, the form
+of the double null metric that is sometimes written down
+(see eg., eqn (70) of [10]) does not fall into the gauge we
+have presented above. This reflects a difference in phi-
+losophy. General relativists are interested in I + as the
+eventual location of gravitational waves from localized
+objects. But if one is interested in graviton scattering,
+as perhaps necessary in quantum gravity, we need access
+to both I + and I −. Our gauge has a natural u ↔ −v
+symmetry which relates I + and I −. This manifests an
+asymptotic CPT invariance [7], which is believed to be a
+symmetry of quantum gravity in flat space [11].
+In [12] we considered the most general asymptotic sym-
+metry algebra in SDN gauge with fall-offs which are
+power laws in the respective null coordinate. Minkowski
+space in double null coordinates can be obtained by writ-
+ing u = t − r and v = t + r:
+ds2 = −du dv + 2
+�v − u
+2
+�2
+γz¯zdzd¯z
+(2)
+which has guA
+=
+gvA
+=
+0.
+This suggested that
+one should allow fall-offs where guA and gvA are at
+most O(v−1) at I +.
+The result was found to con-
+tain new classes of non-trivial asymptotic diffeomor-
+
+2
+phisms on top of the BMS symmetries [13]. These were
+named hypertranslations, subleading hypertranslations
+and subleading hyperrotations1. The algebra of asymp-
+totic Killing vectors that extends the BMS algebra was
+identified and the covariant surface charges [14, 15] were
+computed.
+It was noted that these charges had non-
+trivial dependence on the corresponding “hair” (the met-
+ric parameters affected by these asymptotic diffeomor-
+phisms). But at the same time, they did not contain the
+new asymptotic diffeomorphisms themselves, and there-
+fore the interpretation of these charges was ambiguous.
+Typically for global symmetries that emerge from an
+asymptotic symmetry calculation, both the diffeomor-
+phisms and the associated hair parameters appear in the
+charge expression. On the other hand, for trivial diffeo-
+morphisms, neither the diffeomorphisms nor the parame-
+ters associated to them appear in the charges. This made
+the status of hypertranslations and hyperrotations in-
+termediate between global symmetries and trivial gauge
+transformations, making them challenging to interpret.
+Part of the problem here is that because we are working
+with null directions, the formalism that is most suited
+for our purposes is the covariant phase space approach of
+Wald and followers [14, 15], while a more Hamiltonian-
+like formalism is perhaps more suited for interpretational
+purposes.
+In this paper, we will bypass this problem by identify-
+ing a new set of fall-offs which are not strictly subleading
+to (2), but for which the charges are still finite. These
+fall-offs are presented in (3) and also in more detail in the
+Supplementary Material. In particular, our fall-offs will
+allow guA = O(v0) = gvA. A key feature of these fall-offs
+is that they can change the metric at an order more lead-
+ing than (2), and yet remarkably, we are able to show that
+their charges remain finite. In particular, a striking fact
+that we note is that demanding Riemann flatness allows
+these terms. This allows us to adopt the philosophy that
+there is nothing too sacred about the specific form in ex-
+pression (2), it is the demand of Riemann flatness that
+should be respected in deciding the leading behavior. We
+will find that Riemann flatness still leaves the possibility
+that these modes can be functions of the angular coordi-
+nates (z, ¯z). We will eventually identify these as related
+to the hyperrotation hair. This should be compared to
+the familiar fact that purely angle dependent shear modes
+1In [12], the latter were simply called hyperrotations. But in
+the present paper we will find more leading counterparts to these
+AKVs which are more naturally called (leading) hyperrotations.
+Therefore the ones noted in [12] will be referred to as subleading
+hyperrotations in this paper.
+in Bondi gauge are soft hair associated to supertrans-
+lations, and turning them on can still leave the metric
+Riemann flat [16]. Similarly in SDN gauge, turning on
+supertranslation hair or hypertranslation hair, leaves the
+metric Riemann flat. But both in Bondi gauge as well as
+in SDN gauge, the supertranslation and hypertranslation
+soft modes were subleading to the corresponding conven-
+tional form of the Minkowski metric. The new feature
+of hyperrotation hair here is that it is more leading than
+(2) while remaining Riemann flat. Riemann flatness in
+SDN gauge has many remarkable properties, which will
+be discussed in detail elsewhere [17].
+Once we adopt these relaxed fall-offs the nature of the
+calculation is parallel to that in [12], even though techni-
+cally more involved due to the increased number of metric
+functions that we start with. The result of this exercise is
+that we find that (a) the charges are still finite, (b) there
+is a new set of Diff(S2) transformations (the leading hy-
+perrotations) that appear before subleading hyperrota-
+tions but are subleading to superrotations, (c) both the
+hair parameters as well as the diffeomorphisms associ-
+ated to the leading hypertranslations and leading hyper-
+rotations appear in the charge expressions, on top of the
+conventional BMS quantities, (d) demanding Riemann
+flatness still allows soft hair associated to these diffeo-
+morphisms to appear in the metric, and (e) the algebra
+of the asymptotic symmetries is enhanced with respect
+to both the BMS algebra as well as the BBMS algebra of
+[12].
+The next section contains the main results of this pa-
+per. To avoid repetition, we will only emphasize aspects
+of the discussion that are distinct from those in [12]. In
+particular, we will simply present the final algebra with-
+out presenting the details of the derivation – the approach
+is identical to that in [12], even though technically more
+involved.
+Results: We will work with SDN gauge discussed in [7].
+The fall-offs are presented in great detail in the Supple-
+mentary Material in terms of functions appearing in the
+metric. Here we will write the fall-offs as
+guv = −2 + O
+�
+v−1�
+(3a)
+gAB = 4γAB v−2 + O
+�
+v−3�
+(3b)
+guA = gvA = O
+�
+v−2�
+(3c)
+Even though technically this is a small change from our
+previous paper, we emphasize that this is a pretty sub-
+stantive departure from experience in other gauges. We
+are demanding that the metric be distinct from the con-
+ventional form Minkowski metric (2), already at leading
+order.
+There are three reasons why we believe this is
+
+3
+reasonable. Firstly, the charges remain finite. Secondly,
+demanding Riemann flatness does not force these terms
+to be zero. Thirdly, with this choice, we get a perfectly
+conventional structure for the leading hypertranslation
+and hyperrotation charges.
+The asymptotic Killing vector conditions take the
+form:
+Lξguv = O
+�
+v−1�
+(4a)
+LξguA = O
+�
+v−2�
+(4b)
+LξgvA = O
+�
+v−2�
+(4c)
+LξgAB = O
+�
+v−3�
+(4d)
+These and the exact Killing conditions (21), lead to the
+solutions:
+ξu = f +
+ξu
+(1)
+v
++
+ξu
+(2)
+v2 +
+ξu
+(3)
+v3 + O
+�
+v−4�
+(5a)
+ξv = −ψ
+2 v + ξv
+(0) +
+ξv
+(1)
+v
++
+ξv
+(2)
+v2 + O
+�
+v−3�
+(5b)
+ξA = Y A +
+ξA
+(1)
+v
++
+ξA
+(2)
+v2 +
+ξA
+(3)
+v3 + O
+�
+v−4�
+(5c)
+where
+f = ξu
+(0) = ψ(z, ¯z) u/2 + T (z, ¯z),
+with ψ(z, ¯z) = DAY A
+(6a)
+ξu
+(1) = αA
+2 ∂Af
+(6b)
+ξu
+(2) = 1
+2
+�
+αA
+3 ∂Af + αA
+2 ∂Aξu
+(1)
+�
+(6c)
+ξu
+(3) = 1
+3
+�
+αA
+4 ∂Af + αA
+3 ∂Aξu
+(1) + αA
+2 ∂Aξu
+(2)
+�
+(6d)
+Here T (z, ¯z) denotes supertranslations, and Y z(z), Y ¯z(¯z)
+denote superrotations.
+On top of the BMS diffeo-
+morphisms, the ξv
+(0), ξv
+(1), ξA
+(1) and ξA
+(2) are also de-
+termined by the exact and asymptotic Killing condi-
+tions. The independent functions contained in them are
+hypertranslations φ(z, ¯z), sub-leading hypertranslations
+τ(z, ¯z), hyperrotations XA(z, ¯z) and sub-leading hyper-
+rotations ZA(z, ¯z) respectively. They are related to the
+ξv
+(0), ξv
+(1), ξA
+(1), ξA
+(2) via:
+ξA
+(1) = XA − 2 DAf
+(7a)
+ξv
+(0) = φ + T + △γT − 1
+4aA
+2 DAψ − 1
+2DAXA
+(7b)
+ξv
+(1) = ˜τ + 1
+2A A
+2 DAψ
+(7c)
+ξA
+(2) = ˜ZA + C AB DBψ + A A
+2 ψ − u XA + 2 u DAξv
+(0)
+− u2 DAψ − L1 DAψ
+We have introduced ˜τ and ˜ZA for convenience which are
+related to the sub-leading hypertranslations τ and sub-
+leading hyperrotations ZA via
+˜τ = τ − 1
+4 aA
+3 DAψ
++
+�
+D¯zcz¯z − Dzczz + γz¯zDzD¯zaz
+2 − γz¯zD2
+¯za¯z
+2
+�
+DzT
++
+�
+Dzcz¯z − D¯zc¯z¯z + γz¯zD¯zDza¯z
+2 − γz¯zD2
+zaz
+2
+�
+D¯zT
++ aA
+2 DAξv
+(0) + aA
+2 DAT + 1
+4aA
+2 DA
+�
+aB
+2 DBψ
+�
+(8)
+˜Zz = Zz + czz DzT + cz¯z D¯zT + T Dzczz − T D¯zcz¯z
++ az
+2 ξv
+(0) − 1
+2XBDBaz
+2 + 1
+2aB
+2 DBXz − γz¯zD¯za¯z
+2D¯zT
+− γz¯zD¯zaz
+2DzT + 1
+4az
+2a¯z
+2D¯zψ − γz¯za¯z
+2D2
+¯zT + 1
+4
+�
+az
+2
+�2Dzψ
+− 1
+2az
+2∆γT − γz¯zT DzD¯zaz
+2 + γz¯zT D2
+¯za¯z
+2
+(9)
+˜Z ¯z = Z ¯z + c¯z¯z D¯zT + cz¯z DzT + T D¯zc¯z¯z − T Dzcz¯z
++ a¯z
+2 ξv
+(0) − 1
+2XBDBa¯z
+2 + 1
+2aB
+2 DBX ¯z − γz¯zDzaz
+2DzT
+− γz¯zDza¯z
+2D¯zT + 1
+4a¯z
+2az
+2Dzψ − γz¯zaz
+2D2
+zT + 1
+4
+�
+a¯z
+2
+�2D¯zψ
+− 1
+2a¯z
+2∆γT − γz¯zT D¯zDza¯z
+2 + γz¯zT D2
+zaz
+2
+(10)
+These expressions are significantly more complicated
+than those in [12], so let us pause to explain some of the
+details.
+The integration “constants” in the shear are2
+introduced via
+CAB(u, z, ¯z) = cAB(z, ¯z) +
+� u
+−∞
+du′NAB(u′, z, ¯z)(11)
+with NAB ≡ ∂uCAB, being the SDN news tensor. Sim-
+ilarly, we have defined the integration “constant” in αA
+2
+as
+αA
+2 (u, z, ¯z) = aA
+2 (z, ¯z) +
+� u
+−∞
+du′βA
+2 (u′, z, ¯z)
+(12)
+where βA
+2 ≡ ∂uαA
+2 . See [7, 18] for a discussion on integrals
+of this type that are defined from I +
+− to u. On shell (ie.,
+when Einstein equations hold), we have Nz¯z = 0 and
+βA
+2 = 0, so we will have
+Cz¯z(u, z, ¯z) = cz¯z(z, ¯z)
+αA
+2 (u, z, ¯z) = aA
+2 (z, ¯z)
+(13)
+In addition to this, the Einstein constraints also require
+2The notation here is slightly different from that in [12].
+
+4
+that λ1 = 0. For ξA
+(2), combining all the relevant equa-
+tions, we can write [18]
+∂uξA
+(2) = CAB DBψ − 2 u DAψ + 2 DAξv
+(0) + αA
+2 ψ
+− λ1DAψ − XA
+=⇒ ξA
+(2) = C AB DBψ − u2 DAψ + 2 u DAξv
+(0) + A A
+2 ψ
+− L1DAψ − u XA + ˜ZA(z, ¯z)
+(14)
+The u-independence of ψ, ξv
+(0) and XA has been used in
+writing the integrated version in the second step. Also
+C AB, A A
+2
+and L2 have been defined via
+∂uC AB = CAB(u, z, ¯z)
+(15)
+∂uA A
+2 = αA
+2 (u, z, ¯z)
+(16)
+∂uL1 = λ1(u, z, ¯z)
+(17)
+As in [12], ˜ZA(z, ¯z) is taken as the u-independent piece in
+ξA
+(2). The shift is done on ˜ZA via (9)-(10) and the result
+is what we call sub-leading hyperrotations ZA.
+The rest of the notation follows that of [12]. As empha-
+sized there, the idea in (7) is to do certain shifts so that
+the structure of the diffeomorphisms is cleanest.
+This
+“diagonalizes” the algebra of diffeomorphisms. The phi-
+losophy here is identical, even though the expressions are
+more complicated.
+The hair associated to the various diffeomorphisms are
+therefore as follows: supertranslations T (z, ¯z) are associ-
+ated to the u-independent shifts in Czz and C¯z¯z, hyper-
+translations φ(z, ¯z) are associated to the u-independent
+shifts of Cz¯z, subleading hypertranslations τ(z, ¯z) are as-
+sociated to u-independent shifts of λ2, hyperrotations
+XA(z, ¯z) are associated to u-independent shifts of α A
+2
+and sub-leading hyperrotations ZA(z, ¯z) are associated
+to u-independent shifts of α A
+3 . As in Bondi gauge, we
+also have superrotations Y z(z), Y ¯z(¯z).
+The shifts in-
+volved in the definitions of φ, τ, XA and ZA are detailed
+in the Supplementary Material. As in [12], supertrans-
+lations and leading-&-subleading hypertranslations are
+diffeomorphisms of u and v respectively. Leading hyper-
+rotations were not present in [12], but both leading and
+subleading hyperrotations are subleading to superrota-
+tions on the sphere.
+We will define the “Beyond BBMS” algebra b2-bms4
+as the asymptotic symmetry algebra of the nine non-
+trivial diffeomorphisms – supertranslations, superrota-
+tions, hypertranslations & subleading hypertranslations,
+and hyperrotations & subleading hyperrotations. Follow-
+ing [12, 13], we define the bracket
+��Y , �T, �φ, �τ, �
+X, �Z
+�
+=
+�
+(Y1, T1, φ1, τ1, X1, Z1), (Y2, T2, τ2, φ2, X2, Z2)
+�
+(18)
+The notation is the natural generalization of that in [12,
+13] and the reader should consult those papers for the
+detailed definitions. The new algebra is defined via �Y ,
+�T, �φ, �τ, �
+X and �Z given by the following expressions:
+�Y A = Y B
+1 ∂BY A
+2 − Y B
+2 ∂BY A
+1
+(19a)
+�T = Y A
+1 ∂AT2 − Y A
+2 ∂AT1 + 1
+2 (T1 ψ2 − T2 ψ1).
+(19b)
+�φ = 1
+2(ψ1φ2 − ψ2φ1) +
+�
+Y A
+1 ∂Aφ2 − Y A
+2 ∂Aφ1
+�
+(19c)
+�τ = (ψ1τ2 − ψ2τ1) +
+�
+Y A
+1 ∂Aτ2 − Y A
+2 ∂Aτ1
+�
+(19d)
+�
+XA = 1
+2
+�
+ψ1XA
+2 − ψ2XA
+1
+�
++
+�
+Y B
+1 ∂BXA
+2 − Y B
+2 ∂BXA
+1
+�
++
+�
+XB
+1 ∂BY A
+2 − XB
+2 ∂BY A
+1
+�
+(19e)
+�ZA =
+�
+ψ1ZA
+2 − ψ2ZA
+1
+�
++
+�
+Y B
+1 ∂BZA
+2 − Y B
+2 ∂BZA
+1
+�
++
+�
+ZB
+1 ∂BY A
+2 − ZB
+2 ∂BY A
+1
+�
+(19f)
+This is what we call the b2-bms4 algebra. The fact that
+these nine non-trivial diffeomorphisms form a closed al-
+gebra is checked by the same procedure as outlined in
+[12]. The calculations are straightforward but lengthier
+variations of those there. In order to identify the capped
+quantities, we need to consider the Barnich-Troessaert
+bracket [ξ1, ξ2]M of two AKVs ξ1 and ξ2 [12, 13]. The
+structure is parallel to that presented in [12], with a no-
+
+5
+table difference in the A-component which takes the form
+[ξ1, ξ2]A
+M = �Y A +
+�ξA
+(1)
+v
++
+�ξA
+(2)
+v2 + O
+�
+v−3�
+.
+(20)
+In computing all four components of the Barnich-
+Troessaert bracket, we need �ξv
+(0), �ξv
+(1), �ξA
+(1) and �ξA
+(2), which
+are defined as in (7) but with Y A, T, φ, τ, XA, ZA re-
+placed by their capped versions, defined in (19).
+Equations (19) define the b2-bms4 algebra. Setting the
+hyperrotations XA to zero results in the BBMS algebra
+of [12], and setting φ, τ and ZA as well to zero results in
+the familiar BMS algebra [13].
+Discussion: In this paper, we observed that demand-
+ing finite covariant surface charges in Einstein gravity
+allows fall-offs that are not necessarily subleading to (2).
+Turning on the soft modes associated to supertransla-
+tions and leading hypertranslations/hyperrotations takes
+us beyond (2) even though the metric is still Riemann
+flat. We exploited this fact to work with fall-offs that
+allowed these modes, to show that the covariant surface
+charges contain these diffeomorphisms as well as the as-
+sociated soft hair. This places them on an equal footing
+with conventional global symmetries (eg. supertransla-
+tions), resolving some of the ambiguities pointed out in
+[12].
+Of course, these results open up further questions. Our
+work strongly suggests that the charges associated to hy-
+pertranslations should be interpreted as soft, so it would
+be interesting to connect these results to soft theorems
+(perhaps to the subsubleading soft graviton theorem of
+[20]?) and also to new memory effects. Some of these
+questions are currently under investigation. Hypertrans-
+lations have many similarities to supertranslations, but
+there are also crucial distinctions. The lowest modes of
+supertranslations are simply the action of Poincare trans-
+lations on the boundary (u, z, ¯z). Hypertranslations on
+the other hand are truly distinct from bulk translations –
+we have already subtracted out the supertranslations in
+our shifted diffeomorphisms, when defining hypertransla-
+tions. It should be clear from (5) that the interpretation
+of hypertranslations is more like a bulk diffeomorphism
+at infinity (note that infinity is along the null direction
+v in SDN gauge). It is more naturally compared to ξr
+than ξu in Bondi gauge.
+A related interesting feature of hypertranslations and
+their associated hair is that they can be spherically sym-
+metric.
+This raises a subtlety in the usual statement
+of Birkhoff’s theorem, which will be discussed in an up-
+coming work.
+Note that while supertranslations allow
+soft hair on Schwarzschild, the only spherically symmet-
+ric supertranslation is a time translation, so this subtlety
+does not arise for Schwarzschild in Bondi gauge. It is also
+important to emphasize that hypertranslations should be
+distinguished from the shifts in v at the past boundary
+I −. The latter are simply supertranslations, but now
+acting in the past. What we mean by hypertranslations
+are shifts in v at I +. There is no obvious connection
+between the two (other than the future-past matching at
+i0 that was discussed in [7]) because these coordinates
+live in different charts.
+What about subleading hypertranslations and sublead-
+ing hyperrotations? They do not show up in the charges
+even with the new fall-offs, but their associated hair was
+present both in [12] as well as here. So their interpre-
+tation remains ambiguous. It is natural to consider the
+sub-algebra obtained by setting the subleading hyper-
+translations/hyperrotations to zero.
+This would mean
+that we are working with supertranslations, superrota-
+tions, leading hypertranslations and leading hyperrota-
+tions. This is a natural generalization of the conventional
+BMS algebra in the SDN gauge; it is clearly of interest to
+study it more closely. One could also consider the even
+simpler generalization of BMS, obtained by adding only
+the leading hypertranslations and suppressing the lead-
+ing hyperrotations. This algebra has the advantage that
+we are not turning on diffeomorphisms on the sphere,
+but only the Virasoro (super)rotations.
+While it may
+be difficult to conclusively argue for such a choice from
+a purely asymptotic symmetry perspective, it is natural
+from a celestial holography perspective [19]. This is the
+algebra of supertranslations, (leading) hypertranslations
+and superrotations.
+ACKNOWLEDGMENTS
+We thank Sudip Ghosh and Sarthak Talukdar for dis-
+cussions.
+[1] E.
+Witten,
+“Anti-de
+Sitter
+space
+and
+hologra-
+phy,”
+Adv. Theor. Math. Phys. 2,
+253-291 (1998)
+doi:10.4310/ATMP.1998.v2.n2.a2 [arXiv:hep-th/9802150
+[hep-th]].
+[2] See eg., V. Balasubramanian, J. de Boer and D. Minic,
+“Notes on de Sitter space and holography,”
+Class.
+Quant. Grav. 19, 5655-5700 (2002) doi:10.1016/S0003-
+4916(02)00020-9 [arXiv:hep-th/0207245 [hep-th]], and
+the first few references therein.
+[3] H. Bondi, M. G. J. van der Burg and A. W. K. Met-
+zner, “Gravitational waves in general relativity. 7. Waves
+from axisymmetric isolated systems,” Proc. Roy. Soc.
+
+6
+Lond. A 269, 21-52 (1962) doi:10.1098/rspa.1962.0161.
+R. K. Sachs,
+“Gravitational waves in general rela-
+tivity. 8. Waves in asymptotically flat space-times,”
+Proc.
+Roy.
+Soc.
+Lond.
+A
+270,
+103-126
+(1962)
+doi:10.1098/rspa.1962.0206; R. Sachs, “Asymptotic sym-
+metries in gravitational theory,” Phys. Rev. 128, 2851-
+2864 (1962) doi:10.1103/PhysRev.128.2851
+[4] B. Bhattacharjee and C. Krishnan, “A General Prescrip-
+tion for Semi-Classical Holography,” [arXiv:1908.04786
+[hep-th]].
+[5] C. Krishnan, “Bulk Locality and Asymptotic Causal
+Diamonds,”
+SciPost
+Phys.
+7,
+no.4,
+057
+(2019)
+doi:10.21468/SciPostPhys.7.4.057
+[arXiv:1902.06709
+[hep-th]].
+[6] C. Krishnan, V. Patil and J. Pereira, “Page Curve
+and
+the
+Information
+Paradox
+in
+Flat
+Space,”
+[arXiv:2005.02993 [hep-th]].
+[7] C. Krishnan and J. Pereira, “A New Gauge for Asymp-
+totically Flat Spacetime,” [arXiv:2112.11440 [hep-th]].
+[8] R. M. Wald, “General Relativity,” Chicago Univ. Pr.,
+1984, doi:10.7208/chicago/9780226870373.001.0001
+[9] P. R. Brady, S. Droz, W. Israel and S. M. Morsink,
+“Covariant
+double
+null
+dynamics:
+(2+2)
+splitting
+of
+the
+Einstein
+equations,”
+Class.
+Quant.
+Grav.
+13, 2211-2230 (1996) doi:10.1088/0264-9381/13/8/015
+[arXiv:gr-qc/9510040 [gr-qc]].
+[10] M. Dafermos, G. Holzegel and I. Rodnianski, “The
+linear stability of the Schwarzschild solution to gravi-
+tational perturbations,” Acta Math. 222, 1-214 (2019)
+doi:10.4310/ACTA.2019.v222.n1.a1
+[arXiv:1601.06467
+[gr-qc]].
+[11] A.
+Strominger,
+“On
+BMS
+Invariance
+of
+Grav-
+itational
+Scattering,”
+JHEP
+07,
+152
+(2014)
+doi:10.1007/JHEP07(2014)152
+[arXiv:1312.2229
+[hep-
+th]].
+[12] C. Krishnan and J. Pereira, “Hypertranslations and Hy-
+perrotations,” [arXiv:2205.01422 [hep-th]].
+[13] G.
+Barnich
+and
+C.
+Troessaert,
+“Aspects
+of
+the
+BMS/CFT
+correspondence,”
+JHEP 05,
+062
+(2010)
+doi:10.1007/JHEP05(2010)062
+[arXiv:1001.1541
+[hep-
+th]].
+[14] V.
+Iyer
+and
+R.
+M.
+Wald,
+“Some
+properties
+of
+Noether charge and a proposal for dynamical black
+hole
+entropy,”
+Phys.
+Rev.
+D
+50,
+846-864
+(1994)
+doi:10.1103/PhysRevD.50.846 [arXiv:gr-qc/9403028 [gr-
+qc]].
+[15] G. Barnich and F. Brandt, “Covariant theory of asymp-
+totic symmetries, conservation laws and central charges,”
+Nucl. Phys. B 633,
+3-82 (2002) doi:10.1016/S0550-
+3213(02)00251-1 [arXiv:hep-th/0111246 [hep-th]].
+[16] A. Strominger, “Lectures on the Infrared Structure of
+Gravity and Gauge Theory,” [arXiv:1703.05448 [hep-th]].
+[17] C. Krishnan and J. Pereira, “Asymptotically Riemann-
+flat Spacetimes,” to appear.
+[18] C. Krishnan and J. Pereira,“A New Gauge for Flat Space
+Holography,” to appear.
+[19] J. H. Schwarz, “Diffeomorphism Symmetry in Two Di-
+mensions and Celestial Holography,” [arXiv:2208.13304
+[hep-th]].
+[20] F. Cachazo and A. Strominger, “Evidence for a New Soft
+Graviton Theorem,” [arXiv:1404.4091 [hep-th]].
+[21] G. Barnich and C. Troessaert, “BMS charge algebra,”
+JHEP 12,
+105 (2011) doi:10.1007/JHEP12(2011)105
+[arXiv:1106.0213 [hep-th]].
+Supplementary material
+REFINED FALL-OFFS
+In this section, we will present the falloffs in some detail. Our emphasis will be on the distinctions from those
+presented in [12]. We start with a quick review of the notation: in d + 1 dimensions, the SDN gauge [7] is defined by
+eqn (1). We will restrict ourselves to 3+1 dimensions here. The exact Killing vector equations are
+Lξguu = 0, Lξgvv = 0, LξguA = LξgvA
+(21)
+and we will write the general metric in this gauge as
+ds2 = −eλdu dv +
+�v − u
+2
+�2
+ΩAB(dxA − αAdu − αAdv)(dxB − αBdu − αBdv)
+(22)
+
+7
+In [12], we presented a set of fall-offs in terms of the functions in this ansatz, which the reader should consult. The
+fall-offs we consider in this paper are distinct in the following functions:
+λ(u, v, z, ¯z) = λ1(u, z, ¯z)
+v
++ λ2(u, z, ¯z)
+v2
++ λ3(u, z, ¯z)
+v3
++ λ4(u, z, ¯z)
+v4
++ O
+�
+v−5�
+(23a)
+αz(u, v, z, ¯z) = αz
+2(u, z, ¯z)
+v2
++ αz
+3(u, z, ¯z)
+v3
++ αz
+4(u, z, ¯z)
+v4
++ αz
+5(u, z, ¯z)
+v5
++ O
+�
+v−6�
+(23b)
+α¯z(u, v, z, ¯z) = α¯z
+2(u, z, ¯z)
+v2
++ α¯z
+3(u, z, ¯z)
+v3
++ α¯z
+4(u, z, ¯z)
+v4
++ α¯z
+5(u, z, ¯z)
+v5
++ O
+�
+v−6�
+(23c)
+In terms of the metric, this results in the fall-offs:
+guu = gvv = O
+�
+v−2�
+(24a)
+guv = −1
+2 + O
+�
+v−1�
+(24b)
+gAB = 1
+4 γAB v2 + O(v)
+(24c)
+guA = gvA = O
+�
+v0�
+(24d)
+Compared to the discussion in [12], we also allow αA
+2 as the O(1/v2) term in the αA fall-off. Just like Cz¯z, αA
+2 also
+turns out to be u-independent once we demand Einstein equations. Hence it is an integration “constant” in Einstein
+constraints in the language of [7, 12, 18].
+Demanding Ricci (or Riemann) flatness forces λ1 to be zero and αA
+2 to be functions only of the angles. We have
+kept them general in the discussions of the AKVs because they can be defined on arbitrary backgrounds, without
+worrying about the equations satisfied by those backgrounds. But one can in principle start a-priori with fall-offs
+(23) where λ1 is set to zero and αA
+2 are functions only of z and ¯z. Some of the expressions we have presented will
+simplify somewhat in that case, but the main results do not change.
+DIFFEOMORPHISM SHIFTS
+As in [12] we will define the various diffeomorphisms after a suitable shift in the fall-off coefficient of ξ. This is more
+elaborate in the present case, and we discuss them in detail below. The philosophy behind these shifts was discussed
+in [12].
+Hyperrotations: The simplest case arises for the leading hyperrotations XA(z, ¯z), so we start with them. From
+the exact Lie derivative conditions, we obtain the following constraint on ξA
+1 ,
+∂uξA
+1 = −DAψ
+(25)
+which on integrating both sides becomes
+ξA
+1 = ˜XA − u DAψ
+(26)
+The metric function corresponding to leading hyperrotations is αA
+2 . Under the action of AKVs, the transformation
+of αA
+2 can be obtained by evaluating δξguA = LξguA at O(v−2) as follows
+δαA
+2 =
+�
+f∂u + LY + ψ
+2
+�
+αA
+2 + ˜XA − u DAψ + 2 DAf
+(27)
+where
+LY αA
+2 = Y B ∂BαA
+2 − αB
+2 ∂BY A.
+(28)
+
+8
+is the Lie derivative of αA
+2 with respect to Y A. Recalling that on-shell αA
+2 = aA
+2 (z, ¯z) and substituting f = ψ(z, ¯z) u/2+
+T (z, ¯z), we obtain
+δaA
+2 =
+�
+LY + ψ
+2
+�
+aA
+2 + ˜XA + 2 DAT
+(29)
+Next we would like to interpret XA(z, ¯z) as the diffeomorphism that causes αA
+2 to be turned on if it was initially zero.
+This immediately suggests the following shift
+˜XA = XA − 2 DAT
+(30)
+Substituting this in (26) and using f =
+�
+ψ/2
+�
+u + T yields
+ξA
+1 = XA − 2 DAf
+(31)
+Hypertranslations: In the case of the leading hypertranslations φ(z, ¯z), the shift is of the form
+ξv
+(0) = φ + T + △γT − 1
+4aA
+2 DAψ − 1
+2DAXA
+(32)
+This reduces to the form presented in [12] when the hyperrotations and their hair are set to zero. The change in Cz¯z
+can be computed by evaluating δξgz¯z = Lξgz¯z at O(v−3). The result is
+δCz¯z =
+�
+f ∂u + LY − 1
+2 ψ
+�
+Cz¯z − 4 ∂z∂¯zf + 2 γz¯z
+�
+ξv
+(0) − f − u
+2 ψ + 1
+2DAXA + 1
+4αA
+2 DAψ
+�
+(33)
+Here LY is the Lie derivative of Cz¯z with respect to Y A defined as in [12]:
+LY Cz¯z = Y A∂ACz¯z +
+�
+∂AY A�
+Cz¯z
+(34)
+On-shell we have Cz¯z = cz¯z(z, ¯z) and αA
+2 = aA
+2 (z, ¯z). Using these and substituting f = ψ(z, ¯z) u/2 + T (z, ¯z), we obtain
+δcz¯z =
+�
+LY − 1
+2 ψ
+�
+cz¯z + 2γz¯z
+�
+ξv
+(0) − T − ∆γT + 1
+2DAXA + 1
+4aA
+2 DAψ
+�
+(35)
+It is clear that ξv
+(0) mixes with supertranslations, superrotations and leading hyperrotations. We wish to remove
+this mixing, so that we can interpret φ(z, ¯z) as the diffeomorphism that causes cz¯z to be turned on if it was initially
+zero. From this it follows that the shift is ξv
+(0) = φ + T + △γT − 1
+4aA
+2 DAψ − 1
+2DAXA, as we presented above. This
+defines hypertranslations, φ(z, ¯z). Note that in deriving the algebra for hypertranslations, we have made use of the
+identity
+δξξv
+(0) = −1
+4
+�
+δaA
+2
+�
+DAψ
+(36)
+where we have demanded that δξφ = 0 and δξXA = 0. This shifted definition above of the hypertranslations ensures
+the vanishing of the hatted �φ on the left hand side of algebra, when φ1 and φ2 are zero. As we pointed out in [12], this
+feature can be viewed as one of the motivations behind doing the shifts. This generalizes to the other diffeomorphisms
+as well.
+Subleading Hyperrotations: Now we turn to the case of subleading hyperrotations ZA(z, ¯z) and the correspond-
+ing metric functions αA
+3 . The same procedure as in [12] now yields
+δαz
+3 =
+�
+f ∂u + LY + ψ
+�
+αz
+3 + 2 ξz
+(2) + 4 u Dzf − 2 CzB DBf − 2 αz
+2 ξv
+(0) + XBDBαz
+2 − αB
+2 DBXz
++ 2 γz¯zD¯zα¯z
+2D¯zf + 2 γz¯zD¯zαz
+2DzT − 1
+2αz
+2α¯z
+2D¯zψ + 2 γz¯zα¯z
+2D2
+¯zT + 2u γz¯zα¯z
+2D2
+¯zψ
+− u γz¯zDzαz
+2D¯zψ − 1
+2
+�
+αz
+2
+�2Dzψ − 2u αz
+2ψ + αz
+2∆γT + 2 λ1Dzf + ∂uαz
+2
+�
+αA
+2 DAf
+�
+(37)
+
+9
+where the Lie derivative is defined as
+LY αA
+3 = Y B ∂BαA
+3 − αB
+3 ∂BY A.
+(38)
+Note that in obtaining the above equation, we have used (27) along with
+δλ1 =
+�
+f ∂u + LY + 1
+2ψ
+�
+λ1 + ∂uαA
+2 DAf
+(39)
+which has been obtained by evaluating δξguv = Lξguv at O(v−1) where LY λ1 = Y A∂Aλ1 is the Lie derivative of λ1
+with respect to Y A. On-shell, we have [18]
+∂uαz
+3 = −2 DzCzz + 2 D¯zcz¯z + 2 γz¯zDzD¯zaz
+2 − 2 γz¯zD2
+¯za¯z
+2
+=⇒ αz
+3(u, z, ¯z) = −2 DzC zz + u
+�
+2 D¯zcz¯z + 2 γz¯zDzD¯zaz
+2 − 2 γz¯zD2
+¯za¯z
+2
+�
++ az
+3(z, ¯z)
+(40)
+and a similar equation for α¯z
+3(u, z, ¯z). Recalling that on-shell λ1 = 0, substituting (13), (40), (14) and (6a) into (37)
+and extracting the u-independent terms, we find
+δaz
+3 =
+�
+LY + ψ
+�
+az
+3 + 2 ˜Zz − 2 czz DzT − 2 cz¯z D¯zT − 2 T Dzczz + 2 T D¯zcz¯z − 2 az
+2 ξv
+(0) + XB DBaz
+2
+− aB
+2 DBXz + 2 γz¯zD¯za¯z
+2 D¯zT + 2 γz¯zD¯zaz
+2 DzT − 1
+2az
+2a¯z
+2 D¯zψ + 2 γz¯za¯z
+2 D2
+¯zT − 1
+2
+�
+az
+2
+�2 Dzψ
++ az
+2 ∆γT + 2 γz¯zT DzD¯zaz
+2 − 2 γz¯zT D2
+¯za¯z
+2
+(41)
+As in [12], the inhomogeneous part of the variation gives the shift:
+˜Zz = Zz + czz DzT + cz¯z D¯zT + T Dzczz − T D¯zcz¯z + az
+2 ξv
+(0) − 1
+2XBDBaz
+2
++ 1
+2aB
+2 DBXz − γz¯zD¯za¯z
+2D¯zT − γz¯zD¯zaz
+2DzT + 1
+4az
+2a¯z
+2D¯zψ − γz¯za¯z
+2D2
+¯zT + 1
+4
+�
+az
+2
+�2Dzψ
+− 1
+2az
+2∆γT − γz¯zT DzD¯zaz
+2 + γz¯zT D2
+¯za¯z
+2
+(42)
+For completeness, we also present the result for α¯z
+3(u, z, ¯z), which gives an analogous shift for the ¯z-component of the
+subleading hyperrotations:
+˜Z ¯z = Z ¯z + c¯z¯z D¯zT + cz¯z DzT + T D¯zc¯z¯z − T Dzcz¯z + a¯z
+2 ξv
+(0) − 1
+2XBDBa¯z
+2
++ 1
+2aB
+2 DBX ¯z − γz¯zDzaz
+2DzT − γz¯zDza¯z
+2D¯zT + 1
+4a¯z
+2az
+2Dzψ − γz¯zaz
+2D2
+zT + 1
+4
+�
+a¯z
+2
+�2D¯zψ
+− 1
+2a¯z
+2∆γT − γz¯zT D¯zDza¯z
+2 + γz¯zT D2
+zaz
+2
+(43)
+The point of the shifts is that after doing them, the ZA’s are the independent diffeomorphisms. So it is natural to
+demand
+δξZA = 0.
+(44)
+This leads to
+δξ ˜Zz =
+�
+δczz�
+DzT +
+�
+δcz¯z�
+D¯zT + T
+�
+Dzδczz�
+− T
+�
+D¯zδcz¯z�
++
+�
+δaz
+2
+�
+ξv
+(0) + az
+2
+�
+δξξv
+(0)
+�
+− 1
+2XB�
+DBδaz
+2
+�
++ 1
+2
+�
+δaB
+2
+�
+DBXz − γz¯z�
+D¯zδa¯z
+2
+�
+D¯zT − γz¯z�
+D¯zδaz
+2
+�
+DzT + 1
+4a¯z
+2
+�
+δaz
+2
+�
+D¯zψ + 1
+4az
+2
+�
+δa¯z
+2
+�
+D¯zψ − γz¯z�
+δa¯z
+2
+�
+D2
+¯zT
++ 1
+2az
+2
+�
+δaz
+2
+�
+Dzψ − 1
+2
+�
+δaz
+2
+�
+∆γT − γz¯zT
+�
+DzD¯zδaz
+2
+�
++ γz¯zT
+�
+D2
+¯zδa¯z
+2
+�
+(45)
+with a similar expression for δξ ˜Z ¯z. When computing the algebra for the shifted subleading hyperrotations ZA, these
+expressions come in handy for cancelling out certain unpleasant pieces, and leading to the simple form of our final
+
+10
+algebra (19).
+Subleading Hypertranslations: Following the same procedure as in [12], we find
+δλ2 =
+�
+f∂u + LY + ψ
+�
+λ2 − 1
+4 αA
+3 DAψ + 1
+2 ∂uαA
+3 DAf − ξv
+(1) + αA
+2 DAξv
+(0) + αA
+2 DAT + 1
+4αA
+2 DA
+�
+αB
+2 DBψ
+�
+− λ1 ξv
+(0) + ξA
+(1)DAλ1 + ∂uλ1 αA
+2 DAf + 1
+2∂u
+�
+αA
+2 DAαB
+2
+�
+DBf + αA
+2 ∂uαB
+2 DADBf
+(46)
+with LY λ2 = Y A∂Aλ2. By demanding the Einstein equations as in [12], we can write
+λ2 = λ0
+2(z, ¯z) + u λ1
+2(z, ¯z) + Λ2(u, z, ¯z)
+(47)
+where the form of Λ2(u, z, ¯z) will not be important in what follows. This leads to
+δλ0
+2 =
+�
+ψ + LY
+�
+λ0
+2 + T λ1
+2 − ˜τ − 1
+4 aA
+3 DAψ +
+�
+D¯zcz¯z − Dzczz + γz¯zDzD¯zaz
+2 − γz¯zD2
+¯za¯z
+2
+�
+DzT
++
+�
+Dzcz¯z − D¯zc¯z¯z + γz¯zD¯zDza¯z
+2 − γz¯zD2
+zaz
+2
+�
+D¯zT + aA
+2 DAξv
+(0) + aA
+2 DAT + 1
+4aA
+2 DA
+�
+aB
+2 DBψ
+�
+(48)
+The inhomogeneous part of this is the independent subleading hypertranslation, which takes the form
+˜τ = τ − 1
+4 aA
+3 DAψ +
+�
+D¯zcz¯z − Dzczz + γz¯zDzD¯zaz
+2 − γz¯zD2
+¯za¯z
+2
+�
+DzT
++
+�
+Dzcz¯z − D¯zc¯z¯z + γz¯zD¯zDza¯z
+2 − γz¯zD2
+zaz
+2
+�
+D¯zT + aA
+2 DAξv
+(0) + aA
+2 DAT + 1
+4aA
+2 DA
+�
+aB
+2 DBψ
+�
+(49)
+This results in the modified algebra we presented earlier.
+COVARIANT SURFACE CHARGES
+We will compute the covariant surface charges of [14] as in [12], see also [21]. For the set up in the present paper
+putting all the ingredients together leads to a potentially divergent term
+/δQξ[h; g] =
+1
+16πG lim
+v→∞
+�
+d2Ω
+�1
+8
+�
+ψ DAδαA
+2 − 2 YA δαA
+2 − ψ γABδCAB − δαA
+2 DAψ − YA ∂uδαA
+3
+�
+v + O
+�
+v0��
+(50)
+This should be compared to eqn. (54) of [12]. Substituting
+∂uδαz
+3 = 2 D¯zδCz¯z − 2 DzδCzz + 2 γz¯zDzD¯zδαz
+2 − 2 γz¯zD¯zD¯zδα¯z
+2
+(51a)
+∂uδα¯z
+3 = 2 DzδCz¯z − 2 D¯zδC ¯z¯z + 2 γz¯zDzD¯zδα¯z
+2 − 2 γz¯zDzDzδαz
+2
+(51b)
+we obtain
+/δQξ[h; g] =
+1
+16πG lim
+v→∞
+�
+d2Ω
+�1
+4γz¯z �
+Y z�
+D¯zδCzz − DzδCz¯z
+�
++ Y ¯z�
+DzδC¯z¯z − D¯zδCz¯z
+�
+− ψ δCz¯z
+�
++ 1
+4
+�
+Y z DzDzδαz
+2 − δαz
+2 Dzψ + 1
+2Dz
+�
+ψ δαz
+2
+�
+− Y ¯z DzD¯zδαz
+2 − γz¯zY ¯z δαz
+2
++ Y ¯z D¯zD¯zδα¯z
+2 − δα¯z
+2 D¯zψ + 1
+2D¯z
+�
+ψ δα¯z
+2
+�
+− Y z D¯zDzδα¯z
+2 − γz¯zY z δα¯z
+2
+�
+v + O
+�
+v0��
+(52)
+
+11
+We will establish finiteness of the charges by showing that the O
+�
+v
+�
+term vanishes. The terms on the first line in the
+parenthesis at O
+�
+v
+�
+in the above expression can be rewritten as
+Y z�
+D¯zδCzz − DzδCz¯z
+�
++ Y ¯z�
+DzδC¯z¯z − D¯zδCz¯z
+�
+− ψ δCz¯z
+= Y z D¯zδCzz + Y ¯z DzδC¯z¯z − Y zDzδCz¯z − Y ¯zD¯zδCz¯z −
+�
+DzY z + D¯zY ¯z�
+δCz¯z
+= D¯z
+�
+Y z δCzz
+�
++ Dz
+�
+Y ¯z δC¯z¯z
+�
+− Dz
+�
+Y z δCz¯z
+�
+− D¯z
+�
+Y ¯z δCz¯z
+�
+= Dz
+�
+Y ¯z δC¯z¯z − Y z δCz¯z
+�
++ D¯z
+�
+Y z δCzz − Y ¯z δCz¯z
+�
+(53)
+Similarly, the terms on the second line in the parenthesis at O
+�
+v
+�
+can be rewritten as
+Y z DzDzδαz
+2 − δαz
+2 Dzψ + 1
+2Dz
+�
+ψ δαz
+2
+�
+− Y ¯z DzD¯zδαz
+2 − γz¯zY ¯z δαz
+2
+= Y z DzDzδαz
+2 − δαz
+2 DzDzY z − δαz
+2 DzD¯zY ¯z + 1
+2Dz
+�
+ψ δαz
+2
+�
+− Y ¯z DzD¯zδαz
+2 − γz¯zY ¯z δαz
+2
+=
+�
+Y z DzDzδαz
+2 + Dzδαz
+2 DzY z�
+− Dz
+�
+δαz
+2 DzY z�
+− δαz
+2 DzD¯zY ¯z + 1
+2Dz
+�
+ψ δαz
+2
+�
+− Y ¯z DzD¯zδαz
+2 − γz¯zY ¯z δαz
+2
+= Dz(Y zDzδαz
+2) − Dz
+�
+δαz
+2DzY z�
+−
+�
+δαz
+2 D¯zDzY ¯z − γz¯z δαz
+2 Y ¯z�
++ 1
+2Dz
+�
+ψ δαz
+2
+�
++
+�
+DzY ¯z D¯zδαz
+2
+− Dz
+�
+Y ¯zD¯zδαz
+2
+��
+− γz¯zY ¯z δαz
+2
+= Dz(Y zDzδαz
+2) − Dz
+�
+δαz
+2DzY z�
++ 1
+2Dz
+�
+ψ δαz
+2
+�
+− Dz
+�
+Y ¯zD¯zδαz
+2
+�
+(54)
+Note that in writing down the third equality in the above expression, we have commuted the covariant derivatives
+acting on Y ¯z using the definition of the Riemann tensor. That is, we have evaluated [DA, DB]Y C = RC
+DABY D to
+obtain DzD¯zY ¯z − D¯zDzY ¯z = −γz¯zY ¯z. To simplify and obtain the final expression, we have made use of the fact
+that Y z and Y ¯z are holomorphic functions of z and ¯z respectively. A similar procedure can be implemented for δα¯z
+2
+to rewrite the terms on the third line in the parenthesis at O
+�
+v
+�
+as follows
+Y ¯z D¯zD¯zδα¯z
+2 − δα¯z
+2 D¯zψ + 1
+2D¯z
+�
+ψ δα¯z
+2
+�
+− Y z D¯zDzδα¯z
+2 − γz¯zY z δα¯z
+2
+= D¯z(Y ¯zD¯zδα¯z
+2) − D¯z
+�
+δα¯z
+2D¯zY ¯z�
++ 1
+2D¯z
+�
+ψ δα¯z
+2
+�
+− D¯z
+�
+Y zDzδα¯z
+2
+�
+(55)
+After integration over the 2-sphere, the “total” derivative terms disappear. The vanishing of O
+�
+v
+�
+terms guarantees
+that the surface charges remain finite in the limit v → ∞. This is one of our key results in this paper.
+Due to the vanishing of the O
+�
+v
+�
+terms, only the O
+�
+v0�
+terms remain in the v → ∞ limit. These constitute our
+charge expression and they can be evaluated to be
+/δQξ[h; g] =
+1
+16πG
+�
+d2Ω
+�
+u YA δαA
+2 − 3
+8YA δαA
+3 − 1
+4Y AαB
+2 δCAB − f
+2 γz¯z δCz¯z − 1
+4γAB ξA
+(1) δαB
+2 − ψ
+8 γAB αA
+2 δαB
+2
+− ψ
+4 δλ2 + u ψ
+2 γz¯zδCz¯z − 3
+8ψ γz¯zDz¯z − 1
+4Y A δαB
+2 CAB + 3
+16ψ δCAB CAB + ψ γAC δCAB DCαB
+2 + f
+4 DAδαA
+2
+− 1
+4ξv
+(0) DAδαA
+2 − u
+4 ψ DAδαA
+2 + 1
+4δαA
+2 DAξv
+(0) + u
+4 δαA
+2 DAψ − 1
+8δαA
+3 DAψ + ψ
+8 γz¯zCz¯z DAδαA
+2
++ 1
+4γz¯z ψ αA
+2 DAδCz¯z − 1
+4γz¯zδCz¯z αA
+2 DAψ − 1
+8γz¯zCz¯z δαA
+2 DAψ + 1
+8γz¯z ψ δαA
+2 DACz¯z − 1
+4δαA
+2 DAf
++ ψ
+8 DAδαA
+3 + u ψ
+4 γz¯zδCz¯z − ψ
+8 γz¯zδDz¯z − 1
+8Y A δCAB ∂uαB
+3 + u
+2 YA ∂uδαA
+3 + f
+2 γz¯z ∂uδDz¯z − f
+8 N AB δCAB
+− 1
+8γAB ξA
+(1) ∂uδαB
+3 − ψ
+16 γAB αA
+2 ∂uδαB
+3 − 1
+8Y A CAB ∂uδαB
+3 − 1
+8YA ∂uδαA
+4 − f
+4 CAB δNAB
+�
+(56)
+
+12
+Further on, rearranging the terms and simplifying the above expression gives
+/δQξ[h; g] =
+1
+16πG
+�
+d2Ω
+�
+YA
+�
+u δαA
+2 − 3
+8δαA
+3 + u
+2 ∂uδαA
+3 − 1
+8∂uδαA
+4 − 1
+4δ
+�
+CA
+B αB
+2
+�
+− 1
+8δ
+�
+CA
+B ∂uαB
+3
+��
++ ψ
+�
+− u
+2 DAδαA
+2 + 1
+4DAδαA
+3 + 3
+4u γz¯z δCz¯z − 1
+2 γz¯z δDz¯z − 1
+4δλ2 − 1
+8γAB αA
+2 δαB
+2
++ 3
+16CAB δCAB + DAαB
+2 δCAB + 1
+4γz¯z αA
+2 DAδCz¯z + 1
+4δ
+�
+DA(γz¯z Cz¯z αA
+2 )
+�
+− 1
+16γAB αA
+2 ∂uδαB
+3
+�
++ f
+�1
+2DAδαA
+2 − 1
+2γz¯z δCz¯z + 1
+2γz¯z ∂uδDz¯z − 1
+8N AB δCAB − 1
+4 CAB δNAB
+�
++ ξA
+(1)
+�
+− 1
+4γAB δαB
+2 − 1
+8γAB ∂uδαB
+3
+�
++ ξv
+(0)
+�
+− 1
+2DAδαA
+2
+��
+(57)
+Next we can substitute in the shifts that we obtained earlier, namely
+ξA
+(1) = XA − 2 DAf
+(58a)
+ξv
+(0) = φ + f + ∆γf + u
+2 ψ − 1
+2DAXA − 1
+4αA
+2 DAψ
+(58b)
+to write down the final form of the charge expression as follows:
+/δQξ[h; g] =
+1
+16πG
+�
+d2Ω
+�
+YA
+�
+u δαA
+2 − 3
+8δαA
+3 + u
+2 ∂uδαA
+3 − 1
+8∂uδαA
+4 − 1
+4δ
+�
+CA
+B αB
+2
+�
+− 1
+8δ
+�
+CA
+B ∂uαB
+3
+��
++ ψ
+�
+− 3
+4u DAδαA
+2 + 1
+4DAδαA
+3 + 3
+4u γz¯z δCz¯z − 1
+2 γz¯z δDz¯z − 1
+4δλ2 − 1
+8γAB αA
+2 δαB
+2 + 3
+16CAB δCAB
++ DAαB
+2 δCAB + 1
+4γz¯z αA
+2 DAδCz¯z + 1
+4δ
+�
+DA(γz¯z Cz¯z αA
+2 )
+�
+− 1
+16γAB αA
+2 ∂uδαB
+3 − 1
+8DA
+�
+αA
+2 DBδαB
+2
+��
++ f
+�
+− 1
+2DAδαA
+2 − 1
+2γz¯z δCz¯z + 1
+2γz¯z ∂uδDz¯z − 1
+4∂uDAδαA
+3 − 1
+8N AB δCAB − 1
+4 CAB δNAB
+�
+− 1
+2∆γf DAδαA
+2 + XA
+�
+− 1
+4δαA
+2 − 1
+8∂uδαA
+3 − 1
+4DADBδαB
+2
+�
++ φ
+�
+− 1
+2DAδαA
+2
+��
+(59)
+This can be expanded further by substituting the Einstein constraints on the metric parameters
+∂uαz
+3 = 2 D¯zCz¯z − 2 DzCzz + 2 γz¯zDzD¯zαz
+2 − 2 γz¯zD¯zD¯zα¯z
+2
+(60a)
+∂uα¯z
+3 = 2 DzCz¯z − 2 D¯zC ¯z¯z + 2 γz¯zDzD¯zα¯z
+2 − 2 γz¯zDzDzαz
+2
+(60b)
+but we will not do so here.
+The key observation we take away from the final form of the charges is that both leading hypertranslations and
+leading hyperrotations show up in these charges. This should be contrasted to our previous paper [12] where only the
+metric parameters corresponding to these diffeomorphisms showed up, and not the diffeomorphisms themselves. We
+will investigate the physical significance of hypertranslations and their connections to new memory effects in follow
+up work.
+
diff --git a/5tE3T4oBgHgl3EQfQwkt/content/tmp_files/load_file.txt b/5tE3T4oBgHgl3EQfQwkt/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6a7b25446e9e3bb6484b725b798af7e78c2fbd4f
--- /dev/null
+++ b/5tE3T4oBgHgl3EQfQwkt/content/tmp_files/load_file.txt
@@ -0,0 +1,1004 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf,len=1003
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='04415v1 [hep-th] 11 Jan 2023 Charges for Hypertranslations and Hyperrotations Chethan Krishnana and Jude Pereirab aCentre for High Energy Physics, Indian Institute of Science, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Raman Road, Bangalore 560012, India.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Email: chethan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='krishnan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='physics@gmail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='com b Department of Physics, Arizona State University, Tempe, Arizona 85287-1504, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Email: jude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='pereira@asu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='edu Hypertranslations and hyperrotations are asymptotic symmetries of flat space, on top of the familiar supertranslations and superrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' They were discovered in arXiv:2205.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='01422 by working in the Special Double Null (SDN) gauge, where I + and I − are approached along null directions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' It was observed there that while the hair degrees of freedom associated to these diffeomorphisms show up in the covariant surfaces charges, the diffeomorphisms themselves do not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This made their status intermediate in some ways between global symmetries and trivial gauge transformations, making interpretation ambiguous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In this paper, we revisit the fall-offs considered in arXiv:2205.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='01422 which were strictly subleading to Minkowski in conventional double null coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We identify a new class of fall-offs where this assumption is relaxed, but whose charges nonetheless remain finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Remarkably, the leading behavior is still Riemann flat, indicating that these are soft modes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' With this more refined definition of asymptotic flatness, we show that leading hypertranslations and leading hyperrotations explicitly show up in the charges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This makes them genuine global symmetries of asymptotically flat Einstein gravity in the SDN gauge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We write down the new algebra of asymptotic Killing vectors that subsumes the BMS algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Introduction: When the cosmological constant Λ is negative, it is widely believed that the radial direction of the resulting anti-de Sitter (AdS) spacetime is holograph- ically emergent [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' When Λ is positive, less is known, but there are many suggestions in the literature that the timelike direction of de Sitter (dS) space may have a holo- graphic origin [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' These observations make one suspect that it may be useful to view the holographic direction for flat space, which has Λ = 0, as a null coordinate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Historically however, the null boundary of flat space is typically approached along a spacelike direction, eg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=', in the famous Bondi gauge [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' With quite different motivations, various aspects of flat space were explored from a holographic perspective in [4– 6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Along the way, it was realized that a natural gauge for asymptotically flat space is the Special Double Null (SDN) gauge [7], defined by guu = 0, gvv = 0, guA = gvA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (1) Here u and v are null coordinates and I + and I − are covered (generically) by two separate patches around v → ∞ and u → −∞ respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The holographic directions are v and −u in these patches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The notion of a double null coordinate system has been explored in various contexts in the literature before, see eg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [8, 9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' But usually in these settings, not enough constraints are imposed to fix all the coordinate free- dom;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' they are therefore not genuine gauge choices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In fact in the context of mathematical relativity, the form of the double null metric that is sometimes written down (see eg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=', eqn (70) of [10]) does not fall into the gauge we have presented above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This reflects a difference in phi- losophy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' General relativists are interested in I + as the eventual location of gravitational waves from localized objects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' But if one is interested in graviton scattering, as perhaps necessary in quantum gravity, we need access to both I + and I −.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Our gauge has a natural u ↔ −v symmetry which relates I + and I −.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This manifests an asymptotic CPT invariance [7], which is believed to be a symmetry of quantum gravity in flat space [11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In [12] we considered the most general asymptotic sym- metry algebra in SDN gauge with fall-offs which are power laws in the respective null coordinate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Minkowski space in double null coordinates can be obtained by writ- ing u = t − r and v = t + r: ds2 = −du dv + 2 �v − u 2 �2 γz¯zdzd¯z (2) which has guA = gvA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This suggested that one should allow fall-offs where guA and gvA are at most O(v−1) at I +.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The result was found to con- tain new classes of non-trivial asymptotic diffeomor- 2 phisms on top of the BMS symmetries [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' These were named hypertranslations, subleading hypertranslations and subleading hyperrotations1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The algebra of asymp- totic Killing vectors that extends the BMS algebra was identified and the covariant surface charges [14, 15] were computed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' It was noted that these charges had non- trivial dependence on the corresponding “hair” (the met- ric parameters affected by these asymptotic diffeomor- phisms).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' But at the same time, they did not contain the new asymptotic diffeomorphisms themselves, and there- fore the interpretation of these charges was ambiguous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Typically for global symmetries that emerge from an asymptotic symmetry calculation, both the diffeomor- phisms and the associated hair parameters appear in the charge expression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' On the other hand, for trivial diffeo- morphisms, neither the diffeomorphisms nor the parame- ters associated to them appear in the charges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This made the status of hypertranslations and hyperrotations in- termediate between global symmetries and trivial gauge transformations, making them challenging to interpret.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Part of the problem here is that because we are working with null directions, the formalism that is most suited for our purposes is the covariant phase space approach of Wald and followers [14, 15], while a more Hamiltonian- like formalism is perhaps more suited for interpretational purposes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In this paper, we will bypass this problem by identify- ing a new set of fall-offs which are not strictly subleading to (2), but for which the charges are still finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' These fall-offs are presented in (3) and also in more detail in the Supplementary Material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In particular, our fall-offs will allow guA = O(v0) = gvA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' A key feature of these fall-offs is that they can change the metric at an order more lead- ing than (2), and yet remarkably, we are able to show that their charges remain finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In particular, a striking fact that we note is that demanding Riemann flatness allows these terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This allows us to adopt the philosophy that there is nothing too sacred about the specific form in ex- pression (2), it is the demand of Riemann flatness that should be respected in deciding the leading behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We will find that Riemann flatness still leaves the possibility that these modes can be functions of the angular coordi- nates (z, ¯z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We will eventually identify these as related to the hyperrotation hair.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This should be compared to the familiar fact that purely angle dependent shear modes 1In [12], the latter were simply called hyperrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' But in the present paper we will find more leading counterparts to these AKVs which are more naturally called (leading) hyperrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Therefore the ones noted in [12] will be referred to as subleading hyperrotations in this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' in Bondi gauge are soft hair associated to supertrans- lations, and turning them on can still leave the metric Riemann flat [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Similarly in SDN gauge, turning on supertranslation hair or hypertranslation hair, leaves the metric Riemann flat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' But both in Bondi gauge as well as in SDN gauge, the supertranslation and hypertranslation soft modes were subleading to the corresponding conven- tional form of the Minkowski metric.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The new feature of hyperrotation hair here is that it is more leading than (2) while remaining Riemann flat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Riemann flatness in SDN gauge has many remarkable properties, which will be discussed in detail elsewhere [17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Once we adopt these relaxed fall-offs the nature of the calculation is parallel to that in [12], even though techni- cally more involved due to the increased number of metric functions that we start with.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The result of this exercise is that we find that (a) the charges are still finite,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (b) there is a new set of Diff(S2) transformations (the leading hy- perrotations) that appear before subleading hyperrota- tions but are subleading to superrotations,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (c) both the hair parameters as well as the diffeomorphisms associ- ated to the leading hypertranslations and leading hyper- rotations appear in the charge expressions,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' on top of the conventional BMS quantities,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (d) demanding Riemann flatness still allows soft hair associated to these diffeo- morphisms to appear in the metric,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' and (e) the algebra of the asymptotic symmetries is enhanced with respect to both the BMS algebra as well as the BBMS algebra of [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The next section contains the main results of this pa- per.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' To avoid repetition, we will only emphasize aspects of the discussion that are distinct from those in [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In particular, we will simply present the final algebra with- out presenting the details of the derivation – the approach is identical to that in [12], even though technically more involved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Results: We will work with SDN gauge discussed in [7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The fall-offs are presented in great detail in the Supple- mentary Material in terms of functions appearing in the metric.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Here we will write the fall-offs as guv = −2 + O � v−1� (3a) gAB = 4γAB v−2 + O � v−3� (3b) guA = gvA = O � v−2� (3c) Even though technically this is a small change from our previous paper, we emphasize that this is a pretty sub- stantive departure from experience in other gauges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We are demanding that the metric be distinct from the con- ventional form Minkowski metric (2), already at leading order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' There are three reasons why we believe this is 3 reasonable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Firstly, the charges remain finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Secondly, demanding Riemann flatness does not force these terms to be zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Thirdly, with this choice, we get a perfectly conventional structure for the leading hypertranslation and hyperrotation charges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The asymptotic Killing vector conditions take the form: Lξguv = O � v−1� (4a) LξguA = O � v−2� (4b) LξgvA = O � v−2� (4c) LξgAB = O � v−3� (4d) These and the exact Killing conditions (21),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' lead to the solutions: ξu = f + ξu (1) v + ξu (2) v2 + ξu (3) v3 + O � v−4� (5a) ξv = −ψ 2 v + ξv (0) + ξv (1) v + ξv (2) v2 + O � v−3� (5b) ξA = Y A + ξA (1) v + ξA (2) v2 + ξA (3) v3 + O � v−4� (5c) where f = ξu (0) = ψ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) u/2 + T (z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' with ψ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) = DAY A (6a) ξu (1) = αA 2 ∂Af (6b) ξu (2) = 1 2 � αA 3 ∂Af + αA 2 ∂Aξu (1) � (6c) ξu (3) = 1 3 � αA 4 ∂Af + αA 3 ∂Aξu (1) + αA 2 ∂Aξu (2) � (6d) Here T (z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) denotes supertranslations,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' and Y z(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Y ¯z(¯z) denote superrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' On top of the BMS diffeo- morphisms, the ξv (0), ξv (1), ξA (1) and ξA (2) are also de- termined by the exact and asymptotic Killing condi- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The independent functions contained in them are hypertranslations φ(z, ¯z), sub-leading hypertranslations τ(z, ¯z), hyperrotations XA(z, ¯z) and sub-leading hyper- rotations ZA(z, ¯z) respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' They are related to the ξv (0),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ξv (1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ξA (1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ξA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(2) via: ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ξA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(1) = XA − 2 DAf ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(7a) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) = φ + T + △γT − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4aA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAψ − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DAXA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(7b) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(1) = ˜τ + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2A A ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(7c) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ξA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(2) = ˜ZA + C AB DBψ + A A ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ψ − u XA + 2 u DAξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− u2 DAψ − L1 DAψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='We have introduced ˜τ and ˜ZA for convenience which are ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='related to the sub-leading hypertranslations τ and sub- ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='leading hyperrotations ZA via ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='˜τ = τ − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 aA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 DAψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zcz¯z − Dzczz + γz¯zDzD¯zaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − γz¯zD2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='¯za¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DzT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Dzcz¯z − D¯zc¯z¯z + γz¯zD¯zDza¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − γz¯zD2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='zaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ aA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) + aA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAT + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4aA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='aB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DBψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(8) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='˜Zz = Zz + czz DzT + cz¯z D¯zT + T Dzczz − T D¯zcz¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2XBDBaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2aB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DBXz − γz¯zD¯za¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2D¯zT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− γz¯zD¯zaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DzT + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2a¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2D¯zψ − γz¯za¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2D2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='¯zT + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�2Dzψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2∆γT − γz¯zT DzD¯zaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + γz¯zT D2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='¯za¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(9) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='˜Z ¯z = Z ¯z + c¯z¯z D¯zT + cz¯z DzT + T D¯zc¯z¯z − T Dzcz¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ a¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2XBDBa¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2aB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DBX ¯z − γz¯zDzaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DzT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− γz¯zDza¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2D¯zT + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4a¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2Dzψ − γz¯zaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2D2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='zT + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='a¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�2D¯zψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2a¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2∆γT − γz¯zT D¯zDza¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + γz¯zT D2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='zaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(10) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='These expressions are significantly more complicated ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='than those in [12],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' so let us pause to explain some of the details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The integration “constants” in the shear are2 introduced via CAB(u, z, ¯z) = cAB(z, ¯z) + � u −∞ du′NAB(u′, z, ¯z)(11) with NAB ≡ ∂uCAB, being the SDN news tensor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Sim- ilarly, we have defined the integration “constant” in αA 2 as αA 2 (u, z, ¯z) = aA 2 (z, ¯z) + � u −∞ du′βA 2 (u′, z, ¯z) (12) where βA 2 ≡ ∂uαA 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' See [7, 18] for a discussion on integrals of this type that are defined from I + − to u.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' On shell (ie.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=', when Einstein equations hold), we have Nz¯z = 0 and βA 2 = 0, so we will have Cz¯z(u, z, ¯z) = cz¯z(z, ¯z) αA 2 (u, z, ¯z) = aA 2 (z, ¯z) (13) In addition to this, the Einstein constraints also require 2The notation here is slightly different from that in [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 4 that λ1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' For ξA (2), combining all the relevant equa- tions, we can write [18] ∂uξA (2) = CAB DBψ − 2 u DAψ + 2 DAξv (0) + αA 2 ψ − λ1DAψ − XA =⇒ ξA (2) = C AB DBψ − u2 DAψ + 2 u DAξv (0) + A A 2 ψ − L1DAψ − u XA + ˜ZA(z, ¯z) (14) The u-independence of ψ, ξv (0) and XA has been used in writing the integrated version in the second step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Also C AB, A A 2 and L2 have been defined via ∂uC AB = CAB(u, z, ¯z) (15) ∂uA A 2 = αA 2 (u, z, ¯z) (16) ∂uL1 = λ1(u, z, ¯z) (17) As in [12], ˜ZA(z, ¯z) is taken as the u-independent piece in ξA (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The shift is done on ˜ZA via (9)-(10) and the result is what we call sub-leading hyperrotations ZA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The rest of the notation follows that of [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' As empha- sized there, the idea in (7) is to do certain shifts so that the structure of the diffeomorphisms is cleanest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This “diagonalizes” the algebra of diffeomorphisms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The phi- losophy here is identical, even though the expressions are more complicated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The hair associated to the various diffeomorphisms are therefore as follows: supertranslations T (z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) are associ- ated to the u-independent shifts in Czz and C¯z¯z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' hyper- translations φ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) are associated to the u-independent shifts of Cz¯z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' subleading hypertranslations τ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) are as- sociated to u-independent shifts of λ2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' hyperrotations XA(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) are associated to u-independent shifts of α A 2 and sub-leading hyperrotations ZA(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) are associated to u-independent shifts of α A 3 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' As in Bondi gauge, we also have superrotations Y z(z), Y ¯z(¯z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The shifts in- volved in the definitions of φ, τ, XA and ZA are detailed in the Supplementary Material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' As in [12], supertrans- lations and leading-&-subleading hypertranslations are diffeomorphisms of u and v respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Leading hyper- rotations were not present in [12], but both leading and subleading hyperrotations are subleading to superrota- tions on the sphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We will define the “Beyond BBMS” algebra b2-bms4 as the asymptotic symmetry algebra of the nine non- trivial diffeomorphisms – supertranslations, superrota- tions, hypertranslations & subleading hypertranslations, and hyperrotations & subleading hyperrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Follow- ing [12, 13], we define the bracket ��Y , �T, �φ, �τ, � X, �Z � = � (Y1, T1, φ1, τ1, X1, Z1), (Y2, T2, τ2, φ2, X2, Z2) � (18) The notation is the natural generalization of that in [12, 13] and the reader should consult those papers for the detailed definitions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The new algebra is defined via �Y , �T, �φ, �τ, � X and �Z given by the following expressions: �Y A = Y B 1 ∂BY A 2 − Y B 2 ∂BY A 1 (19a) �T = Y A 1 ∂AT2 − Y A 2 ∂AT1 + 1 2 (T1 ψ2 − T2 ψ1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (19b) �φ = 1 2(ψ1φ2 − ψ2φ1) + � Y A 1 ∂Aφ2 − Y A 2 ∂Aφ1 � (19c) �τ = (ψ1τ2 − ψ2τ1) + � Y A 1 ∂Aτ2 − Y A 2 ∂Aτ1 � (19d) � XA = 1 2 � ψ1XA 2 − ψ2XA 1 � + � Y B 1 ∂BXA 2 − Y B 2 ∂BXA 1 � + � XB 1 ∂BY A 2 − XB 2 ∂BY A 1 � (19e) �ZA = � ψ1ZA 2 − ψ2ZA 1 � + � Y B 1 ∂BZA 2 − Y B 2 ∂BZA 1 � + � ZB 1 ∂BY A 2 − ZB 2 ∂BY A 1 � (19f) This is what we call the b2-bms4 algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The fact that these nine non-trivial diffeomorphisms form a closed al- gebra is checked by the same procedure as outlined in [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The calculations are straightforward but lengthier variations of those there.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' In order to identify the capped quantities, we need to consider the Barnich-Troessaert bracket [ξ1, ξ2]M of two AKVs ξ1 and ξ2 [12, 13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The structure is parallel to that presented in [12], with a no- 5 table difference in the A-component which takes the form [ξ1, ξ2]A M = �Y A + �ξA (1) v + �ξA (2) v2 + O � v−3� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (20) In computing all four components of the Barnich- Troessaert bracket, we need �ξv (0), �ξv (1), �ξA (1) and �ξA (2), which are defined as in (7) but with Y A, T, φ, τ, XA, ZA re- placed by their capped versions, defined in (19).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Equations (19) define the b2-bms4 algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Setting the hyperrotations XA to zero results in the BBMS algebra of [12], and setting φ, τ and ZA as well to zero results in the familiar BMS algebra [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Discussion: In this paper, we observed that demand- ing finite covariant surface charges in Einstein gravity allows fall-offs that are not necessarily subleading to (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Turning on the soft modes associated to supertransla- tions and leading hypertranslations/hyperrotations takes us beyond (2) even though the metric is still Riemann flat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We exploited this fact to work with fall-offs that allowed these modes, to show that the covariant surface charges contain these diffeomorphisms as well as the as- sociated soft hair.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This places them on an equal footing with conventional global symmetries (eg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' supertransla- tions), resolving some of the ambiguities pointed out in [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Of course, these results open up further questions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Our work strongly suggests that the charges associated to hy- pertranslations should be interpreted as soft, so it would be interesting to connect these results to soft theorems (perhaps to the subsubleading soft graviton theorem of [20]?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=') and also to new memory effects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Some of these questions are currently under investigation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Hypertrans- lations have many similarities to supertranslations, but there are also crucial distinctions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The lowest modes of supertranslations are simply the action of Poincare trans- lations on the boundary (u, z, ¯z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Hypertranslations on the other hand are truly distinct from bulk translations – we have already subtracted out the supertranslations in our shifted diffeomorphisms, when defining hypertransla- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' It should be clear from (5) that the interpretation of hypertranslations is more like a bulk diffeomorphism at infinity (note that infinity is along the null direction v in SDN gauge).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' It is more naturally compared to ξr than ξu in Bondi gauge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' A related interesting feature of hypertranslations and their associated hair is that they can be spherically sym- metric.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This raises a subtlety in the usual statement of Birkhoff’s theorem, which will be discussed in an up- coming work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Note that while supertranslations allow soft hair on Schwarzschild, the only spherically symmet- ric supertranslation is a time translation, so this subtlety does not arise for Schwarzschild in Bondi gauge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' It is also important to emphasize that hypertranslations should be distinguished from the shifts in v at the past boundary I −.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The latter are simply supertranslations, but now acting in the past.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' What we mean by hypertranslations are shifts in v at I +.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' There is no obvious connection between the two (other than the future-past matching at i0 that was discussed in [7]) because these coordinates live in different charts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' What about subleading hypertranslations and sublead- ing hyperrotations?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' They do not show up in the charges even with the new fall-offs, but their associated hair was present both in [12] as well as here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' So their interpre- tation remains ambiguous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' It is natural to consider the sub-algebra obtained by setting the subleading hyper- translations/hyperrotations to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This would mean that we are working with supertranslations, superrota- tions, leading hypertranslations and leading hyperrota- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This is a natural generalization of the conventional BMS algebra in the SDN gauge;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' it is clearly of interest to study it more closely.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' One could also consider the even simpler generalization of BMS, obtained by adding only the leading hypertranslations and suppressing the lead- ing hyperrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This algebra has the advantage that we are not turning on diffeomorphisms on the sphere, but only the Virasoro (super)rotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' While it may be difficult to conclusively argue for such a choice from a purely asymptotic symmetry perspective, it is natural from a celestial holography perspective [19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This is the algebra of supertranslations, (leading) hypertranslations and superrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ACKNOWLEDGMENTS We thank Sudip Ghosh and Sarthak Talukdar for dis- cussions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [1] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Witten, “Anti-de Sitter space and hologra- phy,” Adv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Theor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 2, 253-291 (1998) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4310/ATMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='v2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='n2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='a2 [arXiv:hep-th/9802150 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [2] See eg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=', V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Balasubramanian, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' de Boer and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Minic, “Notes on de Sitter space and holography,” Class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Quant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Grav.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 19, 5655-5700 (2002) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1016/S0003- 4916(02)00020-9 [arXiv:hep-th/0207245 [hep-th]], and the first few references therein.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [3] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Bondi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' van der Burg and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Met- zner, “Gravitational waves in general relativity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Waves from axisymmetric isolated systems,” Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Roy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 6 Lond.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' A 269, 21-52 (1962) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1098/rspa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1962.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='0161.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Sachs, “Gravitational waves in general rela- tivity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Waves in asymptotically flat space-times,” Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Roy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Lond.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' A 270, 103-126 (1962) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1098/rspa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1962.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='0206;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Sachs, “Asymptotic sym- metries in gravitational theory,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 128, 2851- 2864 (1962) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1103/PhysRev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='128.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2851 [4] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Bhattacharjee and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan, “A General Prescrip- tion for Semi-Classical Holography,” [arXiv:1908.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='04786 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [5] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan, “Bulk Locality and Asymptotic Causal Diamonds,” SciPost Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 7, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4, 057 (2019) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='21468/SciPostPhys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='057 [arXiv:1902.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='06709 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [6] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Patil and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Pereira, “Page Curve and the Information Paradox in Flat Space,” [arXiv:2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='02993 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [7] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Pereira, “A New Gauge for Asymp- totically Flat Spacetime,” [arXiv:2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='11440 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [8] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Wald, “General Relativity,” Chicago Univ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Pr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=', 1984, doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='7208/chicago/9780226870373.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='0001 [9] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Brady, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Droz, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Israel and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Morsink, “Covariant double null dynamics: (2+2) splitting of the Einstein equations,” Class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Quant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Grav.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 13, 2211-2230 (1996) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1088/0264-9381/13/8/015 [arXiv:gr-qc/9510040 [gr-qc]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [10] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Dafermos, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Holzegel and I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Rodnianski, “The linear stability of the Schwarzschild solution to gravi- tational perturbations,” Acta Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' 222, 1-214 (2019) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4310/ACTA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='v222.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='n1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='a1 [arXiv:1601.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='06467 [gr-qc]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [11] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Strominger, “On BMS Invariance of Grav- itational Scattering,” JHEP 07, 152 (2014) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1007/JHEP07(2014)152 [arXiv:1312.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2229 [hep- th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [12] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Pereira, “Hypertranslations and Hy- perrotations,” [arXiv:2205.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='01422 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [13] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Barnich and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Troessaert, “Aspects of the BMS/CFT correspondence,” JHEP 05, 062 (2010) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1007/JHEP05(2010)062 [arXiv:1001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1541 [hep- th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [14] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Iyer and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Wald, “Some properties of Noether charge and a proposal for dynamical black hole entropy,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' D 50, 846-864 (1994) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='846 [arXiv:gr-qc/9403028 [gr- qc]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [15] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Barnich and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Brandt, “Covariant theory of asymp- totic symmetries, conservation laws and central charges,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' B 633, 3-82 (2002) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1016/S0550- 3213(02)00251-1 [arXiv:hep-th/0111246 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [16] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Strominger, “Lectures on the Infrared Structure of Gravity and Gauge Theory,” [arXiv:1703.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='05448 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [17] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Pereira, “Asymptotically Riemann- flat Spacetimes,” to appear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [18] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Krishnan and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Pereira,“A New Gauge for Flat Space Holography,” to appear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [19] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Schwarz, “Diffeomorphism Symmetry in Two Di- mensions and Celestial Holography,” [arXiv:2208.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='13304 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [20] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Cachazo and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Strominger, “Evidence for a New Soft Graviton Theorem,” [arXiv:1404.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4091 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' [21] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Barnich and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Troessaert, “BMS charge algebra,” JHEP 12, 105 (2011) doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1007/JHEP12(2011)105 [arXiv:1106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='0213 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Supplementary material REFINED FALL-OFFS In this section, we will present the falloffs in some detail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Our emphasis will be on the distinctions from those presented in [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We start with a quick review of the notation: in d + 1 dimensions, the SDN gauge [7] is defined by eqn (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We will restrict ourselves to 3+1 dimensions here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The exact Killing vector equations are Lξguu = 0, Lξgvv = 0, LξguA = LξgvA (21) and we will write the general metric in this gauge as ds2 = −eλdu dv + �v − u 2 �2 ΩAB(dxA − αAdu − αAdv)(dxB − αBdu − αBdv) (22) 7 In [12], we presented a set of fall-offs in terms of the functions in this ansatz, which the reader should consult.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The fall-offs we consider in this paper are distinct in the following functions: λ(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) = λ1(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v + λ2(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v2 + λ3(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v3 + λ4(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v4 + O � v−5� (23a) αz(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) = αz 2(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v2 + αz 3(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v3 + αz 4(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v4 + αz 5(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v5 + O � v−6� (23b) α¯z(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) = α¯z 2(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v2 + α¯z 3(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v3 + α¯z 4(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v4 + α¯z 5(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z) v5 + O � v−6� (23c) In terms of the metric,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' this results in the fall-offs: guu = gvv = O � v−2� (24a) guv = −1 2 + O � v−1� (24b) gAB = 1 4 γAB v2 + O(v) (24c) guA = gvA = O � v0� (24d) Compared to the discussion in [12],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' we also allow αA 2 as the O(1/v2) term in the αA fall-off.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Just like Cz¯z, αA 2 also turns out to be u-independent once we demand Einstein equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Hence it is an integration “constant” in Einstein constraints in the language of [7, 12, 18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Demanding Ricci (or Riemann) flatness forces λ1 to be zero and αA 2 to be functions only of the angles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We have kept them general in the discussions of the AKVs because they can be defined on arbitrary backgrounds, without worrying about the equations satisfied by those backgrounds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' But one can in principle start a-priori with fall-offs (23) where λ1 is set to zero and αA 2 are functions only of z and ¯z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Some of the expressions we have presented will simplify somewhat in that case, but the main results do not change.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' DIFFEOMORPHISM SHIFTS As in [12] we will define the various diffeomorphisms after a suitable shift in the fall-off coefficient of ξ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This is more elaborate in the present case, and we discuss them in detail below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The philosophy behind these shifts was discussed in [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Hyperrotations: The simplest case arises for the leading hyperrotations XA(z, ¯z), so we start with them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' From the exact Lie derivative conditions, we obtain the following constraint on ξA 1 , ∂uξA 1 = −DAψ (25) which on integrating both sides becomes ξA 1 = ˜XA − u DAψ (26) The metric function corresponding to leading hyperrotations is αA 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Under the action of AKVs, the transformation of αA 2 can be obtained by evaluating δξguA = LξguA at O(v−2) as follows δαA 2 = � f∂u + LY + ψ 2 � αA 2 + ˜XA − u DAψ + 2 DAf (27) where LY αA 2 = Y B ∂BαA 2 − αB 2 ∂BY A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (28) 8 is the Lie derivative of αA 2 with respect to Y A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Recalling that on-shell αA 2 = aA 2 (z, ¯z) and substituting f = ψ(z, ¯z) u/2+ T (z, ¯z), we obtain δaA 2 = � LY + ψ 2 � aA 2 + ˜XA + 2 DAT (29) Next we would like to interpret XA(z, ¯z) as the diffeomorphism that causes αA 2 to be turned on if it was initially zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This immediately suggests the following shift ˜XA = XA − 2 DAT (30) Substituting this in (26) and using f = � ψ/2 � u + T yields ξA 1 = XA − 2 DAf (31) Hypertranslations: In the case of the leading hypertranslations φ(z, ¯z), the shift is of the form ξv (0) = φ + T + △γT − 1 4aA 2 DAψ − 1 2DAXA (32) This reduces to the form presented in [12] when the hyperrotations and their hair are set to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The change in Cz¯z can be computed by evaluating δξgz¯z = Lξgz¯z at O(v−3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The result is δCz¯z = � f ∂u + LY − 1 2 ψ � Cz¯z − 4 ∂z∂¯zf + 2 γz¯z � ξv (0) − f − u 2 ψ + 1 2DAXA + 1 4αA 2 DAψ � (33) Here LY is the Lie derivative of Cz¯z with respect to Y A defined as in [12]: LY Cz¯z = Y A∂ACz¯z + � ∂AY A� Cz¯z (34) On-shell we have Cz¯z = cz¯z(z, ¯z) and αA 2 = aA 2 (z, ¯z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Using these and substituting f = ψ(z, ¯z) u/2 + T (z, ¯z), we obtain δcz¯z = � LY − 1 2 ψ � cz¯z + 2γz¯z � ξv (0) − T − ∆γT + 1 2DAXA + 1 4aA 2 DAψ � (35) It is clear that ξv (0) mixes with supertranslations, superrotations and leading hyperrotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We wish to remove this mixing, so that we can interpret φ(z, ¯z) as the diffeomorphism that causes cz¯z to be turned on if it was initially zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' From this it follows that the shift is ξv (0) = φ + T + △γT − 1 4aA 2 DAψ − 1 2DAXA, as we presented above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This defines hypertranslations, φ(z, ¯z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Note that in deriving the algebra for hypertranslations, we have made use of the identity δξξv (0) = −1 4 � δaA 2 � DAψ (36) where we have demanded that δξφ = 0 and δξXA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This shifted definition above of the hypertranslations ensures the vanishing of the hatted �φ on the left hand side of algebra, when φ1 and φ2 are zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' As we pointed out in [12], this feature can be viewed as one of the motivations behind doing the shifts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This generalizes to the other diffeomorphisms as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Subleading Hyperrotations: Now we turn to the case of subleading hyperrotations ZA(z, ¯z) and the correspond- ing metric functions αA 3 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The same procedure as in [12] now yields δαz 3 = � f ∂u + LY + ψ � αz 3 + 2 ξz (2) + 4 u Dzf − 2 CzB DBf − 2 αz 2 ξv (0) + XBDBαz 2 − αB 2 DBXz + 2 γz¯zD¯zα¯z 2D¯zf + 2 γz¯zD¯zαz 2DzT − 1 2αz 2α¯z 2D¯zψ + 2 γz¯zα¯z 2D2 ¯zT + 2u γz¯zα¯z 2D2 ¯zψ − u γz¯zDzαz 2D¯zψ − 1 2 � αz 2 �2Dzψ − 2u αz 2ψ + αz 2∆γT + 2 λ1Dzf + ∂uαz 2 � αA 2 DAf � (37) 9 where the Lie derivative is defined as LY αA 3 = Y B ∂BαA 3 − αB 3 ∂BY A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (38) Note that in obtaining the above equation, we have used (27) along with δλ1 = � f ∂u + LY + 1 2ψ � λ1 + ∂uαA 2 DAf (39) which has been obtained by evaluating δξguv = Lξguv at O(v−1) where LY λ1 = Y A∂Aλ1 is the Lie derivative of λ1 with respect to Y A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' On-shell, we have [18] ∂uαz 3 = −2 DzCzz + 2 D¯zcz¯z + 2 γz¯zDzD¯zaz 2 − 2 γz¯zD2 ¯za¯z 2 =⇒ αz 3(u, z, ¯z) = −2 DzC zz + u � 2 D¯zcz¯z + 2 γz¯zDzD¯zaz 2 − 2 γz¯zD2 ¯za¯z 2 � + az 3(z, ¯z) (40) and a similar equation for α¯z 3(u, z, ¯z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Recalling that on-shell λ1 = 0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' substituting (13),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (40),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (14) and (6a) into (37) and extracting the u-independent terms,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' we find δaz 3 = � LY + ψ � az 3 + 2 ˜Zz − 2 czz DzT − 2 cz¯z D¯zT − 2 T Dzczz + 2 T D¯zcz¯z − 2 az 2 ξv (0) + XB DBaz 2 − aB 2 DBXz + 2 γz¯zD¯za¯z 2 D¯zT + 2 γz¯zD¯zaz 2 DzT − 1 2az 2a¯z 2 D¯zψ + 2 γz¯za¯z 2 D2 ¯zT − 1 2 � az 2 �2 Dzψ + az 2 ∆γT + 2 γz¯zT DzD¯zaz 2 − 2 γz¯zT D2 ¯za¯z 2 (41) As in [12],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' the inhomogeneous part of the variation gives the shift: ˜Zz = Zz + czz DzT + cz¯z D¯zT + T Dzczz − T D¯zcz¯z + az 2 ξv (0) − 1 2XBDBaz 2 + 1 2aB 2 DBXz − γz¯zD¯za¯z 2D¯zT − γz¯zD¯zaz 2DzT + 1 4az 2a¯z 2D¯zψ − γz¯za¯z 2D2 ¯zT + 1 4 � az 2 �2Dzψ − 1 2az 2∆γT − γz¯zT DzD¯zaz 2 + γz¯zT D2 ¯za¯z 2 (42) For completeness,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' we also present the result for α¯z 3(u,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ¯z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' which gives an analogous shift for the ¯z-component of the subleading hyperrotations: ˜Z ¯z = Z ¯z + c¯z¯z D¯zT + cz¯z DzT + T D¯zc¯z¯z − T Dzcz¯z + a¯z 2 ξv (0) − 1 2XBDBa¯z 2 + 1 2aB 2 DBX ¯z − γz¯zDzaz 2DzT − γz¯zDza¯z 2D¯zT + 1 4a¯z 2az 2Dzψ − γz¯zaz 2D2 zT + 1 4 � a¯z 2 �2D¯zψ − 1 2a¯z 2∆γT − γz¯zT D¯zDza¯z 2 + γz¯zT D2 zaz 2 (43) The point of the shifts is that after doing them,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' the ZA’s are the independent diffeomorphisms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' So it is natural to demand δξZA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(44) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='This leads to ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δξ ˜Zz = ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δczz� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DzT + ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δcz¯z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zT + T ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Dzδczz� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− T ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zδcz¯z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) + az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δξξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2XB� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DBδaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δaB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DBXz − γz¯z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zδa¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zT − γz¯z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zδaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DzT + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4a¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zψ + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δa¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D¯zψ − γz¯z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δa¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='¯zT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2az ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Dzψ − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='∆γT − γz¯zT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DzD¯zδaz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ γz¯zT ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='D2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='¯zδa¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(45) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='with a similar expression for δξ ˜Z ¯z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' When computing the algebra for the shifted subleading hyperrotations ZA, these expressions come in handy for cancelling out certain unpleasant pieces, and leading to the simple form of our final 10 algebra (19).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Subleading Hypertranslations: Following the same procedure as in [12], we find δλ2 = � f∂u + LY + ψ � λ2 − 1 4 αA 3 DAψ + 1 2 ∂uαA 3 DAf − ξv (1) + αA 2 DAξv (0) + αA 2 DAT + 1 4αA 2 DA � αB 2 DBψ � − λ1 ξv (0) + ξA (1)DAλ1 + ∂uλ1 αA 2 DAf + 1 2∂u � αA 2 DAαB 2 � DBf + αA 2 ∂uαB 2 DADBf (46) with LY λ2 = Y A∂Aλ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' By demanding the Einstein equations as in [12], we can write λ2 = λ0 2(z, ¯z) + u λ1 2(z, ¯z) + Λ2(u, z, ¯z) (47) where the form of Λ2(u, z, ¯z) will not be important in what follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This leads to δλ0 2 = � ψ + LY � λ0 2 + T λ1 2 − ˜τ − 1 4 aA 3 DAψ + � D¯zcz¯z − Dzczz + γz¯zDzD¯zaz 2 − γz¯zD2 ¯za¯z 2 � DzT + � Dzcz¯z − D¯zc¯z¯z + γz¯zD¯zDza¯z 2 − γz¯zD2 zaz 2 � D¯zT + aA 2 DAξv (0) + aA 2 DAT + 1 4aA 2 DA � aB 2 DBψ � (48) The inhomogeneous part of this is the independent subleading hypertranslation,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' which takes the form ˜τ = τ − 1 4 aA 3 DAψ + � D¯zcz¯z − Dzczz + γz¯zDzD¯zaz 2 − γz¯zD2 ¯za¯z 2 � DzT + � Dzcz¯z − D¯zc¯z¯z + γz¯zD¯zDza¯z 2 − γz¯zD2 zaz 2 � D¯zT + aA 2 DAξv (0) + aA 2 DAT + 1 4aA 2 DA � aB 2 DBψ � (49) This results in the modified algebra we presented earlier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' COVARIANT SURFACE CHARGES We will compute the covariant surface charges of [14] as in [12], see also [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' For the set up in the present paper putting all the ingredients together leads to a potentially divergent term /δQξ[h;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' g] = 1 16πG lim v→∞ � d2Ω �1 8 � ψ DAδαA 2 − 2 YA δαA 2 − ψ γABδCAB − δαA 2 DAψ − YA ∂uδαA 3 � v + O � v0�� (50) This should be compared to eqn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' (54) of [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Substituting ∂uδαz 3 = 2 D¯zδCz¯z − 2 DzδCzz + 2 γz¯zDzD¯zδαz 2 − 2 γz¯zD¯zD¯zδα¯z 2 (51a) ∂uδα¯z 3 = 2 DzδCz¯z − 2 D¯zδC ¯z¯z + 2 γz¯zDzD¯zδα¯z 2 − 2 γz¯zDzDzδαz 2 (51b) we obtain /δQξ[h;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' g] = 1 16πG lim v→∞ � d2Ω �1 4γz¯z � Y z� D¯zδCzz − DzδCz¯z � + Y ¯z� DzδC¯z¯z − D¯zδCz¯z � − ψ δCz¯z � + 1 4 � Y z DzDzδαz 2 − δαz 2 Dzψ + 1 2Dz � ψ δαz 2 � − Y ¯z DzD¯zδαz 2 − γz¯zY ¯z δαz 2 + Y ¯z D¯zD¯zδα¯z 2 − δα¯z 2 D¯zψ + 1 2D¯z � ψ δα¯z 2 � − Y z D¯zDzδα¯z 2 − γz¯zY z δα¯z 2 � v + O � v0�� (52) 11 We will establish finiteness of the charges by showing that the O � v � term vanishes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The terms on the first line in the parenthesis at O � v � in the above expression can be rewritten as Y z� D¯zδCzz − DzδCz¯z � + Y ¯z� DzδC¯z¯z − D¯zδCz¯z � − ψ δCz¯z = Y z D¯zδCzz + Y ¯z DzδC¯z¯z − Y zDzδCz¯z − Y ¯zD¯zδCz¯z − � DzY z + D¯zY ¯z� δCz¯z = D¯z � Y z δCzz � + Dz � Y ¯z δC¯z¯z � − Dz � Y z δCz¯z � − D¯z � Y ¯z δCz¯z � = Dz � Y ¯z δC¯z¯z − Y z δCz¯z � + D¯z � Y z δCzz − Y ¯z δCz¯z � (53) Similarly,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' the terms on the second line in the parenthesis at O ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='v ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='can be rewritten as ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Y z DzDzδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 Dzψ + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ψ δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− Y ¯z DzD¯zδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − γz¯zY ¯z δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='= Y z DzDzδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DzDzY z − δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DzD¯zY ¯z + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ψ δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− Y ¯z DzD¯zδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − γz¯zY ¯z δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='= ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Y z DzDzδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + Dzδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DzY z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DzY z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DzD¯zY ¯z + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ψ δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− Y ¯z DzD¯zδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − γz¯zY ¯z δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='= Dz(Y zDzδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2) − Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DzY z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 D¯zDzY ¯z − γz¯z δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 Y ¯z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ψ δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DzY ¯z D¯zδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Y ¯zD¯zδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− γz¯zY ¯z δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='= Dz(Y zDzδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2) − Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DzY z� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='ψ δαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− Dz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Y ¯zD¯zδαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(54) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Note that in writing down the third equality in the above expression,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' we have commuted the covariant derivatives acting on Y ¯z using the definition of the Riemann tensor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' That is, we have evaluated [DA, DB]Y C = RC DABY D to obtain DzD¯zY ¯z − D¯zDzY ¯z = −γz¯zY ¯z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' To simplify and obtain the final expression, we have made use of the fact that Y z and Y ¯z are holomorphic functions of z and ¯z respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' A similar procedure can be implemented for δα¯z 2 to rewrite the terms on the third line in the parenthesis at O � v � as follows Y ¯z D¯zD¯zδα¯z 2 − δα¯z 2 D¯zψ + 1 2D¯z � ψ δα¯z 2 � − Y z D¯zDzδα¯z 2 − γz¯zY z δα¯z 2 = D¯z(Y ¯zD¯zδα¯z 2) − D¯z � δα¯z 2D¯zY ¯z� + 1 2D¯z � ψ δα¯z 2 � − D¯z � Y zDzδα¯z 2 � (55) After integration over the 2-sphere, the “total” derivative terms disappear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The vanishing of O � v � terms guarantees that the surface charges remain finite in the limit v → ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This is one of our key results in this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' Due to the vanishing of the O � v � terms, only the O � v0� terms remain in the v → ∞ limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' These constitute our charge expression and they can be evaluated to be /δQξ[h;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' g] = ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16πG ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='d2Ω ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='u YA δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8YA δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4Y AαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 δCAB − f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 γz¯z δCz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4γAB ξA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(1) δαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8 γAB αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 δαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 δλ2 + u ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 γz¯zδCz¯z − 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8ψ γz¯zDz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4Y A δαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 CAB + 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16ψ δCAB CAB + ψ γAC δCAB DCαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − u ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 ψ DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) + u ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAψ − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 DAψ + ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8 γz¯zCz¯z DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4γz¯z ψ αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAδCz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4γz¯zδCz¯z αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAψ − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8γz¯zCz¯z δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAψ + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8γz¯z ψ δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DACz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAf ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8 DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + u ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 γz¯zδCz¯z − ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8 γz¯zδDz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8Y A δCAB ∂uαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + u ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 YA ∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 γz¯z ∂uδDz¯z − f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8 N AB δCAB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8γAB ξA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(1) ∂uδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16 γAB αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ∂uδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8Y A CAB ∂uδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8YA ∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 − f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 CAB δNAB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(56) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Further on,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' rearranging the terms and simplifying the above expression gives /δQξ[h;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' g] = ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16πG ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='d2Ω ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='YA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='u δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + u ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='CA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='B αB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='CA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='B ∂uαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− u ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4u γz¯z δCz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 γz¯z δDz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δλ2 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8γAB αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 δαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16CAB δCAB + DAαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 δCAB + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4γz¯z αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAδCz¯z + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DA(γz¯z Cz¯z αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16γAB αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ∂uδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2γz¯z δCz¯z + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2γz¯z ∂uδDz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8N AB δCAB − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 CAB δNAB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ξA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(1) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4γAB δαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8γAB ∂uδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ξv ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(0) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(57) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='Next we can substitute in the shifts that we obtained earlier,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' namely ξA (1) = XA − 2 DAf (58a) ξv (0) = φ + f + ∆γf + u 2 ψ − 1 2DAXA − 1 4αA 2 DAψ (58b) to write down the final form of the charge expression as follows: /δQξ[h;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' g] = ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16πG ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='d2Ω ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='YA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='u δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + u ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='CA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='B αB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='CA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='B ∂uαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ ψ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4u DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 + 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4u γz¯z δCz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 γz¯z δDz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δλ2 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8γAB αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 δαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16CAB δCAB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ DAαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 δCAB + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4γz¯z αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DAδCz¯z + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='DA(γz¯z Cz¯z αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='16γAB αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ∂uδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8DA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='αA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 DBδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ f ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2γz¯z δCz¯z + 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2γz¯z ∂uδDz¯z − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4∂uDAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8N AB δCAB − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4 CAB δNAB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2∆γf DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 + XA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4δαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='8∂uδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 − 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='4DADBδαB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='+ φ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='− 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2DAδαA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(59) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='This can be expanded further by substituting the Einstein constraints on the metric parameters ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='∂uαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 = 2 D¯zCz¯z − 2 DzCzz + 2 γz¯zDzD¯zαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 2 γz¯zD¯zD¯zα¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(60a) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='∂uα¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='3 = 2 DzCz¯z − 2 D¯zC ¯z¯z + 2 γz¯zDzD¯zα¯z ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 − 2 γz¯zDzDzαz ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='(60b) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content='but we will not do so here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' The key observation we take away from the final form of the charges is that both leading hypertranslations and leading hyperrotations show up in these charges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' This should be contrasted to our previous paper [12] where only the metric parameters corresponding to these diffeomorphisms showed up, and not the diffeomorphisms themselves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
+page_content=' We will investigate the physical significance of hypertranslations and their connections to new memory effects in follow up work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/5tE3T4oBgHgl3EQfQwkt/content/2301.04415v1.pdf'}
diff --git a/6NE4T4oBgHgl3EQfcAzU/content/tmp_files/2301.05080v1.pdf.txt b/6NE4T4oBgHgl3EQfcAzU/content/tmp_files/2301.05080v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ba282a8d46aa0ec0dc9642bc4682e5ab136ebf5d
--- /dev/null
+++ b/6NE4T4oBgHgl3EQfcAzU/content/tmp_files/2301.05080v1.pdf.txt
@@ -0,0 +1,1161 @@
+Non-linear correlation analysis in financial markets using
+hierarchical clustering
+J. E. Salgado-Hern´andez and Manan Vyas
+Instituto de Ciencias F´ısicas, Universidad Nacional
+Aut´onoma de M´exico, 62210 Cuernavaca, M´exico
+1
+arXiv:2301.05080v1 [q-fin.ST] 12 Jan 2023
+
+Abstract
+Distance correlation coefficient (DCC) can be used to identify new associations and correlations
+between multiple variables. The distance correlation coefficient applies to variables of any dimen-
+sion, can be used to determine smaller sets of variables that provide equivalent information, is zero
+only when variables are independent, and is capable of detecting nonlinear associations that are
+undetectable by the classical Pearson correlation coefficient (PCC). Hence, DCC provides more
+information than the PCC. We analyze numerous pairs of stocks in S&P500 database with the
+distance correlation coefficient and provide an overview of stochastic evolution of financial market
+states based on these correlation measures obtained using agglomerative clustering.
+I.
+INTRODUCTION
+Correlation coefficient is a number which is used to describe dependence between random
+observations. Most popular correlation coefficient is the Pearson one which is defined on the
+interval [−1, 1] [1]. For random variables X and Y , with finite and positive variances, Pearson
+correlation coefficient (PCC) is defined as PCC(X, Y ) = cov(X, Y )/
+�
+var(X) var(Y ). If
+Pearson correlation coefficient between two random variables is zero, it does not necessarily
+mean that the variables are independent. Distance correlation coefficient does not suffer
+from this drawback.
+The distance correlation coefficient (DCC) is a product-moment correlation and a gener-
+alized form of bivariate measures of dependency [2]. It is a very useful and unexplored area
+for statistical inference. The range of the distance correlation is 0 ≤ DCC ≤ 1 [3]. For two
+real random variables X and Y with finite variances, distance correlation coefficient is de-
+fined as DCC(X, Y ) = dcov(X, Y )/
+�
+dcov(X, X) dcov(Y, Y ). Here, the distance covariance
+dcov is defined in the following way. Let (X, Y ), (X′, Y ′) and (X′′, Y ′′) be i.i.d. copies, then
+dcov2(X, Y ) = E(|X − X′||Y − Y ′|) + E(|X − X′|)E(|Y − Y ′|) − 2E(|X − X′||Y − Y ′′|) .
+Thus, DCC is the correlation between the dot products which the ”double centered” (it
+is the operation of converting the distances to the scalar products while placing the origin
+at the geometric center) matrices are comprised of. It is important to mention that the
+definition of distance correlation coefficient can be extended to variables with finite first
+moments only and lack of DCC defines independence.
+2
+
+As both PCC and DCC quantify strength of dependence, is important to understand
+how large the differences between these two measures can possibly be. A natural question is
+how large the DCC can be for variables for which PCC is zero, since uncorrelatedness only
+means the lack of linear dependence. Importantly, nonlinear or nonmonotone dependence
+may exist. The fact that PCC requires finite second moments while DCC requires finite
+first moments implies that PCC is more sensitive to the tails of the distribution. Although
+methods based on ranks (Spearman rank correlation) can be applied in some problems, these
+methods are effective only for testing linear or monotone types of dependence. Importantly,
+uncorrelatedness (PCC = 0) is too weak to imply a central limit theorem which requires
+independence (DCC = 0) necessarily [4–7].
+We have used stocks listed under S&P 500 for the time period August 2000 to August
+2022 and focus on financial market crisis that occurred in the years 2008 (subprime mortgage
+crisis), 2010 (European debt crisis), 2011 (August 2011 stock market fall), 2015 (Great fall
+of China), 2020 (COVID-19 recession) and 2022 (ongoing Russo-Ukrainian war) along with
+bubble periods of 2002 (stock market downturn of 2002) and 2007 (Chinese stock bubble).
+In order to point out the differences of using DCC, we will also focus on epochs for which
+PCC ≈ 0.
+We analyze the Pearson and Distance correlation matrices and their moments along with
+eigenvalue distributions and participation ratios distribution. Participation ratios quantify
+the number of components that participate significantly in each eigenvector [8, 9]. We show
+that there are strong correlations in all these three measures at the times of crisis. Using
+correlation matrices to represent market states [10–12], we employ agglomerative clustering
+[13] to identify correlation matrices that act similarly and compare the clustering results for
+the selected stocks using PCC and DCC.
+II.
+DATA SET
+We use the 5552 daily closing prices of N = 370 stocks listed under S&P 500 for the
+time period August 2000 to August 2022 downloaded freely from Yahoo finance webpage
+[14]. The selected stocks are the ones that have been continuously traded for the chosen
+time period. Using the daily closing prices Pi(t), with index i representing a given stock and
+time t = 1, 2, . . . , T. daily returns ri(t) = [Pi(t) − Pi−1(t)]/Pi−1(t) are calculated. Here T is
+3
+
+Sector
+Ticker Stocks Weight
+Communication Services
+TS
+11
+0.03
+Consumer Discretionary
+CD
+38
+0.10
+Consumer Staples
+CS
+27
+0.07
+Energy
+EN
+18
+0.05
+Financials
+FI
+49
+0.13
+Health Care
+HC
+51
+0.14
+Industrials
+IN
+54
+0.15
+Information Technology
+IT
+49
+0.13
+Materials
+MA
+21
+0.06
+Real Estate
+RE
+25
+0.07
+Utilities
+UT
+27
+0.07
+TABLE I. Distribution of the constituent sectors of selected stocks of financial market S&P 500.
+total number of the trading days present in the considered time horizon. We then have 5551
+daily returns and use these to compute the equal-time cross-correlation matrices based on
+PCC and DCC. Table I gives the distribution of the sectors.
+The disadvantage of working with long financial time series is the loss of information over
+short periods of time, it is convenient to divide it into short time series (epochs). Computing
+returns and dealing with epochs guarantees (weakly) stationary time series. With this, one
+can study the evolution over time, for example, of the average correlations. This helps focus
+on details in a given particular time interval as financial market is a dynamic entity.
+First, we analyze the distribution of correlation matrix elements, eigenvalues and partici-
+pation ratios obtained using both PCC and DCC for all the 138 time epochs (non-overlapping
+epochs of size 40 days each). We show that there are strong correlations in all these three
+measures at the times of crisis. In other words, there is collective motion during crashes.
+III.
+CORRELATIONS AND SPECTRAL ANALYSIS
+To begin with, we plot the correlation matrices obtained using PCC and DCC in Fig.
+1. As expected, one loses the details due to long time averaging. We plot both PCC and
+4
+
+FIG. 1. Correlation matrices for the total time horizon considered. Left panel shows the PCC
+matrix and the right panel shows the DCC matrix. The minimum values for PCC and DCC are
+0.003 and 0.06 respectively. Similarly, the average PCC and DCC are 0.347 and 0.34.
+DCC correlation matrices on the same scale as there are no negative correlations in the PCC
+matrix computed for the total time horizon. Sectorial correlations are stronger for PCC in
+comparison to DCC.
+As one can not see any specific structures in the plots for the correlation matrices for
+the complete time horizon, we study the distribution of correlation matrix elements for each
+epoch as shown in Fig. 2. DCC and PCC both show a clear shift towards higher values
+of correlation during the crisis periods of interest (2002, 2008, 2010, 2011, 2020 and 2022).
+Also, DCC shows the peaks of distributions at lower values of correlation for the non-crisis
+periods, unlike PCC. Notably, DCC ≥ 0.2 for the time horizon considered implying that
+there are non-monotonic correlations present in financial markets at all times.
+Next, we analyze the time evolution of distribution of eigenvalues of correlation matrices
+as shown in Fig. 3; note that the plot is logarithmic. All the correlation matrices are singular
+and thus, we have a delta peak at zero eigenvalues in addition to bulk distribution (which
+follows random matrix theory predictions) and outliers that represent correlations [9, 15–
+19]. The largest eigenvalue, which is linearly correlated with average correlations, attains
+very large values in crisis periods as seen from distribution of eigenvalues for both PCC and
+DCC. Around end of 2016, the gap between the bulk eigenvalue distribution and outliers
+for PCC is little. One can clearly see a comparatively broad bulk eigenvalue distribution
+5
+
+PCC; August-2000 -- August-2022
+TS
+1.0
+CD
+CS
+ 0.8
+EN
+FI
+ 0.6
+HC
+ 0.4
+IN
+IT
+ 0.2
+MA
+RE
+UT
+8
+Z
+5DCC; August-2000 -- August-2022
+TS
+1.0
+CD
+CS
+0.8
+EN
+FI
+一0.6
+HC
+0.4
+IN
+JI
+0.2
+MA
+RE
+UT
+ 0.0
+Z
+5-1
+-0.5
+ 0
+ 0.5
+ 1 2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+ 0
+ 1
+ 2
+ 3
+ 4
+Cij
+P(Cij)
+ 0
+ 1
+ 2
+ 3
+ 4
+Pearson Correlation
+ 0
+ 0.2
+ 0.4
+ 0.6
+ 0.8
+ 1
+2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+ 0
+ 2
+ 4
+ 6
+ 8
+Cij
+P(Cij)
+ 0
+ 2
+ 4
+ 6
+ 8
+Distance Correlation
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+Time (YYYY/MM/DD)
+-1
+-0.5
+ 0
+ 0.5
+ 1
+Cij
+ 0
+ 1
+ 2
+ 3
+ 4
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+Time (YYYY/MM/DD)
+ 0
+ 0.2
+ 0.4
+ 0.6
+ 0.8
+ 1
+Cij
+ 0
+ 2
+ 4
+ 6
+ 8
+FIG. 2. Time evolution of the distribution of correlation matrix elements P(Cij). Left panel shows
+the P(Cij) for PCC and the right panel shows P(Cij) for the DCC. Bottom panel shows the 2D
+projection of the corresponding figures in the top panel.
+for DCC beyond 2019. This feature is also seen in the plot for PCC, however it is equally
+broad for 2001 when the largest eigenvalue is < 100. Like the distribution of correlation
+matrix elements in Fig. 1, eigenvalue distributions for DCC and PCC both show a clear
+shift towards higher values of outliers during the crisis periods of interest (2002, 2008, 2010,
+2011, 2015, 2020 and 2022).
+We use participation ratio (PR) to quantify the number of components that participate
+significantly in each eigenvector νi,
+PRν =
+� N
+�
+i=1
+|νi|4
+�−1
+.
+(1)
+PR gives the number of elements of an eigenvector that are different from zero that contribute
+significantly to the value of the eigenvector and thus, takes values between 1 (only one
+6
+
+2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-172022-08-30
+Time (YYYY/MM/DD)
+ 0
+ 50
+ 100
+ 150
+ 200
+ 250
+ 300
+Eigenvalues
+10-6
+10-5
+10-4
+10-3
+10-2
+2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-172022-08-30
+Time (YYYY/MM/DD)
+ 0
+ 50
+ 100
+ 150
+ 200
+ 250
+ 300
+Eigenvalues
+10-6
+10-5
+10-4
+10-3
+10-2
+FIG. 3. Time evolution of the distribution of eigenvalues. Left panel shows the eigenvalue distri-
+bution for PCC and the right panel shows for the DCC.
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+Time (YYYY/MM/DD)
+ 0
+ 50
+ 100
+ 150
+ 200
+Participation ratios
+ 0
+ 0.005
+ 0.01
+ 0.015
+ 0.02
+ 0.025
+ 0.03
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+Time (YYYY/MM/DD)
+ 0
+ 50
+ 100
+ 150
+ 200
+Participation ratios
+ 0
+ 0.005
+ 0.01
+ 0.015
+ 0.02
+ 0.025
+ 0.03
+ 0.035
+FIG. 4. Time evolution of the distribution of participation ratios. Left panel shows the participation
+ratios distribution for PCC and the right panel shows for the DCC. The horizontal line shows the
+expectation value obtained from random matrix theory.
+component) and N (all components contributing equally). The expectation value of PR for
+a Gaussian Orthogonal Ensemble (classical random matrix ensemble) has the limiting value
+of ⟨PR⟩ ≈ N/3 [8, 20]. We show the time evolution of distribution of PR for PCC and
+DCC in Fig. 4. The horizontal line in the plots gives the average PR value estimated using
+Gaussian Orthogonal Ensemble. As seen from the plots, the average PR for PCC is ≈ 160
+while that for DCC is ≈ 110. The distribution of PR in case of PCC shows a slight upward
+shift during crisis years of 2008, 2010 and 2011 while we see a slight downward shift in case
+of DCC during the crisis years 2002, 2008, 2010, 2011 and 2020. The lesser the average
+correlation, prominent is the downward shift in the distribution of PR in case of DCC.
+Next we analyze the scatter plots between various moments [21] corresponding to PCC
+and DCC and the results are presented in Fig. 5. Note that each point corresponds to an
+7
+
+epoch and we represent the bubble and crisis periods of interest as solid circles. As seen in
+Fig. 1, the crisis periods appear at higher values of mean correlations µ for both PCC and
+DCC. For PCC, the crisis periods of 2008, 2010, 2011, 2015 and 2020 appear with largest µ
+while the bubble periods of 2002 and 2007 alongwith the ongoing Russo-Ukrainian war have
+relatively lower values of µ. Skewness is negative for all the crisis periods and the bubble
+periods implying that the distribution has a longer left tail and bulk is concentrated towards
+the right side. Kurtosis for the crisis periods of 2008, 2010, 2011, 2015 and 2020 is positive
+implying the distributions are leptokurtic while distributions are platykurtic for the bubble
+periods of 2002 and 2007, and the ongoing Russo-Ukrainian war. Emax reflects a similar
+behavior as average correlations µ and PREmax is also maximum for crisis periods of 2008,
+2010, 2011, 2015 and 2020. In summary, PCC distinguishes the bubble periods of 2002 and
+2007, and the ongoing Russo-Ukrainian war from the crisis periods of 2008, 2010, 2011, 2015
+and 2020 depending on kurtosis of the distribution of correlation matrix elements.
+Similarly, in case of DCC: for µ < 0.5, σ increases with increasing µ and for µ > 0.5, σ
+decreases with increasing µ. The crisis periods of 2008, 2010, 2011, 2015 and 2020 appear
+with largest µ while the bubble periods of 2002 and 2007, and the ongoing Russo-Ukrainian
+war have relatively lower values of µ. Skewness is negative for the crisis periods of 2008,
+2010, 2011, 2015 and 2020 implying that the distribution has a longer left tail and bulk
+is concentrated towards the right side, while distribution has a longer right tail for the
+bubble periods of 2002 and 2007, and distribution is symmetric for the ongoing Russo-
+Ukrainian war. Kurtosis for the crisis periods of 2010, 2011 and 2020 is positive implying
+the distributions are leptokurtic while distributions are platykurtic for the bubble periods
+of 2002 and 2007, crisis periods of 2008 and 2015, and the ongoing Russo-Ukrainian war.
+Emax reflects a similar behavior as average correlations µ and PREmax is constant around the
+maximum value for all the epochs. In summary, DCC distinguishes the bubble periods from
+the crisis periods depending on skewness of the distribution of correlation matrix elements.
+Also, DCC distinguishes the bubble periods of 2002 and 2007, the crisis periods of 2008 and
+2015, and the ongoing Russo-Ukrainian war from the crisis periods of 2010, 2011 and 2020
+depending on kurtosis of the distribution of correlation matrix elements.
+8
+
+FIG. 5. Scatter plots corresponding to PCC (top panel) and DCC (bottom panel) between (a)
+mean correlation µ and standard deviation σ, (b) skewness γ1 and σ, (c) excess kurtosis γ2 and σ,
+(d) largest eigenvalue Emax and σ, (e) PR for the largest eigenvalues PREmax and σ, and (f) PR
+for the largest eigenvalue PREmax and largest eigenvalues Emax.
+9
+
+Pearsoncorrelation
+0.8F
+(a)
+2
+(b)
+(c)
+1
+3
+≤ 0.4
+0
+Y2
+0
+-1
+0
+0
+0.1 0.2 0.3 0.4
+00.1 0.2 0.3 0.4
+0
+0.1 0.2 0.3 0.4
+a
+a
+300
+400
+400
+(d)
+(e)
+200
+i300
+300
+xeu
+100
+200
+008
+200
+98
+(f)
+0
+100
+100
+0
+0.1 0.2 0.3 0.4
+0
+0.1 0.2 0.3 0.4
+0
+100
+200
+300
+a
+E
+maxDistance correlation
+0.8
+N
+(a)
+(b)
+(c)
+1
+≤. 0.4
+0
+2
+0
+-1
+0
+0
+0.1
+0.2
+0
+0.1
+0.2
+0
+0.1
+0.2
+300
+400
+400
+200
+300
+300
+max
+E
+100
+200
+200
+(d)
+(e)
+().
+0
+100
+0.1
+0.2
+0.1
+0.2
+100
+0
+100
+200
+300
+a
+a2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+0
+50
+100
+150
+200
+250
+2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+0
+20
+40
+60
+80
+100
+120
+140
+160
+FIG. 6. Euclidean distance matrix obtained using Eq. (2) for PCC (left panel) and DCC (right
+panel).
+IV.
+AGGLOMERATIVE CLUSTERING
+In this section, we compare the clustering results for the selected stocks using PCC and
+DCC. We employ agglomerative clustering that creates clusters by successively merging
+epochs starting with singleton clusters. Using the linkage criterion in each iteration, the
+clusters are joined together until obtaining a single cluster [13].
+Dendrograms give the
+representation of this hierarchy. Choosing the threshold value then decides the number of
+clusters that will be obtained. We cluster similar correlation matrices into these optimized
+n number of “market states”.
+This is a variance-minimizing approach tackled with an
+agglomerative hierarchical approach.
+Dendrograms obtained for the PCC and DCC are
+given in Appendix V.
+In order to implement this algorithm, we need to compute the distance matrix ξ based
+on correlation coefficients C’s,
+ξ(ti, tj) = dE|C(ti) − C(tj)| ,
+(2)
+with dE representing the Euclidean norm and indices i, j = 1, 2, , . . . , 138 representing dif-
+ferent epochs. Figure 6 gives the Euclidean matrices for PCC and DCC respectively. Note
+that the crash periods of 2008, 2010, 2011 and 2020 are visible in these. Once the algorithm
+was trained with its respective distance matrix, the average correlation coefficients PCC and
+DCC were used as inputs to be able to group them into n = 5 clusters that were considered
+10
+
+FIG. 7. Average correlation matrices for each market state obtained using agglomerative clustering
+for PCC [(a)-(e)] and DCC [(f)-(j)]. The average correlation coefficients (from left to right) are
+PCC: 0.12, 0.22, 0.37, 0.52, and 0.65; DCC: 0.35, 0.41, 0.46, 0.54, and 0.66, respectively.
+adequate; see Figs. 10 and 11 for corresponding dendrograms.
+The average correlation matrices of each market states corresponding to both (a) PCC
+and (b) DCC are shown in Fig. 7. The correlation structures vary for each market state
+corresponding to PCC and DCC. The average correlation coefficients (from left to right) are
+(a) PCC: 0.12, 0.22, 0.37, 0.52, and 0.65, (b) DCC: 0.35, 0.41, 0.46, 0.54, and 0.66. The
+number of matrices that are grouped together in each of the market states (from left to right)
+are (a) PCC: 9, 49, 66, 7, and 7 and (b) DCC: 51, 23, 47, 10, and 7. The market states with
+highest correlation coefficient are 7 for both PCC and DCC. For PCC, the market state
+with highest average correlation includes the crash periods of 2008, 2010, 2015 and 2022
+with two matrices not belonging to crash periods. For DCC, the market state with highest
+average correlation includes the crash periods of 2008, 2010, 2011 and 2020. For PCC, the
+market state with second highest average correlation includes epochs in the vicinity of the
+crash periods of 2008, 2010, 2011 and 2020 and for DCC, the market state with second
+highest average correlation includes epochs in the vicinity of the crash periods of 2015 and
+2022. The bubble periods of years 2002 and 2007 are included in the market state with third
+highest average correlation for both PCC and DCC. There are two epochs for which PCC
+11
+
+(a)PCC,mar-(b)PCC,mar-(c)PCC,mar-(
+(d) PCC, mar- (e) PCC, mar-
+ket state1
+ket state 2
+ket state 3
+ket state 4
+ketstate5
+(f) DCC,mar-(g)DCC,mar-(h)DCC,mar-
+(i) DCC, mar-
+G) DCC,
+mar-
+ket state 1
+ket state 2
+ket state 3
+ket state 4
+ket state52000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+1
+2
+3
+4
+5
+States
+2000-10-02
+2003-11-10
+2007-02-05
+2010-04-09
+2013-06-13
+2016-10-12
+2019-12-17
+2022-08-30
+1
+2
+3
+4
+5
+States
+FIG. 8. Dynamical evolution of financial market in time: PCC (top panel) and DCC (bottom
+panel).
+The market states 1, 2, . . . , 5 obtained using agglomerative clustering are arranged in
+increasing order of average correlation coefficients for both PCC and DCC.
+≈ 0 and these epochs are in the market state corresponding to the lowest average correlation
+coefficient both for PCC and DCC. Note that this market state has respectively 9 and 51
+matrices in the cluster for PCC and DCC.
+Dynamical evolution of the financial market can be studied by the transitions between
+these market states. The financial market can remain in a particular market state, can
+jump to another market state and bounce back or evolve to another market state. In Fig.
+8 the results of the temporal evolution of the market are shown based on both PCC and
+DCC and Fig. 9 shows the corresponding transition matrices. For each market state, the
+average correlation coefficients are ordered in ascending order. Transitions are counted when
+changing epoch, either from one market state to another or if it remained in the same market
+state. Most of the values stay close to the diagonal, this means that the transitions occur in
+12
+
+1
+2
+3
+4
+5
+1
+2
+3
+4
+5
+3
+5
+1
+0
+0
+4
+26
+16
+1
+2
+1
+16
+42
+4
+2
+0
+1
+5
+1
+0
+0
+1
+2
+1
+3
+1
+2
+3
+4
+5
+1
+2
+3
+4
+5
+31
+10
+7
+1
+2
+9
+5
+8
+1
+0
+8
+8
+26
+3
+1
+1
+0
+6
+2
+1
+1
+0
+0
+3
+3
+FIG. 9. Transition matrices corresponding to PCC (left panel) and DCC (right panel) showing
+transition between the five market states obtained using agglomerative clustering.
+small jumps towards the closest market states or continue in itself and transitions between
+states with low average correlation and high average correlations are avoided [12, 22].
+In case of PCC, the state with lowest average correlation (1) never connects to state with
+highest (5) or second highest (4) average correlation coefficient. There is a transition from
+state state 2 to 5 and state 2 to 5 which are indirect transitions as they are in the sequence
+1 → 2 → 5 and 2 → 3 → 5 and these correspond to the crash periods of 2020 and 2011
+respectively. Similarly, for DCC, the state with second lowest average correlation (2) never
+connects to state with highest (5) average correlation coefficient. However, there are two
+transitions between 5 and 1 and one transition from 1 to 5. These correspond to the crash
+periods of 2010, 2020 and 2011 respectively. There is also a transition between 1 and 4 that
+corresponds to the crash period of 2011. This is an indirect one as first transition happens
+between 1 and 3 and then to 4.
+V.
+CONCLUSIONS
+We analyzed correlations in S&P 500 market data for the time period August 2000 to
+August 2022 using both PCC and DCC. Notably, DCC ≥ 0.2 for the time horizon considered
+implying that there are non-monotonic correlations present in financial markets at all times.
+Eigenvalue distributions for DCC and PCC both show a clear shift towards higher values of
+13
+
+outliers during the crisis periods of interest (2002, 2008, 2010, 2011, 2015, 2020 and 2022).
+The distribution of PR in case of PCC shows a slight upward shift during crisis years of
+2008, 2010 and 2011 while we see a slight downward shift in case of DCC during the bubble
+period of 2002 and crisis years 2008, 2010, 2011 and 2020. The lesser the average correlation,
+prominent is the downward shift in the distribution of PR in case of DCC.
+PCC distinguishes the bubble periods of 2002 and 2007, and the ongoing Russo-Ukrainian
+war from the crisis periods of 2008, 2010, 2011, 2015 and 2020 depending on kurtosis of the
+distribution of correlation matrix elements.
+DCC distinguishes the bubble periods from
+the crisis periods depending on skewness of the distribution of correlation matrix elements.
+Also, DCC distinguishes the bubble periods of 2002 and 2007, the crisis periods of 2008 and
+2015, and the ongoing Russo-Ukrainian war from the crisis periods of 2010, 2011 and 2020
+depending on kurtosis of the distribution of correlation matrix elements.
+Going further, we compare the clustering results for correlation matrices obtained for
+the selected stocks using PCC and DCC. We employ agglomerative clustering that uses
+Euclidean distances and minimizes the sum of squared differences within all clusters. We
+obtain five market states corresponding to both PCC and DCC. The crisis periods are in
+market states with largest and second largest average correlation coefficients. Bubble periods
+are in the market state with third largest average correlation coefficient. The two epochs for
+PCC ≈ 0 are in the market state with smallest average correlation coefficient; note that this
+market state has respectively 9 and 51 matrices in the cluster for PCC and DCC. We also
+compare the transitions between these market states for both PCC and DCC. In summary,
+results for clustering depend upon the linear (PCC) and non-linear (DCC) nature of the
+correlation coefficient employed. Preliminary results on financial markets can be viewed in
+a bachelor thesis [23].
+ACKNOWLEDGMENTS
+Authors thank Harinder Pal for many useful discussions on clustering algorithms and
+help with many figures. Authors acknowledge financial support from CONACYT project
+14
+
+Fronteras 10872.
+[1] R. N. Mantegna and H. E. Stanley, Introduction to Econophysics: Correlations and Complex-
+ity in Finance, (Cambridge University Press, 1999).
+[2] G. J. Sz´ekely and M. L. Rizzo, The Annals of Applied Statistics, 3, 1236 (2009).
+[3] D. Edelmann, T. F. M´ori, G. J. Sz´ekely, Statistics and Probability Letters 169, 108960 (2021).
+[4] R. C. Bradley, J. Multivariate Anal. 11, 1 (1981).
+[5] R. C. Bradley, Ann. Probab. 16, 313 (1988).
+[6] R. C. Bradley, Introduction to Strong Mixing Condition, 1–3, (Kendrick Press, , Heber City
+(Utah), 2007).
+[7] G. J. Sz´ekely and N. K. Bakirov, Brownian covariance and CLT for stationary sequences,
+Technical Report No. 08-01, Dept. Mathematics and Statistics, Bowling Green State Univ.,
+Bowling Green, OH (2008).
+[8] T. Guhr, A. Mueller, H. A. Weidenmueller, Physics Reports 299, 189 (1998).
+[9] V. Plerou, P. Gopikrishnan, B. Rosenow, L. A. Nunes Amaral, and H. E. Stanley, Phys. Rev.
+Lett. 83, 1471 (1999).
+[10] M. C. M¨unnix et. al. , Scientific Reports 2, 644 (2012).
+[11] D. Chetalova, R Sch¨afer, and T. Guhr, J. Stat. Mech. 2015, P01029 (2015).
+[12] H. K. Pharasi et. al. , New Journal of Physics 20, 103041 (2018).
+[13] N. Musmeci, T. Aste, and T. Di Matteo, PLoS ONE 10, 1 (2015).
+[14] Yahoo finance database, https://finance.yahoo.com/, accessed on 10 October, 2022 for S&P
+500.
+[15] A. Edelman, SIAM Journal on Matrix Analysis and Applications, 9, 543 (1988).
+[16] L. Laloux, P. Cizeau, J. -P. Bouchaud, and M. Potters, Phys. Rev. Lett. 83, 1467 (1999).
+[17] M. Vyas, T. Guhr, and T. H. Seligman, Scientific reports 8, 1 (2018).
+[18] P. Bhadola and N. Deo, ”Spectral and Network Method in Financial Time Series Analysis:
+A Study on Stock and Currency Market”, in A. S. Chakrabarti et al. (eds.), Network Theory
+and Agent-Based Modeling in Economics and Finance (2019) pp. 331-352.
+[19] H. K. Pharasi, K. Sharma, A. Chakraborti, and T. H. Seligman, ”Complex market dynamics
+in the light of random matrix theory”, in New Perspectives and Challenges in Econophysics
+15
+
+and Sociophysics, edited by F. Abergel, B. K. Chakrabarti, A. Chakraborti, N. Deo, and K.
+Sharma (Springer International Publishing, Cham, 2019) pp. 13–34.
+[20] V. K. B. Kota, Embedded Random Matrix Ensembles in Quantum Physics (Springer, Heidel-
+berg, 2014).
+[21] A. Stuart and J. K. Ord, Kendall’s Advanced Theory of Statistics : Distribution Theory
+(Oxford University Press, New York, 1987).
+[22] A. J. Heckens and T. Guhr, J. Stat. Mech. 2022, 043401 (2022)
+[23] J. E. Salgado-Hern´andez, (Licenciatura thesis, UNAM) Correlaci´on y agrupaciones de series
+de tiempo financieras (2023).
+16
+
+APPENDIX A: DENDROGRAMS OBTAINED USING PCC AND DCC
+114
+73
+136
+93
+62
+53
+61
+60
+70
+69
+50
+51
+68
+122
+22
+44
+28
+35
+75
+82
+87
+17
+24
+31
+36
+54
+55
+56
+63
+48
+46
+47
+13
+15
+16
+11
+14
+74
+76
+77
+79
+80
+81
+84
+90
+91
+49
+134
+130
+125
+133
+126
+99
+123
+83
+92
+95
+58
+65
+72
+94
+100
+135
+137
+110
+118
+12
+88
+89
+57
+40
+45
+42
+43
+67
+52
+78
+124
+96
+109
+120
+0
+107
+1
+2
+101
+112
+105
+108
+5
+7
+8
+3
+39
+33
+38
+127
+131
+121
+106
+103
+104
+111
+102
+115
+113
+117
+85
+97
+132
+128
+129
+6
+119
+98
+116
+59
+64
+66
+34
+26
+29
+19
+20
+23
+27
+25
+18
+21
+32
+30
+41
+71
+86
+37
+4
+9
+10
+Epochs
+0
+500
+1000
+1500
+2000
+2500
+3000
+Euclidean distance
+Dendrogram (PCC)
+FIG. 10. Dendrogram obtained for PCC using agglomerative clustering.
+17
+
+50
+69
+70
+51
+60
+68
+122
+97
+98
+75
+87
+31
+6
+17
+24
+35
+85
+119
+113
+117
+76
+84
+80
+82
+36
+11
+22
+77
+28
+44
+0
+1
+107
+3
+8
+2
+5
+120
+108
+105
+112
+7
+9
+37
+39
+21
+33
+38
+71
+115
+102
+106
+101
+86
+103
+104
+30
+32
+29
+25
+26
+18
+34
+41
+4
+10
+19
+20
+23
+131
+128
+129
+127
+132
+116
+111
+121
+27
+64
+59
+66
+93
+53
+61
+123
+114
+136
+62
+73
+52
+67
+99
+110
+118
+88
+109
+12
+65
+72
+89
+94
+95
+45
+42
+43
+96
+57
+78
+49
+47
+48
+124
+126
+134
+135
+137
+100
+90
+92
+81
+83
+74
+79
+56
+63
+58
+54
+55
+91
+40
+46
+130
+125
+133
+13
+15
+14
+16
+Epochs
+0
+500
+1000
+1500
+2000
+Euclidean Distance
+Dendrogram (DCC)
+FIG. 11. Dendrogram obtained for DCC using agglomerative clustering.
+18
+
diff --git a/6NE4T4oBgHgl3EQfcAzU/content/tmp_files/load_file.txt b/6NE4T4oBgHgl3EQfcAzU/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..248acc6cd24922e7ef38f70a3b79bbb139c7a417
--- /dev/null
+++ b/6NE4T4oBgHgl3EQfcAzU/content/tmp_files/load_file.txt
@@ -0,0 +1,774 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf,len=773
+page_content='Non-linear correlation analysis in financial markets using hierarchical clustering J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Salgado-Hern´andez and Manan Vyas Instituto de Ciencias F´ısicas, Universidad Nacional Aut´onoma de M´exico, 62210 Cuernavaca, M´exico 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='05080v1 [q-fin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='ST] 12 Jan 2023 Abstract Distance correlation coefficient (DCC) can be used to identify new associations and correlations between multiple variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The distance correlation coefficient applies to variables of any dimen- sion, can be used to determine smaller sets of variables that provide equivalent information, is zero only when variables are independent, and is capable of detecting nonlinear associations that are undetectable by the classical Pearson correlation coefficient (PCC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Hence, DCC provides more information than the PCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We analyze numerous pairs of stocks in S&P500 database with the distance correlation coefficient and provide an overview of stochastic evolution of financial market states based on these correlation measures obtained using agglomerative clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' INTRODUCTION Correlation coefficient is a number which is used to describe dependence between random observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Most popular correlation coefficient is the Pearson one which is defined on the interval [−1, 1] [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For random variables X and Y , with finite and positive variances, Pearson correlation coefficient (PCC) is defined as PCC(X, Y ) = cov(X, Y )/ � var(X) var(Y ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' If Pearson correlation coefficient between two random variables is zero, it does not necessarily mean that the variables are independent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Distance correlation coefficient does not suffer from this drawback.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The distance correlation coefficient (DCC) is a product-moment correlation and a gener- alized form of bivariate measures of dependency [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' It is a very useful and unexplored area for statistical inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The range of the distance correlation is 0 ≤ DCC ≤ 1 [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For two real random variables X and Y with finite variances, distance correlation coefficient is de- fined as DCC(X, Y ) = dcov(X, Y )/ � dcov(X, X) dcov(Y, Y ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Here, the distance covariance dcov is defined in the following way.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Let (X, Y ), (X′, Y ′) and (X′′, Y ′′) be i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' copies, then dcov2(X, Y ) = E(|X − X′||Y − Y ′|) + E(|X − X′|)E(|Y − Y ′|) − 2E(|X − X′||Y − Y ′′|) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Thus, DCC is the correlation between the dot products which the ”double centered” (it is the operation of converting the distances to the scalar products while placing the origin at the geometric center) matrices are comprised of.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' It is important to mention that the definition of distance correlation coefficient can be extended to variables with finite first moments only and lack of DCC defines independence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 2 As both PCC and DCC quantify strength of dependence, is important to understand how large the differences between these two measures can possibly be.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' A natural question is how large the DCC can be for variables for which PCC is zero, since uncorrelatedness only means the lack of linear dependence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Importantly, nonlinear or nonmonotone dependence may exist.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The fact that PCC requires finite second moments while DCC requires finite first moments implies that PCC is more sensitive to the tails of the distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Although methods based on ranks (Spearman rank correlation) can be applied in some problems, these methods are effective only for testing linear or monotone types of dependence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Importantly, uncorrelatedness (PCC = 0) is too weak to imply a central limit theorem which requires independence (DCC = 0) necessarily [4–7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We have used stocks listed under S&P 500 for the time period August 2000 to August 2022 and focus on financial market crisis that occurred in the years 2008 (subprime mortgage crisis), 2010 (European debt crisis), 2011 (August 2011 stock market fall), 2015 (Great fall of China), 2020 (COVID-19 recession) and 2022 (ongoing Russo-Ukrainian war) along with bubble periods of 2002 (stock market downturn of 2002) and 2007 (Chinese stock bubble).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In order to point out the differences of using DCC, we will also focus on epochs for which PCC ≈ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We analyze the Pearson and Distance correlation matrices and their moments along with eigenvalue distributions and participation ratios distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Participation ratios quantify the number of components that participate significantly in each eigenvector [8, 9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We show that there are strong correlations in all these three measures at the times of crisis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Using correlation matrices to represent market states [10–12], we employ agglomerative clustering [13] to identify correlation matrices that act similarly and compare the clustering results for the selected stocks using PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' DATA SET We use the 5552 daily closing prices of N = 370 stocks listed under S&P 500 for the time period August 2000 to August 2022 downloaded freely from Yahoo finance webpage [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The selected stocks are the ones that have been continuously traded for the chosen time period.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Using the daily closing prices Pi(t), with index i representing a given stock and time t = 1, 2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' , T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' daily returns ri(t) = [Pi(t) − Pi−1(t)]/Pi−1(t) are calculated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Here T is 3 Sector Ticker Stocks Weight Communication Services TS 11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='03 Consumer Discretionary CD 38 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='10 Consumer Staples CS 27 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='07 Energy EN 18 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='05 Financials FI 49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='13 Health Care HC 51 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='14 Industrials IN 54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='15 Information Technology IT 49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='13 Materials MA 21 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='06 Real Estate RE 25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='07 Utilities UT 27 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='07 TABLE I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Distribution of the constituent sectors of selected stocks of financial market S&P 500.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' total number of the trading days present in the considered time horizon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We then have 5551 daily returns and use these to compute the equal-time cross-correlation matrices based on PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Table I gives the distribution of the sectors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The disadvantage of working with long financial time series is the loss of information over short periods of time, it is convenient to divide it into short time series (epochs).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Computing returns and dealing with epochs guarantees (weakly) stationary time series.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' With this, one can study the evolution over time, for example, of the average correlations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' This helps focus on details in a given particular time interval as financial market is a dynamic entity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' First, we analyze the distribution of correlation matrix elements, eigenvalues and partici- pation ratios obtained using both PCC and DCC for all the 138 time epochs (non-overlapping epochs of size 40 days each).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We show that there are strong correlations in all these three measures at the times of crisis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In other words, there is collective motion during crashes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' CORRELATIONS AND SPECTRAL ANALYSIS To begin with, we plot the correlation matrices obtained using PCC and DCC in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' As expected, one loses the details due to long time averaging.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We plot both PCC and 4 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Correlation matrices for the total time horizon considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Left panel shows the PCC matrix and the right panel shows the DCC matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The minimum values for PCC and DCC are 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='003 and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='06 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Similarly, the average PCC and DCC are 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='347 and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' DCC correlation matrices on the same scale as there are no negative correlations in the PCC matrix computed for the total time horizon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Sectorial correlations are stronger for PCC in comparison to DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' As one can not see any specific structures in the plots for the correlation matrices for the complete time horizon, we study the distribution of correlation matrix elements for each epoch as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' DCC and PCC both show a clear shift towards higher values of correlation during the crisis periods of interest (2002, 2008, 2010, 2011, 2020 and 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Also, DCC shows the peaks of distributions at lower values of correlation for the non-crisis periods, unlike PCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Notably, DCC ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 for the time horizon considered implying that there are non-monotonic correlations present in financial markets at all times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Next, we analyze the time evolution of distribution of eigenvalues of correlation matrices as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 3;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' note that the plot is logarithmic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' All the correlation matrices are singular and thus, we have a delta peak at zero eigenvalues in addition to bulk distribution (which follows random matrix theory predictions) and outliers that represent correlations [9, 15– 19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The largest eigenvalue, which is linearly correlated with average correlations, attains very large values in crisis periods as seen from distribution of eigenvalues for both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Around end of 2016, the gap between the bulk eigenvalue distribution and outliers for PCC is little.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' One can clearly see a comparatively broad bulk eigenvalue distribution 5 PCC;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' August-2000 -- August-2022 TS 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 CD CS 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 EN FI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='6 HC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 IN IT 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 MA RE UT 8 Z 5DCC;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' August-2000 -- August-2022 TS 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 CD CS 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 EN FI 一0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='6 HC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 IN JI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 MA RE UT 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 Z 5-1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5 1 2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 0 1 2 3 4 Cij P(Cij) 0 1 2 3 4 Pearson Correlation 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 1 2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 0 2 4 6 8 Cij P(Cij) 0 2 4 6 8 Distance Correlation 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 Time (YYYY/MM/DD) 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5 1 Cij 0 1 2 3 4 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 Time (YYYY/MM/DD) 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 1 Cij 0 2 4 6 8 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Time evolution of the distribution of correlation matrix elements P(Cij).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Left panel shows the P(Cij) for PCC and the right panel shows P(Cij) for the DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bottom panel shows the 2D projection of the corresponding figures in the top panel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' for DCC beyond 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' This feature is also seen in the plot for PCC, however it is equally broad for 2001 when the largest eigenvalue is < 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Like the distribution of correlation matrix elements in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 1, eigenvalue distributions for DCC and PCC both show a clear shift towards higher values of outliers during the crisis periods of interest (2002, 2008, 2010, 2011, 2015, 2020 and 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We use participation ratio (PR) to quantify the number of components that participate significantly in each eigenvector νi, PRν = � N � i=1 |νi|4 �−1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' (1) PR gives the number of elements of an eigenvector that are different from zero that contribute significantly to the value of the eigenvector and thus,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' takes values between 1 (only one 6 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-172022-08-30 Time (YYYY/MM/DD) 0 50 100 150 200 250 300 Eigenvalues 10-6 10-5 10-4 10-3 10-2 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-172022-08-30 Time (YYYY/MM/DD) 0 50 100 150 200 250 300 Eigenvalues 10-6 10-5 10-4 10-3 10-2 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Time evolution of the distribution of eigenvalues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Left panel shows the eigenvalue distri- bution for PCC and the right panel shows for the DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 Time (YYYY/MM/DD) 0 50 100 150 200 Participation ratios 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='005 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='015 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='025 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='03 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 Time (YYYY/MM/DD) 0 50 100 150 200 Participation ratios 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='005 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='015 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='025 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='035 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Time evolution of the distribution of participation ratios.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Left panel shows the participation ratios distribution for PCC and the right panel shows for the DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The horizontal line shows the expectation value obtained from random matrix theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' component) and N (all components contributing equally).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The expectation value of PR for a Gaussian Orthogonal Ensemble (classical random matrix ensemble) has the limiting value of ⟨PR⟩ ≈ N/3 [8, 20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We show the time evolution of distribution of PR for PCC and DCC in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The horizontal line in the plots gives the average PR value estimated using Gaussian Orthogonal Ensemble.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' As seen from the plots, the average PR for PCC is ≈ 160 while that for DCC is ≈ 110.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The distribution of PR in case of PCC shows a slight upward shift during crisis years of 2008, 2010 and 2011 while we see a slight downward shift in case of DCC during the crisis years 2002, 2008, 2010, 2011 and 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The lesser the average correlation, prominent is the downward shift in the distribution of PR in case of DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Next we analyze the scatter plots between various moments [21] corresponding to PCC and DCC and the results are presented in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Note that each point corresponds to an 7 epoch and we represent the bubble and crisis periods of interest as solid circles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' As seen in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 1, the crisis periods appear at higher values of mean correlations µ for both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For PCC, the crisis periods of 2008, 2010, 2011, 2015 and 2020 appear with largest µ while the bubble periods of 2002 and 2007 alongwith the ongoing Russo-Ukrainian war have relatively lower values of µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Skewness is negative for all the crisis periods and the bubble periods implying that the distribution has a longer left tail and bulk is concentrated towards the right side.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Kurtosis for the crisis periods of 2008, 2010, 2011, 2015 and 2020 is positive implying the distributions are leptokurtic while distributions are platykurtic for the bubble periods of 2002 and 2007, and the ongoing Russo-Ukrainian war.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Emax reflects a similar behavior as average correlations µ and PREmax is also maximum for crisis periods of 2008, 2010, 2011, 2015 and 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In summary, PCC distinguishes the bubble periods of 2002 and 2007, and the ongoing Russo-Ukrainian war from the crisis periods of 2008, 2010, 2011, 2015 and 2020 depending on kurtosis of the distribution of correlation matrix elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Similarly, in case of DCC: for µ < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5, σ increases with increasing µ and for µ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5, σ decreases with increasing µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The crisis periods of 2008, 2010, 2011, 2015 and 2020 appear with largest µ while the bubble periods of 2002 and 2007, and the ongoing Russo-Ukrainian war have relatively lower values of µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Skewness is negative for the crisis periods of 2008, 2010, 2011, 2015 and 2020 implying that the distribution has a longer left tail and bulk is concentrated towards the right side, while distribution has a longer right tail for the bubble periods of 2002 and 2007, and distribution is symmetric for the ongoing Russo- Ukrainian war.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Kurtosis for the crisis periods of 2010, 2011 and 2020 is positive implying the distributions are leptokurtic while distributions are platykurtic for the bubble periods of 2002 and 2007, crisis periods of 2008 and 2015, and the ongoing Russo-Ukrainian war.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Emax reflects a similar behavior as average correlations µ and PREmax is constant around the maximum value for all the epochs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In summary, DCC distinguishes the bubble periods from the crisis periods depending on skewness of the distribution of correlation matrix elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Also, DCC distinguishes the bubble periods of 2002 and 2007, the crisis periods of 2008 and 2015, and the ongoing Russo-Ukrainian war from the crisis periods of 2010, 2011 and 2020 depending on kurtosis of the distribution of correlation matrix elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 8 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Scatter plots corresponding to PCC (top panel) and DCC (bottom panel) between (a) mean correlation µ and standard deviation σ, (b) skewness γ1 and σ, (c) excess kurtosis γ2 and σ, (d) largest eigenvalue Emax and σ, (e) PR for the largest eigenvalues PREmax and σ, and (f) PR for the largest eigenvalue PREmax and largest eigenvalues Emax.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 9 Pearsoncorrelation 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8F (a) 2 (b) (c) 1 3 ≤ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0 Y2 0 1 0 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 00.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 a a 300 400 400 (d) (e) 200 i300 300 xeu 100 200 008 200 98 (f) 0 100 100 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0 100 200 300 a E maxDistance correlation 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 N (a) (b) (c) 1 ≤.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 0 2 0 1 0 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 300 400 400 200 300 300 max E 100 200 200 (d) (e) ().' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 0 100 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 100 0 100 200 300 a a2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 0 50 100 150 200 250 2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 0 20 40 60 80 100 120 140 160 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Euclidean distance matrix obtained using Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' (2) for PCC (left panel) and DCC (right panel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' AGGLOMERATIVE CLUSTERING In this section, we compare the clustering results for the selected stocks using PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We employ agglomerative clustering that creates clusters by successively merging epochs starting with singleton clusters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Using the linkage criterion in each iteration, the clusters are joined together until obtaining a single cluster [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Dendrograms give the representation of this hierarchy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Choosing the threshold value then decides the number of clusters that will be obtained.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We cluster similar correlation matrices into these optimized n number of “market states”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' This is a variance-minimizing approach tackled with an agglomerative hierarchical approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Dendrograms obtained for the PCC and DCC are given in Appendix V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In order to implement this algorithm, we need to compute the distance matrix ξ based on correlation coefficients C’s, ξ(ti, tj) = dE|C(ti) − C(tj)| , (2) with dE representing the Euclidean norm and indices i, j = 1, 2, , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' , 138 representing dif- ferent epochs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Figure 6 gives the Euclidean matrices for PCC and DCC respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Note that the crash periods of 2008, 2010, 2011 and 2020 are visible in these.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Once the algorithm was trained with its respective distance matrix, the average correlation coefficients PCC and DCC were used as inputs to be able to group them into n = 5 clusters that were considered 10 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Average correlation matrices for each market state obtained using agglomerative clustering for PCC [(a)-(e)] and DCC [(f)-(j)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The average correlation coefficients (from left to right) are PCC: 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='12, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='22, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='37, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='52, and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='65;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' DCC: 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='35, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='41, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='46, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='54, and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='66, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' adequate;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' see Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 10 and 11 for corresponding dendrograms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The average correlation matrices of each market states corresponding to both (a) PCC and (b) DCC are shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The correlation structures vary for each market state corresponding to PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The average correlation coefficients (from left to right) are (a) PCC: 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='12, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='22, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='37, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='52, and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='65, (b) DCC: 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='35, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='41, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='46, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='54, and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The number of matrices that are grouped together in each of the market states (from left to right) are (a) PCC: 9, 49, 66, 7, and 7 and (b) DCC: 51, 23, 47, 10, and 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The market states with highest correlation coefficient are 7 for both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For PCC, the market state with highest average correlation includes the crash periods of 2008, 2010, 2015 and 2022 with two matrices not belonging to crash periods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For DCC, the market state with highest average correlation includes the crash periods of 2008, 2010, 2011 and 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For PCC, the market state with second highest average correlation includes epochs in the vicinity of the crash periods of 2008, 2010, 2011 and 2020 and for DCC, the market state with second highest average correlation includes epochs in the vicinity of the crash periods of 2015 and 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The bubble periods of years 2002 and 2007 are included in the market state with third highest average correlation for both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' There are two epochs for which PCC 11 (a)PCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='mar-(b)PCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='mar-(c)PCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='mar-( (d) PCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' mar- (e) PCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' mar- ket state1 ket state 2 ket state 3 ket state 4 ketstate5 (f) DCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='mar-(g)DCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='mar-(h)DCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='mar- (i) DCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' mar- G) DCC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' mar- ket state 1 ket state 2 ket state 3 ket state 4 ket state52000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 1 2 3 4 5 States 2000-10-02 2003-11-10 2007-02-05 2010-04-09 2013-06-13 2016-10-12 2019-12-17 2022-08-30 1 2 3 4 5 States FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Dynamical evolution of financial market in time: PCC (top panel) and DCC (bottom panel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The market states 1, 2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' , 5 obtained using agglomerative clustering are arranged in increasing order of average correlation coefficients for both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' ≈ 0 and these epochs are in the market state corresponding to the lowest average correlation coefficient both for PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Note that this market state has respectively 9 and 51 matrices in the cluster for PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Dynamical evolution of the financial market can be studied by the transitions between these market states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The financial market can remain in a particular market state, can jump to another market state and bounce back or evolve to another market state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 8 the results of the temporal evolution of the market are shown based on both PCC and DCC and Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 9 shows the corresponding transition matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' For each market state, the average correlation coefficients are ordered in ascending order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Transitions are counted when changing epoch, either from one market state to another or if it remained in the same market state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Most of the values stay close to the diagonal, this means that the transitions occur in 12 1 2 3 4 5 1 2 3 4 5 3 5 1 0 0 4 26 16 1 2 1 16 42 4 2 0 1 5 1 0 0 1 2 1 3 1 2 3 4 5 1 2 3 4 5 31 10 7 1 2 9 5 8 1 0 8 8 26 3 1 1 0 6 2 1 1 0 0 3 3 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Transition matrices corresponding to PCC (left panel) and DCC (right panel) showing transition between the five market states obtained using agglomerative clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' small jumps towards the closest market states or continue in itself and transitions between states with low average correlation and high average correlations are avoided [12, 22].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In case of PCC, the state with lowest average correlation (1) never connects to state with highest (5) or second highest (4) average correlation coefficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' There is a transition from state state 2 to 5 and state 2 to 5 which are indirect transitions as they are in the sequence 1 → 2 → 5 and 2 → 3 → 5 and these correspond to the crash periods of 2020 and 2011 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Similarly, for DCC, the state with second lowest average correlation (2) never connects to state with highest (5) average correlation coefficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' However, there are two transitions between 5 and 1 and one transition from 1 to 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' These correspond to the crash periods of 2010, 2020 and 2011 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' There is also a transition between 1 and 4 that corresponds to the crash period of 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' This is an indirect one as first transition happens between 1 and 3 and then to 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' CONCLUSIONS We analyzed correlations in S&P 500 market data for the time period August 2000 to August 2022 using both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Notably, DCC ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 for the time horizon considered implying that there are non-monotonic correlations present in financial markets at all times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Eigenvalue distributions for DCC and PCC both show a clear shift towards higher values of 13 outliers during the crisis periods of interest (2002, 2008, 2010, 2011, 2015, 2020 and 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The distribution of PR in case of PCC shows a slight upward shift during crisis years of 2008, 2010 and 2011 while we see a slight downward shift in case of DCC during the bubble period of 2002 and crisis years 2008, 2010, 2011 and 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The lesser the average correlation, prominent is the downward shift in the distribution of PR in case of DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' PCC distinguishes the bubble periods of 2002 and 2007, and the ongoing Russo-Ukrainian war from the crisis periods of 2008, 2010, 2011, 2015 and 2020 depending on kurtosis of the distribution of correlation matrix elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' DCC distinguishes the bubble periods from the crisis periods depending on skewness of the distribution of correlation matrix elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Also, DCC distinguishes the bubble periods of 2002 and 2007, the crisis periods of 2008 and 2015, and the ongoing Russo-Ukrainian war from the crisis periods of 2010, 2011 and 2020 depending on kurtosis of the distribution of correlation matrix elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Going further, we compare the clustering results for correlation matrices obtained for the selected stocks using PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We employ agglomerative clustering that uses Euclidean distances and minimizes the sum of squared differences within all clusters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We obtain five market states corresponding to both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The crisis periods are in market states with largest and second largest average correlation coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bubble periods are in the market state with third largest average correlation coefficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' The two epochs for PCC ≈ 0 are in the market state with smallest average correlation coefficient;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' note that this market state has respectively 9 and 51 matrices in the cluster for PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' We also compare the transitions between these market states for both PCC and DCC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' In summary, results for clustering depend upon the linear (PCC) and non-linear (DCC) nature of the correlation coefficient employed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Preliminary results on financial markets can be viewed in a bachelor thesis [23].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' ACKNOWLEDGMENTS Authors thank Harinder Pal for many useful discussions on clustering algorithms and help with many figures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Authors acknowledge financial support from CONACYT project 14 Fronteras 10872.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [1] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Mantegna and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Stanley, Introduction to Econophysics: Correlations and Complex- ity in Finance, (Cambridge University Press, 1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [2] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Sz´ekely and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Rizzo, The Annals of Applied Statistics, 3, 1236 (2009).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [3] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Edelmann, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' M´ori, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Sz´ekely, Statistics and Probability Letters 169, 108960 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [4] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bradley, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Multivariate Anal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 11, 1 (1981).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [5] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bradley, Ann.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Probab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 16, 313 (1988).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [6] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bradley, Introduction to Strong Mixing Condition, 1–3, (Kendrick Press, , Heber City (Utah), 2007).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [7] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Sz´ekely and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bakirov, Brownian covariance and CLT for stationary sequences, Technical Report No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 08-01, Dept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Mathematics and Statistics, Bowling Green State Univ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=', Bowling Green, OH (2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [8] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Guhr, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Mueller, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Weidenmueller, Physics Reports 299, 189 (1998).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [9] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Plerou, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Gopikrishnan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Rosenow, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Nunes Amaral, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Stanley, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 83, 1471 (1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [10] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' M¨unnix et.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' , Scientific Reports 2, 644 (2012).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [11] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Chetalova, R Sch¨afer, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Guhr, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Stat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Mech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 2015, P01029 (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [12] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Pharasi et.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' , New Journal of Physics 20, 103041 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [13] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Musmeci, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Aste, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Di Matteo, PLoS ONE 10, 1 (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [14] Yahoo finance database, https://finance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='yahoo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='com/, accessed on 10 October, 2022 for S&P 500.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [15] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Edelman, SIAM Journal on Matrix Analysis and Applications, 9, 543 (1988).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [16] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Laloux, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Cizeau, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' -P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bouchaud, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Potters, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 83, 1467 (1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [17] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Vyas, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Guhr, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Seligman, Scientific reports 8, 1 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [18] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Bhadola and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Deo, ”Spectral and Network Method in Financial Time Series Analysis: A Study on Stock and Currency Market”, in A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Chakrabarti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' ), Network Theory and Agent-Based Modeling in Economics and Finance (2019) pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 331-352.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [19] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Pharasi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Sharma, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Chakraborti, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Seligman, ”Complex market dynamics in the light of random matrix theory”, in New Perspectives and Challenges in Econophysics 15 and Sociophysics, edited by F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Abergel, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Chakrabarti, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Chakraborti, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Deo, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Sharma (Springer International Publishing, Cham, 2019) pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 13–34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [20] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Kota, Embedded Random Matrix Ensembles in Quantum Physics (Springer, Heidel- berg, 2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [21] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Stuart and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Ord, Kendall’s Advanced Theory of Statistics : Distribution Theory (Oxford University Press, New York, 1987).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' [22] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Heckens and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Guhr, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Stat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Mech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 2022, 043401 (2022) [23] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Salgado-Hern´andez, (Licenciatura thesis, UNAM) Correlaci´on y agrupaciones de series de tiempo financieras (2023).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='APPENDIX A: DENDROGRAMS OBTAINED USING PCC AND DCC ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='114 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='73 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='136 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='93 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='62 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='53 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='61 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='70 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='69 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='51 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='68 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='122 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='44 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='28 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='75 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='82 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='87 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='31 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='36 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='54 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='55 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='56 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='63 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='48 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='46 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='47 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='74 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='76 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='77 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='79 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='80 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='81 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='84 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='90 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='91 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='49 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='134 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='130 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='125 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='133 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='126 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='99 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='123 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='83 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='92 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='95 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='58 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='65 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='72 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='94 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='100 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='135 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='137 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='110 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='118 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='88 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='89 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='57 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='42 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='43 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='67 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='52 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='78 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='124 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='96 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='109 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='120 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='107 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='101 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='112 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='105 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='108 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='39 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='33 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='38 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='127 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='131 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='121 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='106 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='103 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='104 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='111 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='102 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='115 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='113 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='117 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='85 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='97 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='132 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='128 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='129 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='119 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='98 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='116 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='59 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='64 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='66 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='34 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='26 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='29 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='27 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='41 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='71 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='86 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='37 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='Epochs ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='500 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1000 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1500 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2000 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2500 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3000 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='Euclidean distance ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='Dendrogram (PCC) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Dendrogram obtained for PCC using agglomerative clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='69 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='70 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='51 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='68 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='122 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='97 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='98 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='75 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='87 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='31 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='85 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='119 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='113 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='117 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='76 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='84 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='80 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='82 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='36 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='77 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='28 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='44 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='107 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='120 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='108 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='105 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='112 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='37 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='39 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='33 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='38 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='71 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='115 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='102 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='106 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='101 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='86 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='103 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='104 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='29 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='26 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='34 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='41 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='131 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='128 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='129 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='127 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='132 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='116 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='111 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='121 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='27 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='64 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='59 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='66 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='93 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='53 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='61 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='123 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='114 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='136 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='62 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='73 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='52 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='67 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='99 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='110 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='118 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='88 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='109 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='65 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='72 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='89 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='94 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='95 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='42 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='43 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='96 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='57 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='78 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='49 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='47 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='48 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='124 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='126 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='134 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='135 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='137 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='100 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='90 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='92 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='81 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='83 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='74 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='79 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='56 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='63 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='58 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='54 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='55 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='91 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='46 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='130 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='125 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='133 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='Epochs ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='500 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1000 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='1500 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='2000 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='Euclidean Distance ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='Dendrogram (DCC) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content='FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' Dendrogram obtained for DCC using agglomerative clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
+page_content=' 18' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6NE4T4oBgHgl3EQfcAzU/content/2301.05080v1.pdf'}
diff --git a/6dE1T4oBgHgl3EQf7AUK/content/2301.03528v1.pdf b/6dE1T4oBgHgl3EQf7AUK/content/2301.03528v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..404917b325b3adedb29d22aae32c236951b5cbc0
--- /dev/null
+++ b/6dE1T4oBgHgl3EQf7AUK/content/2301.03528v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfdbf0caa1502c6a7d92ebab18305b855dcf68a82e0fcbb7d605ec65cfc99ae6
+size 1075593
diff --git a/6dE1T4oBgHgl3EQf7AUK/vector_store/index.faiss b/6dE1T4oBgHgl3EQf7AUK/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..e72576d557a937e6546986f3c982cf9e85e7be32
--- /dev/null
+++ b/6dE1T4oBgHgl3EQf7AUK/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f088b90fd6373a9ee176e8683716bf1098ded93806d8cfa29b84d086407520b
+size 2097197
diff --git a/6tE1T4oBgHgl3EQfnAQs/content/tmp_files/2301.03303v1.pdf.txt b/6tE1T4oBgHgl3EQfnAQs/content/tmp_files/2301.03303v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4aaf0e3021067e19f9bc7614484247233b6b5027
--- /dev/null
+++ b/6tE1T4oBgHgl3EQfnAQs/content/tmp_files/2301.03303v1.pdf.txt
@@ -0,0 +1,2259 @@
+
+
+1
+How Effective are COVID-19 Vaccine Health Messages in Reducing Vaccine Skepticism?
+Heterogeneity in Messages’ Effectiveness by Just-World Beliefs
+
+
+
+Juliane Wiese, corresponding author
+
+
+
+ Nattavudh Powdthavee
+
+Warwick Business School
+
+
+
+ Nanyang Technological University
+University of Warwick
+
+
+
+
+
+ 50 Nanyang Avenue
+Scarman Road
+
+
+
+
+
+
+
+ 639798 Singapore
+Coventry CV4 7AL
+
+
+
+
+
+
+
+United Kingdom
+
+
+
+
+
+
+
+
+
+ORCID ID : 0000-0002-4314-5934
+
+ ORCID ID: 0000-0002-9345-4882
+
+juliane.wiese@warwick.ac.uk
++33 6 68 88 18 27
+
+
+Declarations of interest: none.
+
+Abstract
+
+To end the COVID-19 pandemic, policymakers have relied on various public health messages to
+boost vaccine take-up rates amongst people across wide political spectra, backgrounds, and
+worldviews. However, much less is understood about whether these messages affect different
+people in the same way. One source of heterogeneity is the belief in a just world (BJW), which is
+the belief that in general, good things happen to good people, and bad things happen to bad people.
+This study investigates the effectiveness of two common messages of the COVID-19 pandemic:
+vaccinate to protect yourself and vaccinate to protect others in your community. We then examine
+whether BJW moderates the effectiveness of these messages. We hypothesize that just-world
+believers react negatively to the prosocial pro-vaccine message, as it charges individuals with the
+responsibility to care for others around them. Using an unvaccinated sample of UK residents before
+vaccines were made widely available (N=526), we demonstrate that the individual-focused
+message significantly reduces overall vaccine skepticism, and that this effect is more robust for
+
+
+
+2
+individuals with a low BJW, whereas the community-focused message does not. Our findings
+highlight the importance of individual differences in the reception of public health messages to
+reduce COVID-19 vaccine skepticism.
+
+Keywords: vaccine skepticism; health messages; justice beliefs; individual differences; COVID-
+19
+
+
+
+
+
+3
+1. Introduction
+
+Before the vaccine rollout in the UK, 28% of the British population, particularly those in Black
+and South Asian minority ethnic groups, were skeptical about getting vaccinated (Robertson et al.,
+2021). To maximize vaccine take-up, governments have been delivering simple messages that
+emphasize people’s responsibility to themselves and the community. For example, the National
+Health Services in the UK urges the public to “join the millions already vaccinated, to protect
+yourself and others” (NHS UK, 2021). These foci, given their central role in public health
+messaging during the COVID-19 pandemic so far, have shaped the two themes of messages
+examined in this study: individual and community responsibilities.
+
+Despite the extensive literature on the framing approaches of public health messages around
+vaccines (e.g., Gallagher & Updegraff, 2012; McPhee et al., 2003; Kelly & Kornik, 2016), the
+overall effectiveness of COVID-19 vaccine messages on individual or community responsibility
+is currently imperfectly understood. While recent evidence suggests that individual-focused
+messages more effectively increase vaccine uptake and support for mandates than community-
+focused messages, these effects are heterogeneous across individualistic and communitarian
+worldviews (Yuan & Chu, 2022). Furthermore, we do not know which underlying beliefs about
+the vaccine are best addressed by these messages. Nevertheless, they continue to be used by
+public health officials worldwide.
+
+In contexts of extreme urgency, who are the types of people who might respond poorly to these
+messages and experience stronger vaccine skepticism? We build our investigation around the
+strong theoretical link between belief in a just world (BJW) and vaccine skepticism. Just-world
+
+
+
+4
+believers conceive a universal justice structure which holds that both normatively and positively
+speaking, good things tend to happen to good people and vice versa (Furnham, 2003). This
+adaptive function (Dalbert, 2009), manifesting at varying levels of intensity and therefore
+influencing a large portion of the population (White et al., 2019), allows individuals to
+rationalize negative consequences in the world as justified, predictable, and manageable. Doing
+so promotes well-being and a sense of stability in the world (Correia et al., 2009; Jiang et al.,
+2016). In the context of the COVID-19 pandemic, where an unprecedented public health
+emergency and sweeping government regulations significantly reduced individual freedoms,
+just-world believers struggled to make sense of such undeserved restrictions. This sense of
+unfairness fosters a resistance against the government-promoted solution to the problem:
+specifically, a vaccine that has been developed in record speed. Suggestive evidence of this link
+between just-world believers and anti-vaxxers is demonstrated by their numerous shared
+psychological traits, including conspiracy thinking (Nestik et al., 2020; Jolley & Douglas, 2014)
+and individualistic attitudes (Wenzel et al., 2017; Motta et al., 2021). Government-sponsored
+pro-vaccine messages, particularly ones that focus on the responsibility we hold to our
+communities, are therefore likely to threaten the just-world believers’ worldview, as their
+personal role in the pandemic is limited, and others’ health outcomes are independent of their
+own decision to get vaccinated. Their worldview threatened, just-world believers defensively
+dismiss the message that threatens their BJW, and deny the existence of a problem in the first
+place (Furnham, 2003).
+
+This study makes two main contributions to the literature. First, we experimentally investigate the
+effectiveness of two commonly used pro-vaccine messages. Second, we examine whether BJW
+
+
+
+5
+moderates the effectiveness of each message. Given policymakers’ priority to increase COVID-19
+vaccine uptake, understanding individual differences in the messages’ effectiveness by BJW is
+critical to understanding the potential threats to their overall effectiveness on the entire population.
+
+2. Existing literature and hypotheses
+Before the vaccine rollout, researchers’ main concern was whether the COVID-19 vaccines safely
+reduce illness and transmissibility. Having established this (Katella, 2021; Pritchard et al., 2021),
+vaccine uptake has emerged as a more enduring challenge for public health officials. A nationally
+representative survey of 316 Americans shows that demonstrating its efficacy and emphasizing
+the costs of the pandemic encourages vaccine uptake (Pogue et al., 2020). However, their survey
+did not engage with messages that focus on the simple facts that give value to the vaccine: that it
+protects its recipients and their community. These facts have been central to policymakers’
+messaging during the COVID-19 pandemic, and there continues to be little empirical investigation
+into their effectiveness in shifting public perception around the vaccine’s effectiveness.
+
+The decision to vaccinate weighs the benefits against the risks of vaccination, which could range
+from fears of side effects and needles to mistrust of healthcare authorities. Previous research
+demonstrates the importance of highlighting vaccines’ protective benefits, as doing so can crowd
+out concerns about risks (Porter et al., 2018). Similarly, a COVID-19 vaccine message highlighting
+the vaccine’s protective benefits to the individual has been shown to increase intended vaccine
+uptake (Yuan & Chu, 2022). Our work examines how such an individualistic message can drive
+the underlying beliefs around the vaccine’s protective function to its recipients.
+
+
+
+
+6
+In addition, researchers have found prosocial vaccine messages to have a positive impact on
+vaccination rates (Betsch et al., 2017; Betsch & Böhm, 2018; McPhee et al., 2003). For example,
+messages that emphasize the benefits of an avian flu vaccine to others significantly increase
+vaccination intentions, compared to messages which emphasize its benefits to the individual (Kelly
+& Hornick, 2016). While these findings link the community-oriented message to increased
+vaccination intentions, they do not examine how such a message impacts beliefs around
+transmission rates, which is the mechanism that connects the prosocial messages with increased
+vaccine uptake. We aim to show experimentally that prosocial messages increase confidence in
+the underlying belief that the vaccines reduce transmission.
+
+Based on this evidence, we predict that the individual message will more effectively decrease
+overall skepticism than the community message, and that this effect is driven by the fact that the
+individual message shifts the underlying belief that the vaccine protects its recipients. The
+prosocial messages will more moderately increase confidence that the vaccine reduces
+transmission.
+
+Despite the predicted overall success of the two messages, the question remains around
+heterogeneous effects, specifically around moral worldviews that play a role in the decision to
+vaccinate. While Devereux et al. (2021) discover a link between stronger BJW and a greater
+likelihood to adhere to COVID-19 measures, such as social distancing, these measures come at
+essentially zero risk, resulting in a very different cost-benefit analysis. In contrast, accepting a
+vaccine requires accepting the risk of potential negative side-effects, and might therefore have a
+different relationship with BJW.
+
+
+
+7
+
+Demographic factors (Peretti-Watel et al., 2020; Khubchandani et al., 2021), psychological traits
+(Browne et al., 2015; Jolley & Douglas, 2014), and beliefs about vaccine safety (Karlsson et al.,
+2021) predict vaccine attitudes. However, studies that examine how such traits, like BJW, interfere
+with public health messages are scarce. While recent evidence has shown that people with more
+individualistic, rather than communitarian, values respond more favorably to individual-centered
+COVID-19 vaccine messages (Yuan & Chu, 2022), it remains unclear how such worldviews
+moderate individuals’ understanding of the many ways in which the vaccine protects the public.
+Furthermore, rather than simply capturing individualistic or community-oriented worldviews,
+BJW contains a deeper moral around one’s deservingness of one’s place in the world, telling us
+more about the reasoning behind an individual’s action (or inaction).
+
+While people who see public health as a moral issue tend to consider prosocial (vs. self-centered)
+social distancing messages more persuasive (Luttrell & Petty, 2020), BJW is not an altruistic moral
+belief system. Instead, it holds individuals responsible for their own fate. BJW inherently commits
+fundamental attribution error, in which individuals place more weight on dispositional, as opposed
+to environmental or situational, factors (Ross, 1977). By further emphasising societal
+responsibility as a motive to get vaccinated, public health officials transfer the responsibility for a
+COVID patient’s health onto the community’s vaccination decision-making. This clashes with the
+tendency of just-world believers to blame patients for their own misfortunes and to separate the
+consequences of their own actions from the outcomes of others (Lerner & Simmons, 1966; Lucas
+et al., 2009). Therefore, by asking people to take responsibility for others’ health and safety during
+the COVID-19 pandemic, policymakers inevitably challenge the justice structure of the world in
+
+
+
+8
+which individuals are responsible for their own fate. In response, just-world believers might
+discredit the vaccine altogether. We therefore hypothesise that for individuals with a strong BJW,
+the prosocial messages are less effective at reducing vaccine skepticism.
+
+3. Method
+3.1 Data
+In this pre-registered experiment (tinyurl.com/bxv23), 600 UK-based Prolific (www.prolific.co)
+users aged between 18 and 49 joined a longitudinal online study on attitudes towards COVID-19
+and vaccination. At the time, the UK general public under 50 years of age was not yet eligible to
+receive a COVID-19 vaccine. Just over a quarter of the UK population had received its first dose,
+and only 1% of the population had received both doses (Vaccinations in United Kingdom, 30 April,
+2021).
+
+Part one of the study (𝑇!) took place on 24 February 2021, and part two (𝑇") on 1 March 2021. We
+collected data at two points in time to reduce the likelihood that (i) participants suspect the study
+purpose and bias their responses, and (ii) participants’ responses to vaccine skepticism questions
+are biased by exposure to questions around justice beliefs (Zizzo, 2010). Participants gave
+informed consent and were compensated £0.25 at 𝑇! and £1.00 at 𝑇".
+
+527 participants (88%) remained at 𝑇" and were randomised evenly across Control, Individual-
+Treatment, and Prosocial-Treatment (N = 172, 181, and 174, respectively). Only one participant
+failed all three attention checks and was removed from the sample, resulting in 526 participants
+with complete longitudinal data. This sample size (i) allowed sufficient power for a reasonable
+
+
+
+9
+minimal detectable effect size and (ii) is slightly larger than what was used in a similar research
+design studying BJW and climate change messaging (Feinberg & Willer, 2011). Of the final
+sample of 526 individuals, 70% were females, 87% were ethnically White, and 59% have an annual
+income of £30,000 or over. The mean age was 31. Balance checks confirm that our sample was
+balanced on observable characteristics across all groups; see Table A.1 in the appendix.
+
+3.2 Measures and procedure
+3.2.1 BJW scales
+Because vaccination evokes concepts of justice both for the individual and for society, participants
+completed the general BJW scale, six questions about the justice structure in the world in general
+(Dalbert et al., 1987), and the personal BJW scale, seven questions which posit that the world is
+just for me personally but not for others (Dalbert, 1999) at 𝑇!. The two scales have a correlation
+coefficient of 0.52. To attain a linear combination of BJW factors, we conducted a separate factor
+analysis on each scale, yielding two distinct factors (a = 0.78 for general BJW and a = 0.88 for
+personal BJW), and then conducted a factor analysis on these factors, resulting in a combined BJW
+factor (a = 0.68); the factor analysis results are in Table A.2. The resulting combined BJW factor
+was standardized to a mean of 0 and a standard deviation of 1. It was transformed into a dummy
+variable which marks above- or below-median strength of BJW. This allows us to investigate the
+differential effects of the treatments on vaccine skepticism by the strength of BJW.
+
+3.2.2 Vaccine skepticism
+
+
+
+10
+At 𝑇! and 𝑇", participants completed four questions on COVID-19 vaccine skepticism, with
+possible answers ranging from 0 (not at all certain/likely) to 100 (extremely certain/likely). The
+precise wording of the questions was:
+• “How certain are you that the COVID-19 vaccines are a useful tool in fighting the
+pandemic?”
+• “How likely are you to accept the COVID-19 vaccine when offered?”
+• “How certain are you that the COVID-19 vaccine reduces transmission between
+individuals?”
+•
+“How certain are you that the COVID-19 vaccine would prevent you personally from
+getting very ill due to COVID-19?”
+For simplicity, we reversed the responses so that higher values represent higher levels of vaccine
+skepticism in each of the four outcomes. The baseline mean responses are 16.8, 13.0, 29.1, and
+22.4, respectively, which suggest that at 𝑇!, the study population was relatively prepared to take
+the vaccine but was more skeptical of its illness and transmission prevention. These outcomes are
+moderately correlated, with correlations ranging from 0.53 to 0.76. To circumvent the multiple
+comparisons problem, we also derived an overall skepticism outcome by conducting a factor
+analysis on the four reversed individual skepticism variables for both outcomes at 𝑇! (a = 0.87)
+and 𝑇" (a = 0.89); see Tables A.3 and A.4 in the appendix for the estimates. All skepticism
+variables were standardized to have a mean of 0 and a standard deviation of 1 and were included
+in analysis.
+
+At 𝑇", 5 days after 𝑇!, participants were randomised into one of three groups: control (no article),
+individual, and community responsibility treatment. In both treatments, participants were asked to
+
+
+
+11
+read a news-style article. The articles, Figure A.3.1 in the appendix, are identical in the first
+paragraphs, which discuss the context of the pandemic and vaccine development at the time of
+writing. They deviate towards the end by treatment group. The individual responsibility article
+explains that the vaccine reduces the risk of severe COVID-19 illness to vaccine recipients, and
+the prosocial article explains that to combat the virus, individuals must accept the vaccine to reduce
+community transmission.
+
+3.2.3 Attention and manipulation check
+Participants in the treatment groups were asked two fact-based questions from the article, as well
+as whether taking the recommended steps during the pandemic will mainly protect them, or mainly
+protect others, from COVID-19 illness. Amongst the final sample of participants who passed all
+three attention checks, we find a significant difference between the two treatments on the
+manipulation-check item, t(352) = 13.64, p = 0.000 for indicating that the vaccine protects
+yourself, and t(352) = -13.07, p = 0.000 for indicating that the vaccine protects others.
+
+3.2.4 Sociodemographic controls
+Participants also completed a post-experiment questionnaire, which elicited their ethnicity,
+education level, region, income, political views, optimism, risk attitudes, COVID-19 history, and
+adhesion to government guidelines. Age and gender were collected automatically by Prolific.
+
+Figure A.1 shows the procedural flow of the experiment and consort diagram, and Figures A.2 and
+A.3 present screenshots of the materials used.
+
+
+
+
+12
+3.3 Analysis
+We conduct all analyses of vaccine skepticism using Ordinary Least Squares (OLS) regression
+with robust standard errors clustered on the participant-level. Our primary analysis examines the
+treatment effects on the overall vaccine skepticism factor. We regress Equation (1) and present the
+results in column 3 of Table 1:
+∆𝑆#$ = 𝑎 + 𝛽"𝑇# + 𝑋#
+%𝛾 + 𝛽&𝐵𝐽𝑊# + 𝛽'(𝑇# × 𝐵𝐽𝑊#) + 𝑒, (1)
+
+where 𝑖 = 1, … , 𝑁; 𝑡 = 1, … ,2. ∆𝑆#$ represents the change in the overall vaccine skepticism factor
+from t=0 to t=1, where a higher value represents greater vaccine skepticism; Ti represents the
+treatment condition (control, individual, or community message) and 𝛽" is the effect of this
+condition on skepticism; 𝑋#
+% represents the matrix of covariates, including a standardized optimism
+factor (a=0.8134), age, age-squared, gender dummy, £30,000+ annual income (vs. below £30,000
+annual income) dummy, London (vs. non-London) dummy, undergraduate education (vs. non-
+undergraduate education) dummy, white (vs. non-white) dummy, Labour party (vs. non-Labour)
+dummy; 𝛽& is the effect of holding a strong (vs. weak) BJW; 𝛽' represents the interaction of
+treatment and BJW, i.e. the differential effect of the treatment when participants have either a
+stronger or a weaker BJW; and e is the error term. Columns 1 and 2 model the parsimonious
+specifications of Eq. (1), with covariates excluding and including BJW, respectively.
+
+Table 2 models the effects of the interaction between treatment and BJW on each of the four
+skepticism outcomes. Their forms are identical to Eq. (1), with the exception that the outcome
+variable is replaced by each of the four vaccine skepticism subscales, standardized to mean of 0
+and standard deviation of 1.
+
+
+
+13
+∆𝑆()$
+; = 𝑎 + 𝛽"𝑇# + 𝑋#
+%𝛾 + 𝛽&𝐵𝐽𝑊# + 𝛽'(𝑇# × 𝐵𝐽𝑊#) + 𝑒, (2)
+where 𝑖 = 1, … , 𝑁; 𝑗 = 1, … ,4; 𝑡 = 1, … ,2. Here, ∆𝑆()
+; = 𝑆()$
+; − 𝑆()$*"
+? , where the notation j
+represents different domains of beliefs, e.g., 𝑆"# represents the belief that the vaccine is not useful;
+𝑆 represents the likelihood of not accepting the vaccine; 𝑆'# represents the belief that the vaccine
+will not reduce transmission; and 𝑆+# represents the belief that the vaccine will not prevent serious
+illness. The rest of the specification is identical to Eq. (1).
+
+Note that we deviate from the pre-registered document in two respects. First, we include an overall
+skepticism factor as an outcome variable in our primary analysis, circumventing the multiple
+comparisons problem in our primary analysis. Second, we run OLS regressions with standard
+errors clustered at the participant level as the primary analysis rather than using analysis of
+variance (ANOVA). This change is made due to the inclusion of continuous independent variables
+in the regression.
+
+4. Results
+4.1 Message effectiveness
+We begin by examining the within-person changes in vaccine skepticism by treatment group. As
+predicted, Figure 1 shows that the individual message significantly reduces overall skepticism by
+0.04 standard deviation, compared to the control group which increases overall skepticism by 0.07
+standard deviation (Wilcoxon signed-rank test, p = 0.030). There is weaker evidence that the
+community message also reduces overall skepticism, which decreased by 0.02 standard deviation
+(Wilcoxon signed-rank test, p = 0.103). Figure 1 thus provides raw data evidence that individual-
+focused public health message is most effective at reducing overall vaccine skepticism.
+
+
+
+14
+[Figure 1 here]
+To understand this result more thoroughly, Table 1 estimates regression equations that adjust for
+other covariates, i.e., Eq.1. We find the regression results to be consistent with Figure 1’s findings.
+The individual-focused message decreases overall skepticism more robustly than the community
+message, b = -0.11, [95% C.I.: -0.20, -0.02], p = 0.014, versus b = -0.09, [95% C.I.: - 0.19, 0.01],
+p = 0.083, respectively.
+[Table 1 here]
+4.2 BJW as a moderator of pro-vaccine message impacts
+To formally test for the heterogeneous effect of public health messages by BJW, Tables 1 and 2
+include the interaction terms between treatment and a high BJW dummy. Column 3 of Table 1
+shows that for people with a low BJW, the individual message is extremely effective at lowering
+their overall skepticism factor, b = - 0.19, [95% C.I.: - 0.32, -0.06], p = 0.004. As discussed
+earlier, columns 1 and 2 of Table 1 demonstrate a greater effectiveness of the individual message
+on average. The results of column 3 suggest that the effectiveness of this individualistic message
+is more robust for people with a low BJW, whereas we see no such differential effect for the
+collective message. Figures 2 and 3 show this distinction visually, with the predictive margins
+plots of the control and individual treatment overlapping (Figure 2), and the predictive margins
+plots of the control and collective treatment (Figure 3) not overlapping. When examining the
+interaction regressions for each sub-scale of vaccine skepticism (Table 2), we find that the strong
+effect of the individual treatment on overall skepticism for people with a low BJW is driven by a
+reduction in skepticism around the belief that the vaccine will not prevent illness, b = - 0.32,
+[95% C.I.: - 0.50, -0.14], p < 0.001. This suggests that people with a low BJW, i.e. those who do
+not believe that there is a justice system which ensures that overall good things happen to good
+
+
+
+15
+people and bad things happen to bad people, are extremely reactive to the individualistic message.
+It increases their confidence in the vaccine being able to protect them from serious illness. In other
+words, receiving the individualistic message, which accurately highlights that receiving the
+vaccine can prevent serious illness, correctly updates the beliefs around this issue for those with a
+low BJW, but not for those with a strong BJW. This suggests that for someone with a strong BJW,
+the belief in this just world order overpowers the belief in the science of the vaccine, as perhaps
+the deservingness of a person to fall ill would govern their likelihood of sickness moreso than the
+vaccine’s protective properties.
+[Figures 2 and 3 here]
+Furthermore, we do not find evidence that people with a strong BJW react particularly poorly to
+the community message, b = 0.03, [95% C.I.: - 0.16, 0.22], p = 0.778. This suggests that a
+message which urges the public to take care of its community does not come into strong conflict
+with believers of a just world who may not feel responsible for the pandemic. This lack of
+resistance is consistent with just world believers’ willingness to engage in other COVID-19
+preventative measures (Devereux et al., 2021).
+[Table 2 here]
+
+5. Discussions
+Our findings that the individual and community messages concerning the COVID-19 vaccine can
+shift beliefs around the vaccine’s various protective functions demonstrates an unsurprising link
+between the presentation of fact and its influence on a corresponding attitude. Nevertheless, in
+their desperate attempts to convince the public to get vaccinated, policymakers have sometimes
+turned to extreme measures, such as million-dollar lotteries, rifle giveaways, and free beer and
+
+
+
+16
+donuts (Lewis 2021). However, while policymakers may have expected a clear increase in uptake,
+emerging evidence suggests that there is limited evidence in favor of these creative incentivizing
+strategies (Walkey et al., 2022; Acharya & Dhakal, 2021), perhaps due to newfound suspicion of
+such gimmicky programs. Instead, policymakers should provide truthful information about the
+capacities of the COVID-19 vaccine, relying on existing evidence that these strategies effectively
+lower vaccine skepticism (Pennycook et al., 2020; Yuan & Chu, 2022).
+
+Our messages do not easily shift the belief that the vaccine reduces transmission of the virus. This
+is especially important as new evidence emerges around the limited effectiveness of the vaccines
+against mutations of the coronavirus and in preventing transmission. Early studies suggest that the
+COVID-19 vaccines may not be as effective in preventing transmission as previously thought
+(Reuters, 2021). While policymakers should highlight the protective benefits of the vaccine, they
+must be cautious in not overstating the vaccine’s effectiveness around transmission. Doing so
+could give vaccinated individuals a false sense of security, and ultimately reduced trust in public
+health authorities, resulting in less social distancing and respect for COVID-19 guidelines. As new
+scientific evidence about the vaccine emerges, officials must update their messaging content
+accordingly.
+
+The literature shows that prosocial messages play an important role in motivating COVID-19
+preventative actions, like signing up for contact-tracing apps (Jordan et al., 2020). In contrast,
+vaccine skepticism responds differently. Consistent with previous findings (Yuan & Chu, 2022),
+we show that individual responsibility messages work as well, and sometimes better, than the
+community messages in reducing vaccine skepticism, depending on the dimension of skepticism
+
+
+
+17
+in question. This discrepancy between non-vaccine COVID-19 prevention and vaccine messages
+could be because general preventative measures are perceived to be less risky than taking the
+vaccine. Riskier behaviors require more self-gain, which explains why the individual message is
+more successful.
+
+Furthermore, the pro-vaccine messages used in this experiment affect different domains of vaccine
+skepticism differently. More specifically, they do not convince the population that the vaccine is
+useful to ending the pandemic, nor do they influence vaccination intentions. In the urgent
+pandemic context, while attitudes matter, vaccination behaviors are even more critical. Alternative
+strategies to motivate behavior must not be overlooked or confounded with strategies that target
+attitudes in future research.
+
+When further examining heterogeneous treatment affects by intensity of BJW, we find that the
+overall success of the individual message is more robust among individuals with a low BJW,
+compared to those with a high BJW. The individual message, which focusses on the primary effect
+of the vaccine, may speak more particularly to people with a weak BJW because they see the world
+in a more factual, cartesian way. Someone with a strong BJW, on the other hand, may consider
+competing justice-related reasonings for the spread of or protection against COVID. The same is
+not true of the effects of the community message. Individuals with a strong BJW were found to be
+unmoved by the community message, possibly because this prosocial message sets an expectation
+that challenges the distribution of responsibility in a just world, as previously discussed. While
+individuals who see public health as a moral issue are more persuaded by other-focused (rather
+than self-focused) social distancing messages (Luttrell & Petty, 2020), BJW is not a worldview
+
+
+
+18
+based on altruistic morals. Rather, where others may fall ill due to COVID-19, strong believers of
+a just world would blame the patients for their own misfortune, rather than assuming responsibility
+over the pandemic via mass collective vaccination.
+
+Our results suggest that evidence-based messages (e.g.: the vaccine will protect you) have
+heterogeneous effects according to worldview. This heterogeneity replicates the findings of Yuan
+& Chu, who recently demonstrate that the individual-centered COVID-19 vaccine message is more
+impactful than a community-centered one, largely due to people whose worldview aligns with a
+more individualistic outlook (2022). Our studies differ in that we examine BJW, rather than
+individualism/communitarianism, and our sample was based in the UK, rather than the US.
+However, broadly speaking, the results confirm one another’s findings, which is that the
+individual-centered message works best overall, but that this effect is driven largely by people with
+a worldview that places themselves, the individual, independent of a larger community or justice
+structure, at the center. Authorities ought to take into consideration the extent to which their
+vaccine messaging can have heterogeneous effects according to the worldviews of their
+population, especially as they encourage vaccine take-up amongst people with more extreme
+worldviews.
+
+6. Conclusions
+Simple messages that promote the COVID-19 vaccine effectively reduce vaccine skepticism of
+the corresponding beliefs around the vaccine’s effectiveness. This reassuringly highlights the
+importance for policymakers to focus the information of their vaccination campaigns on the
+specific concerns of the public. The differences we find in effectiveness by psychological outlook
+
+
+
+19
+are important for policymakers to consider, especially as the remaining unvaccinated likely hold
+more extreme world views. Messages that work well for people with low-level BJW evidently
+work less well for those with a more extreme worldview, suggesting that policymakers must
+reconsider how to motivate those harder-to-reach populations to get vaccinated. Custom messages
+that directly target people with such views could be an interesting line of research to follow.
+
+This research is not without limitations. First, the data is restricted to a specific age-group in the
+United Kingdom and therefore has not been tested in other contexts, where just-world beliefs and
+vaccine skepticism differ. For example, in the United States, conservatism links with both BJW
+(Furnham, 2003) and COVID-19 vaccine skepticism (Latkin et al., 2021), suggesting that BJW
+might be negatively correlated with pro-vaccine attitudes. Second, the sample in our study is not
+quota matched to the U.K. population, nor was it obtained using probability sampling. Hence, the
+results cannot be considered nationally representative, and there is likely a degree of selection bias
+amongst users of Prolific. Third, our dataset does not capture whether participants ultimately took
+up the vaccination, as it only captures attitudes and intentions. As previously discussed, behaviors
+in this context are more important than attitudes, and would be valuable to follow up on.
+
+The authors declare no conflicts of interest.
+
+
+
+
+
+20
+References
+
+Acharya, B., & Dhakal, C. (2021). Implementation of state vaccine incentive lottery programs
+and uptake of COVID-19 vaccinations in the United States. JAMA Network Open, 4(12),
+e2138238-e2138238.
+Betsch, C., & Böhm, R. (2018). Moral values do not affect prosocial vaccination. Nature Human
+Behaviour, 2(12), 881-882. https://doi.org/10.1038/s41562-018-0478-1.
+Betsch, C., Böhm, R., Korn, L., & Holtmann, C. (2017). On the benefits of explaining herd
+immunity in vaccine advocacy. Nature Human Behaviour, 1(3), 1-6.
+https://doi.org/10.1038/s41562-017-0056.
+Browne, M., Thomson, P., Rockloff, M. J., & Pennycook, G. (2015). Going against the herd:
+psychological and cultural factors underlying the ‘vaccination confidence gap’. PLoS
+one, 10(9). https://doi.org/10.1371/journal.pone.0132562.
+Correia, I., Batista, M. T., & Lima, M. L. (2009). Does the belief in a just world bring happiness?
+Causal relationships among belief in a just world, life satisfaction and mood. Australian
+Journal of Psychology, 61(4), 220-227. https://doi.org/10.1080/00049530802579515.
+Dalbert, C. (2009). Belief in a just world. Handbook of individual differences in social behavior,
+288-297.
+Dalbert, C. (1999). The world is more just for me than generally: About the personal belief in a
+just world scale's validity. Social justice research, 12(2), 79-98.
+ https://doi.org/10.1023/A:1022091609047
+Dalbert, C., Montada, L., & Schmitt, M. (1987). Glaube an eine gerechte Welt als Motiv:
+Validierungskorrelate zweier Skalen. Psychologische Beitrage.
+Devereux, P. G., Miller, M. K., & Kirshenbaum, J. M. (2021). Moral disengagement, locus of
+
+
+
+21
+control, and belief in a just world: Individual differences relate to adherence to COVID-
+19 guidelines. Personality and Individual Differences, 182, 111069.
+https://doi.org/10.1016/j.paid.2021.111069
+Feinberg, M., & Willer, R. (2011). Apocalypse soon? Dire messages reduce belief in global
+warming by contradicting just-world beliefs. Psychological science, 22(1), 34-38.
+https://doi.org/10.1177/0956797610391911.
+Furnham, A. (2003). ‘Belief in a just world: Research progress over the past decade’, Personality
+and Individual Differences, 34(5). https://doi.org/10.1016/S0191-8869(02)00072-7.
+Gallagher, K. M., & Updegraff, J. A. (2012). Health message framing effects on attitudes,
+intentions, and behavior: a meta-analytic review. Annals of Behavioral Medicine, 43(1),
+101-116. https://doi.org/10.1007/s12160-012-9446-6.
+Jiang, F., Yue, X., Lu, S., Yu, G., & Zhu, F. (2016). How belief in a just world benefits mental
+health: The effects of optimism and gratitude. Social Indicators Research, 126(1), 411-
+423. https://doi.org/10.1007/s11205-015-0877-x.
+Jolley, D., & Douglas, K. M. (2014). The effects of anti-vaccine conspiracy theories on
+vaccination intentions. PloS one, 9(2), e89177.
+https://doi.org/10.1371/journal.pone.0089177
+Jordan, J., Yoeli, E., & Rand, D. (2020). Don’t get it or don’t spread it? Comparing self-
+interested versus prosocially framed COVID-19 prevention messaging. PsyArXiv, 10.
+Karlsson, L. C., Soveri, A., Lewandowsky, S., Karlsson, L., Karlsson, H., Nolvi, S., ... &
+Antfolk, J. (2021). Fearing the disease or the vaccine: The case of COVID-
+19. Personality and individual differences, 172, 110590.
+https://doi.org/10.1016/j.paid.2020.110590.
+
+
+
+22
+Katella, K. (2021). Comparing the COVID-19 Vaccines: How are They Different?. Yale
+Medicine. Available at: http://www.yalemedicine.org/news/covid-19-vaccine-comparison
+(Accessed: 30 April, 2021).
+Kelly, B.J. and Hornik, R.C., 2016. Effects of framing health messages in terms of benefits to
+loved ones or others: An experimental study. Health Communication, 31(10), pp.1284-
+1290. https://doi.org/10.1080/10410236.2015.1062976
+Khubchandani, J., Sharma, S., Price, J. H., Wiblishauser, M. J., Sharma, M., & Webb, F. J.
+(2021). COVID-19 vaccination hesitancy in the United States: a rapid national
+assessment. Journal of Community Health, 46(2), 270-277.
+
+https://doi.org/10.1007/s10900-020-00958-x.
+Latkin, C. A., Dayton, L., Yi, G., Colon, B., & Kong, X. (2021). Mask usage, social distancing,
+racial, and gender correlates of COVID-19 vaccine intentions among adults in the
+US. PloS one, 16(2), e0246970. https://doi.org/10.1371/journal.pone.0246970.
+Lerner, M. J., & Simmons, C. H. (1966). Observer's reaction to the" innocent victim":
+Compassion or rejection?. Journal of Personality and social Psychology, 4(2), 203.
+https://doi.org/10.1037/h0023562.
+Lewis, T. (2021). From $1-Million Lotteries to Free Beer: Do COVID Vaccination Incentives
+Work? Scientific American. https://www.scientificamerican.com/article/from-1-million-
+lotteries-to-free-beer-do-covid-vaccination-incentives-work1/.
+Lucas, T., Alexander, S., Firestone, I., & Lebreton, J. M. (2009). Belief in a just world, social
+influence and illness attributions: Evidence of a just world boomerang effect. Journal of
+Health Psychology, 14(2), 258-266. https://doi.org/10.1177/1359105308100210.
+Luttrell, A., & Petty, R. E. (2021). Evaluations of self-focused versus other-focused arguments
+
+
+
+23
+for social distancing: An extension of moral matching effects. Social Psychological and
+Personality Science, 12(6), 946-954. https://doi.org/10.1177/1948550620947853.
+McPhee, S.J., Nguyen, T., Euler, G.L., Mock, J., Wong, C., Lam, T., Nguyen, W., Nguyen, S.,
+Ha, M.Q.H., Do, S.T. and Buu, C., 2003. Successful promotion of hepatitis B
+vaccinations among Vietnamese-American children ages 3 to 18: results of a controlled
+trial. Pediatrics, 111(6), pp.1278-1288. https://doi.org/10.1542/peds.111.6.1278.
+Motta, M., Callaghan, T., Sylvester, S., & Lunz-Trujillo, K. (2021). Identifying the prevalence,
+correlates, and policy consequences of anti-vaccine social identity. Politics, Groups, and
+Identities, 1-15. https://doi.org/10.1080/21565503.2021.1932528.
+Nestik, T. A., & Deyneka, O. S. (2020). Socio-psychological predictors of belief in conspiracy
+theories of the origin of COVID-19 and involvement in social media. Social Psychology
+and Society, 11(4), 87-104. https://doi.org/10.17759/sps.
+NHS UK (2021) 25 April. Available at https://twitter.com/NHSuk (Accessed: 18 May 2021).
+Pennycook, G., McPhetres, J., Zhang, Y., Lu, J. G., & Rand, D. G. (2020). Fighting COVID-19
+misinformation on social media: Experimental evidence for a scalable accuracy-nudge
+intervention. Psychological Science, 31(7), 770-780.
+https://doi.org/10.1177/0956797620939054.
+Peretti-Watel, P., Seror, V., Cortaredona, S., Launay, O., Raude, J., Verger, P., ... & Ward, J. K.
+(2020). A future vaccination campaign against COVID-19 at risk of vaccine hesitancy
+
+and politicisation. The Lancet Infectious Diseases, 20(7), 769-770.
+https://doi.org/10.1016/S1473-3099(20)30426-6.
+Pogue, K., Jensen, J. L., Stancil, C. K., Ferguson, D. G., Hughes, S. J., Mello, E. J., ... & Poole,
+B. D. (2020). Influences on attitudes regarding potential COVID-19 vaccination in the
+
+
+
+24
+United States. Vaccines, 8(4), 582. https://doi.org/10.3390/vaccines8040582.
+Porter, R. M., Amin, A. B., Bednarczyk, R. A., & Omer, S. B. (2018). Cancer-salient messaging
+for human papillomavirus vaccine uptake: A randomized controlled
+trial. Vaccine, 36(18), 2494-2500. https://doi.org/10.1016/j.vaccine.2018.01.040.
+Pritchard, E., Matthews, P. C., Stoesser, N., Eyre, D. W., Gethings, O., Vihta, K., ... & Pouwels,
+ K. (2021). Impact of vaccination on new SARS-CoV-2 infections in the UK. Nature
+Medicine. https://doi.org/10.1038/s41591-021-01410-w
+Reuters. (2021, August 6). Early signs COVID-19 vaccines may not stop Delta transmission,
+England says. https://www.reuters.com/world/uk/england-says-delta-infections-produce-
+similar-virus-levels-regardless-vaccine-2021-08-06/.
+Robertson, E., Reeve, K. S., Niedzwiedz, C. L., Moore, J., Blake, M., Green, M., ... & Benzeval,
+M. J. (2021). Predictors of COVID-19 vaccine hesitancy in the UK household
+longitudinal study. Brain, Behavior, and Immunity, 94, 41-50.
+
+https://doi.org/10.1016/j.bbi.2021.03.008.
+Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution
+process. In Advances in experimental social psychology (Vol. 10, pp. 173-220).
+Academic Press. https://doi.org/10.1016/S0065-2601(08)60357-3.
+Vaccinations in United Kingdom (30 April, 2021). Gov.UK Coronavirus (COVID-10) in the UK.
+Available at: https://coronavirus.data.gov.uk/details/vaccinations (Accessed: 30 April,
+
+2021).
+Walkey, A. J., Law, A., & Bosch, N. A. (2021). Lottery-based incentive in Ohio and COVID-19
+vaccination rates. JAMA, 326(8), 766-767.
+Wenzel, K., Schindler, S., & Reinhard, M. A. (2017). General belief in a just world is positively
+
+
+
+25
+associated with dishonest behavior. Frontiers in psychology, 8, 1770.
+https://doi.org/10.3389/fpsyg.2017.01770.
+White, C. J., Norenzayan, A., & Schaller, M. (2019). The content and correlates of belief in
+Karma across cultures. Personality and Social Psychology Bulletin, 45(8), 1184-1201.
+https://doi.org/10.1177/0146167218808502.
+Yuan, S., & Chu, H. (2022). Vaccine for yourself, your community, or your country? Examining
+audiences’ response to distance framing of COVID-19 vaccine messages. Patient
+Education and Counseling, 105(2), 284-289.
+Zizzo, D. J. (2010). Experimenter demand effects in economic experiments. Experimental
+Economics, 13(1), 75-98. https://doi.org/10.1007/s10683-009-9230-z
+
+
+
+26
+
+1
+
+How Effective are COVID-19 Vaccine Health Messages in Reducing Vaccine Skepticism?
+Heterogeneity in Messages’ Effectiveness by Just-World Beliefs
+
+Tables and Figures
+
+
+
+
+Figure 1. Proportions of overall skepticism changes across control, individual, and community messages.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+Change in standardized vaccine skepticism
+Change in skepticism (SD) from T=0 to T=1
+5
+.05
+0
+Overall skepticism
+Control
+Individual
+Community
+95% CI
+n = 526Table 1: The effects of public health messages on overall vaccine skepticism factor
+outcome: OLS regressions
+
+
+
+
+
+(1)
+(2)
+(3)
+
+∆ Skepticism
+factor (std)
+∆ Skepticism
+factor (std)
+∆ Skepticism
+factor (std)
+
+
+
+
+Individual
+-0.113**
+-0.113**
+-0.189***
+
+(0.0453)
+(0.0457)
+(0.0657)
+Community
+-0.0872
+-0.0872
+-0.101
+
+(0.0502)
+(0.0502)
+(0.0639)
+High BJW
+
+0.000134
+-0.0569
+
+
+(0.0434)
+(0.0658)
+Individual x High BJW
+
+
+0.147
+
+
+
+(0.0911)
+Community x High BJW
+
+
+0.0274
+
+
+
+(0.0969)
+Optimism (std)
+-0.00429
+-0.00432
+-0.00569
+
+(0.0201)
+(0.0215)
+(0.0215)
+Age
+0.00640
+0.00640
+0.00625
+
+(0.0190)
+(0.0191)
+(0.0188)
+Age squared
+-8.56e-05
+-8.56e-05
+-8.20e-05
+
+(0.000293)
+(0.000294)
+(0.000290)
+Female
+-0.000917
+-0.000910
+0.00305
+
+(0.0438)
+(0.0440)
+(0.0436)
+£ 30k+
+-0.0179
+-0.0179
+-0.0153
+
+(0.0410)
+(0.0414)
+(0.0411)
+London
+-0.0313
+-0.0313
+-0.0292
+
+(0.0635)
+(0.0636)
+(0.0633)
+University+
+0.0467
+0.0467
+0.0431
+
+(0.0430)
+(0.0434)
+(0.0429)
+White
+0.0210
+0.0210
+0.0254
+
+(0.0703)
+(0.0705)
+(0.0708)
+Labour
+-0.00769
+-0.00768
+-0.00469
+
+(0.0375)
+(0.0372)
+(0.0378)
+Constant
+-0.0700
+-0.0701
+-0.0499
+
+(0.303)
+(0.307)
+(0.300)
+Cluster individuals
+526
+526
+526
+R-squared
+0.024
+0.024
+0.029
+Note: *** p<0.001, ** p<0.05. Robust standard errors clustered at the individual level and are in parentheses.
+Dependent variables represent the change from #! to #" and are standardized to have a mean of 0 and a standard
+deviation of 1.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+Table 2: The effects of public health messages on individual skepticism outcomes: OLS
+regressions with BJW interactions
+
+
+(1)
+(2)
+(3)
+(4)
+
+∆ Vaccine not
+useful (std)
+∆ Not accept
+vaccine (std)
+∆ Not reduce
+transmission (std)
+∆ Not prevent
+illness (std)
+
+
+
+
+
+Individual
+-0.166
+-0.0212
+-0.0458
+-0.323***
+
+(0.102)
+(0.0466)
+(0.107)
+(0.0917)
+Community
+-0.101
+0.0175
+-0.0904
+-0.139
+
+(0.0951)
+(0.0545)
+(0.110)
+(0.0972)
+High BJW
+-0.0605
+-0.00896
+0.0244
+-0.103
+
+(0.0951)
+(0.0543)
+(0.116)
+(0.113)
+Individual x High BJW
+0.147
+0.0565
+0.0416
+0.214
+
+(0.135)
+(0.0790)
+(0.158)
+(0.146)
+Community x High BJW
+0.0651
+-0.0288
+-0.189
+0.0998
+
+(0.134)
+(0.0827)
+(0.170)
+(0.155)
+Optimism (std)
+-0.0281
+-4.72e-05
+0.00949
+0.0194
+
+(0.0253)
+(0.0188)
+(0.0378)
+(0.0344)
+Age
+-0.00199
+0.0131
+0.0512
+-0.00812
+
+(0.0276)
+(0.0156)
+(0.0311)
+(0.0281)
+Age squared
+5.24e-05
+-0.000208
+-0.000845
+0.000158
+
+(0.000424)
+(0.000227)
+(0.000481)
+(0.000433)
+Female
+0.0213
+-0.0378
+0.0621
+-0.0174
+
+(0.0616)
+(0.0397)
+(0.0760)
+(0.0696)
+£ 30k+
+-0.00347
+0.0365
+-0.0707
+-0.0745
+
+(0.0606)
+(0.0381)
+(0.0706)
+(0.0662)
+London
+-0.00687
+-0.0415
+0.00830
+-0.0467
+
+(0.0989)
+(0.0448)
+(0.0875)
+(0.0791)
+University+
+0.0975
+-0.0117
+-0.0588
+0.0427
+
+(0.0601)
+(0.0326)
+(0.0678)
+(0.0667)
+White
+0.0705
+-0.0778
+0.122
+-0.0191
+
+(0.108)
+(0.0591)
+(0.118)
+(0.115)
+Labour
+-0.0578
+0.0123
+-0.0130
+0.0311
+
+(0.0557)
+(0.0321)
+(0.0687)
+(0.0605)
+Constant
+-0.0239
+-0.114
+-0.716
+0.284
+
+(0.458)
+(0.259)
+(0.498)
+(0.444)
+Cluster individuals
+526
+526
+526
+526
+R-squared
+0.023
+0.021
+0.032
+0.034
+Note: *** p<0.001, ** p<0.05. Robust standard errors clustered at the individual level and are in parentheses.
+Dependent variables represent the change from #! to #" and are standardized to have a mean of 0 and a standard
+deviation of 1.
+
+
+
+
+
+
+Figure 2: Predictive margins of the individual treatment and control group, over the
+standardized BJW factor
+
+
+
+
+Predictive Margins of treat with 95% Cls
+2
+Linear Prediction
+0
+2
+-4
+-2
+0
+2
+4
+Standardized BJW Factor
+Control
+IndividualFigure 3: Predictive margins of the community treatment and control group, over the
+standardized BJW factor
+
+
+
+
+
+
+Predictive Margins of treat with 95% Cls
+Linear Prediction
+2
+0
+2
+-4
+-2
+0
+2
+4
+Standardized BJW Factor
+Control
+Community 1
+How Effective are COVID-19 Vaccine Health Messages in Reducing Vaccine Skepticism?
+Heterogeneity in Messages’ Effectiveness by Just-World Beliefs
+
+Appendix
+
+
+
+ 2
+Table A.1: Balance checks on all observable characteristics amongst the final analysis sample.
+
+
+Control
+(0)
+Individual
+(1)
+Community
+(2)
+(0) vs. (1),
+p-value
+(0) vs. (2),
+p-value
+(1) vs. (2),
+p-value
+𝑇!(baseline)
+Not vaccine useful
+17.0
+16.2
+17.3
+0.704
+0.904
+0.616
+
+(1.5)
+(1.4)
+(1.6)
+
+
+
+
+172
+180
+174
+
+
+
+Not accept vaccine
+12.5
+12.6
+13.9
+0.988
+0.628
+0.625
+
+(1.9)
+(1.7)
+(2.0)
+
+
+
+
+172
+180
+174
+
+
+
+Not reduce transmission
+29.9
+28.7
+29.0
+0.645
+0.670
+0.983
+
+(1.9)
+(1.9)
+(2.0)
+
+
+
+
+172
+180
+174
+
+
+
+Not prevent illness
+22.0
+22.3
+22.8
+0.895
+0.741
+0.839
+
+(1.8)
+(1.7)
+(1.8)
+
+
+
+
+172
+180
+174
+
+
+
+𝑇" (endline)
+Not vaccine useful
+15.6
+13.5
+14.5
+0.253
+0.615
+0.566
+
+(1.4)
+(1.2)
+(1.5)
+
+
+
+
+172
+180
+174
+
+
+
+Not accept vaccine
+11.6
+12.0
+13.0
+0.892
+0.602
+0.684
+
+(1.8)
+(1.7)
+(1.8)
+
+
+
+
+172
+180
+174
+
+
+
+Not reduce transmission
+30.7
+29.0
+25.0
+0.548
+0.039
+0.124
+
+(2.0)
+(1.9)
+(1.8)
+
+
+
+
+172
+180
+174
+
+
+
+Not prevent illness
+22.1
+17.8
+20.8
+0.054
+0.581
+0.187
+
+(1.7)
+(1.5)
+(1.7)
+
+
+
+
+172
+180
+174
+
+
+
+
+Quartile 1 BJW
+0.3
+0.2
+0.3
+0.626
+0.671
+0.358
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+Quartile 4 BJW
+0.2
+0.3
+0.3
+0.529
+0.352
+0.756
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+Optimism factor (Std)
+0.0
+0.0
+-0.0
+0.928
+0.779
+0.850
+
+
+(0.1)
+(0.1)
+(0.1)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+Age
+30.4
+31.0
+31.5
+0.479
+0.231
+0.600
+
+
+(0.7)
+(0.6)
+(0.7)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+Female
+0.7
+0.8
+0.7
+0.309
+0.459
+0.781
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+166
+171
+170
+
+
+
+
+£30,000+
+0.7
+0.7
+0.8
+0.987
+0.455
+0.423
+
+
+(0.1)
+(0.0)
+(0.1)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+London
+0.1
+0.1
+0.2
+0.652
+0.248
+0.472
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+Undergraduate+
+0.6
+0.6
+0.6
+0.778
+0.808
+0.597
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+171
+179
+174
+
+
+
+
+White
+0.9
+0.9
+0.9
+0.525
+0.724
+0.779
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+172
+180
+174
+
+
+
+
+Labour party
+0.3
+0.4
+0.4
+0.825
+0.438
+0.578
+
+ 3
+
+
+(0.0)
+(0.0)
+(0.0)
+
+
+
+
+
+169
+172
+172
+
+
+
+Note: standard deviations in parenthesis, sample size of respondents in italics.
+
+
+
+
+
+ 4
+Figure A.1: Experimental process and consort diagram
+
+
+
+
+
+
+T0
+600participants
+Completed BJW scale and baseline
+skepticism outcomes
+T1
+600 participants
+Invited to return for part 2
+Control
+Individual treatmentCollective treatment
+172 participants
+181 participants
+174 participants
+Read article about
+Read article about
+individual benefits
+communitybenefits
+to vaccination
+tovaccination
+Passed attention
+Passed attention
+checks
+checks
+180 participants
+174 participants
+526 participants
+Completed endline skepticism outcomes
+and demographic questions 5
+Figure A.2: Survey design: questions at 𝑇!.
+
+
+
+
+
+Please read each statement carefully and indicate the extent to which you personally
+agree or disagree with it.
+Very
+Very
+strongly
+Slightly
+Slightly
+strongly
+disagree
+Disagree
+disagree
+agree
+Agree
+agree
+I think basically the
+0
+0
+0
+0
+0
+0
+world is a justplace.
+I believe that, by and
+large, people get what
+0
+0
+0
+0
+0
+0
+they deserve.
+Iam confidentthat
+justice always prevails
+0
+0
+0
+0
+0
+over injustice.
+I am convinced that in
+the long run, people
+0
+will be compensated
+for injustices.
+I firmly believe that
+injustices inallareas
+of life (e.g.
+professional, family,
+0
+0
+0
+politics) are the
+exception rather than
+the rule.
+I think people try to be
+fair when making
+0
+0
+0
+important decisions. 6
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+Please read each statement carefully and indicate the extent to which you personally
+agree ordisagree with it.
+Very
+Very
+strongly
+Slightly
+Slightly
+strongly
+disagree
+Disagree
+disagree
+agree
+Agree
+agree
+I believe that, by and
+large, I deserve what
+0
+0
+0
+0
+0
+0
+happens to me.
+I am usually treated
+0
+0
+0
+0
+0
+0
+fairly.
+I believe that I usually
+0
+0
+0
+0
+0
+0
+get what I deserve.
+Overall, events in my
+0
+0
+0
+0
+0
+0
+life are just.
+In my life injustice is
+theexceptionrather
+0
+0
+0
+0
+0
+0
+than the rule.
+I believe that most of
+0
+0
+0
+0
+0
+0
+the things that happen
+in my life are fair.
+I think that important
+decisionsthatare
+0
+made concerning me
+are usually just. 7
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+How certainareyouthattheCOViD-19vaccinesare ausefultool infightingthepandemic?
+Not at all certain
+Extremelycertain
+0
+10
+20
+30
+40
+50
+60
+70
+80
+90
+100
+How likelyareyouto accept the CovID-19vaccinewhen offered?
+Not at all likely
+Extremely likely
+0
+10
+20
+30
+40
+50
+60
+70
+80
+90
+100
+HowcertainareyouthattheCOViD-19vaccinereducestransmissionbetweenindividuals?
+0
+10
+20
+30
+40
+50
+60
+70
+80
+90
+100
+How certainareyouthattheCoviD-19vaccinewouldpreventyoupersonallyfromgettingveryill dueto
+COVID-19?
+0
+10
+20
+30
+40
+50
+60
+70
+80
+90
+100 8
+Figure A.3.1: Survey design: treatment messages at 𝑇". Control participants were asked to
+respond to the same four skepticism outcomes shown in Figure A.2. Individual (left) and
+community (right) messages participants were first asked to read the following fictitious news
+articles and were then prompted to respond to the four skepticism outcomes.
+
+
+
+
+
+
+Belowisanewsstorysimilartoothernewsstoriesyoumighthavereadbefore.Please
+readthestoryandrespondtothequestionsthatfollow.
+BOsTON --"Coronavirus disease (COVID-19)is a highly contagious illness, caused bythe
+transmission of the SARS-CoV-2 virus. First identified in December 2019, the virus has
+causedapandemicthatresulted inshutdownsall aroundtheglobe.It wasfirst widely
+haswreakedhavocontheglobe.Claimingmillionsoflives,thispandemichascreateda
+cleardemarcationintime:pre-covid,andpost-covid,"says ProfessorArthurMichali,a
+publichealthexpertfromaleadingresearchuniversity."Beforethispandemic,Tcouldhave
+attendedaconferenceinTokyo oneday,ledaresearchcollaboration inGenevathenext,
+andarrivedback inBostonthethirdday.Thiskindoftravel issimplynolongerpossible
+under currentcircumstances,andit's likelythatthis sort ofbehaviourcontributedtothe
+rapidspreadofthediseaseworldwide."
+ProfessorMichali,whohaswonnumerousawardsforhis researchoverthelasttwo
+decades,ispartoftheCOViD-19EmergencyCommitteeattheWorldHealthOrganisation.
+Amongstothertopics,thiscommitteeisworkingtobetterunderstandthevarious
+responses and interventions that can help curb the spread of the disease.
+Michali is co-authoring a forthcoming pamphlet, entitled"The COViD-19Vaccine:what
+circulatingandforthcomingvaccines.ThepamphletdescribesCOviD-19as"adangerous
+disease,particularlyforthe elderlyand clinicallyvulnerable,astheyare more likelyto
+suffersevere,andpossiblyfatal,respiratoryillness.Nevertheless,anyone,regardlessof
+ageormedical background,isatriskof sufferingaharshillness.Michaliwishesto
+emphasisethatthebestthingyoucandotoprotectyourselffromthisdiseaseisto
+takeupthevaccinewhenyouareofferedit."Some ofthevaccinesonthemarketare
+boasting95%efficacyrates.Thismeansthatreceivingthevaccinedramaticallyreduces
+yourriskofdevelopingseriousCOViD-19symptomsifyouareexposedtotheviruslater
+downtheline."Althoughexpertsarecontinuingtoemphasisetheimportanceofsocial
+distancingandwearingmasks,thesemeasuresare notperfect,andthereremainsariskof
+inadvertentlycatchingthediseasethatcouldleaveyoubed-riddenforweeks,even
+months.Receivingthe vaccine isthesingle most important stepan individualcantaketo
+protect him orherself fromthe virus.Michali reflects intheconcludingthoughts of the
+pamphlet,"thereisnotmuchthatwecancontrolintimeslikethese,butyouneedto
+do whatyoucantoprotectyourself inthese uncertain times.Takingup thevaccine
+whenoffered isthebestactionyoucantaketokeepyourself safe!"Importantly,
+Michali wants individuals to rememberthat it is their personal responsibility to keep
+themselvesprotected.Below isanewsstorysimilartoothernewsstoriesyoumighthavereadbefore.Please
+readthestoryand respondtothequestionsthatfollow.
+BOsTON--“Coronavirus disease (COVID-19)is a highly contagious illness, caused bythe
+transmission oftheSARS-CoV-2 virus.First identified inDecember2019,the virushas
+caused apandemic that resulted in shutdowns all aroundtheglobe.Itwasfirst widely
+spreadbetweenhumansatawholesaleseafoodmarket inWuhan,China."Thisdisease
+haswreakedhavocontheglobe.Claimingmillionsof lives,thispandemichascreateda
+cleardemarcation intime:pre-covid,and post-covid,says ProfessorArthurMichali,a
+publichealthexpertfromaleadingresearchuniversity."Beforethispandemic,Icouldhave
+attendedaconference inTokyooneday,ledaresearchcollaboration inGenevathenext,
+andarrivedbackinBostonthethirdday.Thiskindoftravelissimplynolongerpossible
+undercurrentcircumstances,andit's likelythatthissortof behaviourcontributedtothe
+rapidspreadofthediseaseworldwide."
+ProfessorMichali,whohaswonnumerousawardsforhisresearchoverthelasttwo
+decades, is part of the COviD-19 Emergency Committee at the World Health Organisation.
+Amongstothertopics,thiscommitteeisworkingtobetterunderstandthevarious
+Michali is co-authoring a forthcoming pamphlet, entitled “"The COviD-19 Vaccine: what
+circulatingandforthcomingvaccines.ThepamphletdescribesCOviD-19as"adangerous
+disease,particularly forthe elderly and clinically vulnerable,as they are more likely to
+suffersevere,andpossiblyfatal,respiratoryillness.Nevertheless,anyone,regardlessof
+ageormedical background,is atrisk ofsufferingaharsh illness."Michali wishesto
+emphasisethatthebestthingyoucandotoprotectothersfromthisdiseaseisto
+takeupthevaccinewhenyouareoffered it."Someofthevaccinesonthemarketare
+boasting95%efficacyrates.Thismeansthatreceivingthevaccinedramaticallyreduces
+yourriskofdevelopingseriousCOviD-19symptomsifyouareexposedtotheviruslater
+downtheline.CommunitytransmissionhasbeenshowntobelowerwhensevereCOviD
+19symptomsdonotpresent,soyouareprotectingyourneighbours,parents,
+grandparents,andfriendsbyreceivingthevaccine."Althoughexpertsarecontinuingto
+emphasisetheimportanceofsocialdistancingandwearingmasks,thesemeasuresare
+notperfect,astheviruscanstillspreadbetweenpeople.Theworryisnotsomuchabout
+individual cases, but rather, it is about reducing transmission in communities, as it is that
+type oftransmissionthat will preventus from everseeing anendtothispandemic.
+thecommunityfromthevirus.Michali reflectsintheconcludingthoughtsofthepamphlet,
+"thereisnotmuchthatwe can control intimes likethese,butweneedtotake
+collectiveactiontofightthispandemic!Takingupthevaccinewhenofferedisthe
+bestactionyoucantakeforyourfamily,friends,andforyourcommunity!
+Importantly,Michaliwantspeopletorememberthatitistheirresponsibilitytokeeppeople
+intheircommunity,especiallythosewhoarevulnerabletothedisease,protected 9
+Figure A.3.2: Survey design : manipulation check at 𝑇".
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+According to the article, where was the COVID-19 virus first widely spread?
+Geneva,Switzerland
+Boston, USA
+Wuhan, China
+Tokyo, Japan
+Accordingtothearticle,whatisProfessorArthurMichali currentlyworkingon?
+A strategy to liaise with journalists and media about COVID-19
+Apamphletto informthe average citizenaboutthe currentandforthcoming COviD-19vaccines
+Atravel itineraryfromTokyotoGenevato Boston
+Asociologicalstudyonthespreadof COviD-19
+According to the article, whom will you primarily protect by taking up a CoOVID-19 vaccine?
+Yourself
+Healthworkersinothercountries
+Peoplewho have justdiedofCOvID-19-related illness
+Others in your community 10
+Figure A.3.3: Survey design: demographic questions at 𝑇".
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+Areyou generally aperson who tries to avoid taking risks orare youfully prepared to take
+risks?
+Won't take risks
+Ready to take risks
+0
+1
+2
+3
+4
+5
+6
+7
+8
+6
+10
+HaveyoubeendiagnosedwithCovID-19atanypoint?
+Howfrequentlydoyoufollowgovernmentguidelinesonfacecoveringswheninshops?
+I neverwearafacecoveringbecauseIamexemptfromwearingone
+I neverwearafacecoveringand I amnotexemptfromwearingone.
+Most of the time I do not wear a face covering.
+HalfofthetimeIwearafacecovering,halfI donot.
+Most ofthe time I wearaface covering.
+Ialwayswearafacecovering
+How confident are you thatface coverings area useful tool in fighting the pandemic?
+Not at all confident
+Extremelyconfident
+0
+10
+20
+30
+40
+50
+60
+70
+80
+90
+100What isyourethnicity?
+What is the highest level of education that you have completed?
+Inwhichregiondoyoucurrentlyreside?
+What is youryearlyhousehold incomebeforetax?
+V
+Whichpolitical party do you consideryourself to beclosest to?
+Please indicateyourattitudesto eachof thefollowingstatements.
+I neither
+I disagree a
+I disagree a
+agree nor
+Iagree a
+lot
+little
+disagree
+little
+I agree a lot
+In uncertain times,
+usuallyexpectthe
+0
+0
+0
+0
+0
+best.
+I'm always optimistic
+0
+0
+0
+0
+0
+aboutmy future.
+Overall,Iexpectmore
+good things to happen
+0
+0
+0
+0
+0
+to me than bad. 11
+Table A.2: Factor analysis on the personal and general BJW factors, which produce the
+combined BJW factor.
+
+Factor analysis/correlation
+
+
+Factor
+Eigenvalue
+Difference
+Proportion
+Cumulative
+Factor1
+ 0.79 1.04
+ 1.46
+1.46
+
+Factor loadings (pattern matrix) and unique variances
+Variable
+Factor1
+Uniqueness
+General BJW
+0.63
+0.61
+Personal BJW
+0.63
+0.61
+
+
+Scoring coefficients
+Variable
+Factor1
+General BJW
+0.41
+Personal BJW
+0.41
+
+Cronbach’s alpha
+a
+0.69
+
+
+Table A.3: Factor analysis on the skepticism outcomes at 𝑇!.
+Factor analysis/correlation
+
+
+Factor
+Eigenvalue
+Difference
+Proportion
+Cumulative
+Factor1
+2.57
+2.60
+1.10
+1.10
+
+Factor loadings (pattern matrix) and unique variances
+Variable
+Factor1
+Uniqueness
+Vaccine Useful
+0.86
+0.26
+Accept Vaccine
+0.83
+0.31
+Reduce Transmission
+0.64
+0.59
+Prevent Illness
+0.85
+0.28
+
+
+Scoring coefficients
+Variable
+Factor1
+Vaccine Useful
+0.35
+Accept Vaccine
+0.28
+Reduce Transmission
+0.12
+Prevent Illness
+0.31
+
+
+
+ 12
+Cronbach’s alpha
+a
+0.88
+
+
+Table A.4: Factor analysis on the skepticism outcomes at 𝑇".
+
+Factor analysis/correlation
+
+
+Factor
+Eigenvalue
+Difference
+Proportion
+Cumulative
+Factor1
+2.69
+2.70
+1.09
+1.09
+
+Factor loadings (pattern matrix) and unique variances
+Variable
+Factor1
+Uniqueness
+Vaccine Useful
+0.88
+ 0.22
+Accept Vaccine
+0.82
+ 0.33
+Reduce Transmission
+0.70
+ 0.51
+Prevent Illness
+0.87
+0.25
+
+
+Scoring coefficients
+Variable
+Factor1
+Vaccine Useful
+0.37
+Accept Vaccine
+ 0.23
+Reduce Transmission
+0.13
+Prevent Illness
+0.32
+
+
+Cronbach’s alpha
+a
+0.89
+
+
+
+
diff --git a/6tE1T4oBgHgl3EQfnAQs/content/tmp_files/load_file.txt b/6tE1T4oBgHgl3EQfnAQs/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..868fbf46c2a7b3a820c71a835eb4acb3cd8cee03
--- /dev/null
+++ b/6tE1T4oBgHgl3EQfnAQs/content/tmp_files/load_file.txt
@@ -0,0 +1,1363 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf,len=1362
+page_content='1 How Effective are COVID-19 Vaccine Health Messages in Reducing Vaccine Skepticism?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Heterogeneity in Messages’ Effectiveness by Just-World Beliefs Juliane Wiese, corresponding author Nattavudh Powdthavee Warwick Business School Nanyang Technological University University of Warwick 50 Nanyang Avenue Scarman Road 639798 Singapore Coventry CV4 7AL United Kingdom ORCID ID : 0000 0002 4314 5934 ORCID ID: 0000 0002 9345 4882 juliane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='wiese@warwick.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='uk +33 6 68 88 18 27 Declarations of interest: none.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Abstract To end the COVID-19 pandemic, policymakers have relied on various public health messages to boost vaccine take-up rates amongst people across wide political spectra, backgrounds, and worldviews.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' However, much less is understood about whether these messages affect different people in the same way.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' One source of heterogeneity is the belief in a just world (BJW), which is the belief that in general, good things happen to good people, and bad things happen to bad people.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This study investigates the effectiveness of two common messages of the COVID-19 pandemic: vaccinate to protect yourself and vaccinate to protect others in your community.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We then examine whether BJW moderates the effectiveness of these messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We hypothesize that just-world believers react negatively to the prosocial pro-vaccine message, as it charges individuals with the responsibility to care for others around them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Using an unvaccinated sample of UK residents before vaccines were made widely available (N=526), we demonstrate that the individual-focused message significantly reduces overall vaccine skepticism, and that this effect is more robust for 2 individuals with a low BJW, whereas the community-focused message does not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Our findings highlight the importance of individual differences in the reception of public health messages to reduce COVID-19 vaccine skepticism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Keywords: vaccine skepticism;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' health messages;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' justice beliefs;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' individual differences;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' COVID- 19 3 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Introduction Before the vaccine rollout in the UK, 28% of the British population, particularly those in Black and South Asian minority ethnic groups, were skeptical about getting vaccinated (Robertson et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' To maximize vaccine take-up, governments have been delivering simple messages that emphasize people’s responsibility to themselves and the community.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' For example, the National Health Services in the UK urges the public to “join the millions already vaccinated, to protect yourself and others” (NHS UK, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' These foci, given their central role in public health messaging during the COVID-19 pandemic so far, have shaped the two themes of messages examined in this study: individual and community responsibilities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Despite the extensive literature on the framing approaches of public health messages around vaccines (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Gallagher & Updegraff, 2012;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' McPhee et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2003;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Kelly & Kornik, 2016), the overall effectiveness of COVID-19 vaccine messages on individual or community responsibility is currently imperfectly understood.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While recent evidence suggests that individual-focused messages more effectively increase vaccine uptake and support for mandates than community- focused messages, these effects are heterogeneous across individualistic and communitarian worldviews (Yuan & Chu, 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Furthermore, we do not know which underlying beliefs about the vaccine are best addressed by these messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Nevertheless, they continue to be used by public health officials worldwide.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In contexts of extreme urgency, who are the types of people who might respond poorly to these messages and experience stronger vaccine skepticism?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We build our investigation around the strong theoretical link between belief in a just world (BJW) and vaccine skepticism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Just-world 4 believers conceive a universal justice structure which holds that both normatively and positively speaking, good things tend to happen to good people and vice versa (Furnham, 2003).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This adaptive function (Dalbert, 2009), manifesting at varying levels of intensity and therefore influencing a large portion of the population (White et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2019), allows individuals to rationalize negative consequences in the world as justified, predictable, and manageable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Doing so promotes well-being and a sense of stability in the world (Correia et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2009;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Jiang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In the context of the COVID-19 pandemic, where an unprecedented public health emergency and sweeping government regulations significantly reduced individual freedoms, just-world believers struggled to make sense of such undeserved restrictions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This sense of unfairness fosters a resistance against the government-promoted solution to the problem: specifically, a vaccine that has been developed in record speed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Suggestive evidence of this link between just-world believers and anti-vaxxers is demonstrated by their numerous shared psychological traits, including conspiracy thinking (Nestik et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Jolley & Douglas, 2014) and individualistic attitudes (Wenzel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2017;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Motta et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Government-sponsored pro-vaccine messages, particularly ones that focus on the responsibility we hold to our communities, are therefore likely to threaten the just-world believers’ worldview, as their personal role in the pandemic is limited, and others’ health outcomes are independent of their own decision to get vaccinated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Their worldview threatened, just-world believers defensively dismiss the message that threatens their BJW, and deny the existence of a problem in the first place (Furnham, 2003).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This study makes two main contributions to the literature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' First, we experimentally investigate the effectiveness of two commonly used pro-vaccine messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Second, we examine whether BJW 5 moderates the effectiveness of each message.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Given policymakers’ priority to increase COVID-19 vaccine uptake, understanding individual differences in the messages’ effectiveness by BJW is critical to understanding the potential threats to their overall effectiveness on the entire population.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Existing literature and hypotheses Before the vaccine rollout, researchers’ main concern was whether the COVID-19 vaccines safely reduce illness and transmissibility.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Having established this (Katella, 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Pritchard et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021), vaccine uptake has emerged as a more enduring challenge for public health officials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A nationally representative survey of 316 Americans shows that demonstrating its efficacy and emphasizing the costs of the pandemic encourages vaccine uptake (Pogue et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' However, their survey did not engage with messages that focus on the simple facts that give value to the vaccine: that it protects its recipients and their community.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' These facts have been central to policymakers’ messaging during the COVID-19 pandemic, and there continues to be little empirical investigation into their effectiveness in shifting public perception around the vaccine’s effectiveness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The decision to vaccinate weighs the benefits against the risks of vaccination, which could range from fears of side effects and needles to mistrust of healthcare authorities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Previous research demonstrates the importance of highlighting vaccines’ protective benefits, as doing so can crowd out concerns about risks (Porter et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Similarly, a COVID-19 vaccine message highlighting the vaccine’s protective benefits to the individual has been shown to increase intended vaccine uptake (Yuan & Chu, 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Our work examines how such an individualistic message can drive the underlying beliefs around the vaccine’s protective function to its recipients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 6 In addition, researchers have found prosocial vaccine messages to have a positive impact on vaccination rates (Betsch et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2017;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Betsch & Böhm, 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' McPhee et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2003).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' For example, messages that emphasize the benefits of an avian flu vaccine to others significantly increase vaccination intentions, compared to messages which emphasize its benefits to the individual (Kelly & Hornick, 2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While these findings link the community-oriented message to increased vaccination intentions, they do not examine how such a message impacts beliefs around transmission rates, which is the mechanism that connects the prosocial messages with increased vaccine uptake.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We aim to show experimentally that prosocial messages increase confidence in the underlying belief that the vaccines reduce transmission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Based on this evidence, we predict that the individual message will more effectively decrease overall skepticism than the community message, and that this effect is driven by the fact that the individual message shifts the underlying belief that the vaccine protects its recipients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The prosocial messages will more moderately increase confidence that the vaccine reduces transmission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Despite the predicted overall success of the two messages, the question remains around heterogeneous effects, specifically around moral worldviews that play a role in the decision to vaccinate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While Devereux et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021) discover a link between stronger BJW and a greater likelihood to adhere to COVID-19 measures, such as social distancing, these measures come at essentially zero risk, resulting in a very different cost-benefit analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In contrast, accepting a vaccine requires accepting the risk of potential negative side-effects, and might therefore have a different relationship with BJW.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 7 Demographic factors (Peretti-Watel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Khubchandani et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021), psychological traits (Browne et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2015;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Jolley & Douglas, 2014), and beliefs about vaccine safety (Karlsson et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021) predict vaccine attitudes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' However, studies that examine how such traits, like BJW, interfere with public health messages are scarce.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While recent evidence has shown that people with more individualistic, rather than communitarian, values respond more favorably to individual-centered COVID-19 vaccine messages (Yuan & Chu, 2022), it remains unclear how such worldviews moderate individuals’ understanding of the many ways in which the vaccine protects the public.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Furthermore, rather than simply capturing individualistic or community-oriented worldviews, BJW contains a deeper moral around one’s deservingness of one’s place in the world, telling us more about the reasoning behind an individual’s action (or inaction).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While people who see public health as a moral issue tend to consider prosocial (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' self-centered) social distancing messages more persuasive (Luttrell & Petty, 2020), BJW is not an altruistic moral belief system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Instead, it holds individuals responsible for their own fate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' BJW inherently commits fundamental attribution error, in which individuals place more weight on dispositional, as opposed to environmental or situational, factors (Ross, 1977).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' By further emphasising societal responsibility as a motive to get vaccinated, public health officials transfer the responsibility for a COVID patient’s health onto the community’s vaccination decision-making.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This clashes with the tendency of just-world believers to blame patients for their own misfortunes and to separate the consequences of their own actions from the outcomes of others (Lerner & Simmons, 1966;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Lucas et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2009).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Therefore, by asking people to take responsibility for others’ health and safety during the COVID-19 pandemic, policymakers inevitably challenge the justice structure of the world in 8 which individuals are responsible for their own fate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In response, just-world believers might discredit the vaccine altogether.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We therefore hypothesise that for individuals with a strong BJW, the prosocial messages are less effective at reducing vaccine skepticism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Method 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 Data In this pre-registered experiment (tinyurl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='com/bxv23), 600 UK-based Prolific (www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='prolific.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='co) users aged between 18 and 49 joined a longitudinal online study on attitudes towards COVID-19 and vaccination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' At the time, the UK general public under 50 years of age was not yet eligible to receive a COVID-19 vaccine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Just over a quarter of the UK population had received its first dose, and only 1% of the population had received both doses (Vaccinations in United Kingdom, 30 April, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Part one of the study (𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=') took place on 24 February 2021, and part two (𝑇") on 1 March 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We collected data at two points in time to reduce the likelihood that (i) participants suspect the study purpose and bias their responses, and (ii) participants’ responses to vaccine skepticism questions are biased by exposure to questions around justice beliefs (Zizzo, 2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Participants gave informed consent and were compensated £0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='25 at 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' and £1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00 at 𝑇".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 527 participants (88%) remained at 𝑇" and were randomised evenly across Control, Individual- Treatment, and Prosocial-Treatment (N = 172, 181, and 174, respectively).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Only one participant failed all three attention checks and was removed from the sample, resulting in 526 participants with complete longitudinal data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This sample size (i) allowed sufficient power for a reasonable 9 minimal detectable effect size and (ii) is slightly larger than what was used in a similar research design studying BJW and climate change messaging (Feinberg & Willer, 2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Of the final sample of 526 individuals, 70% were females, 87% were ethnically White, and 59% have an annual income of £30,000 or over.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The mean age was 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Balance checks confirm that our sample was balanced on observable characteristics across all groups;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' see Table A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 in the appendix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 Measures and procedure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 BJW scales Because vaccination evokes concepts of justice both for the individual and for society, participants completed the general BJW scale, six questions about the justice structure in the world in general (Dalbert et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 1987), and the personal BJW scale, seven questions which posit that the world is just for me personally but not for others (Dalbert, 1999) at 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='. The two scales have a correlation coefficient of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' To attain a linear combination of BJW factors, we conducted a separate factor analysis on each scale, yielding two distinct factors (a = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='78 for general BJW and a = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='88 for personal BJW), and then conducted a factor analysis on these factors, resulting in a combined BJW factor (a = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='68);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' the factor analysis results are in Table A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The resulting combined BJW factor was standardized to a mean of 0 and a standard deviation of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' It was transformed into a dummy variable which marks above- or below-median strength of BJW.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This allows us to investigate the differential effects of the treatments on vaccine skepticism by the strength of BJW.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 Vaccine skepticism 10 At 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' and 𝑇", participants completed four questions on COVID-19 vaccine skepticism, with possible answers ranging from 0 (not at all certain/likely) to 100 (extremely certain/likely).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The precise wording of the questions was: • “How certain are you that the COVID-19 vaccines are a useful tool in fighting the pandemic?”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' • “How likely are you to accept the COVID-19 vaccine when offered?”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' • “How certain are you that the COVID-19 vaccine reduces transmission between individuals?”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' • “How certain are you that the COVID-19 vaccine would prevent you personally from getting very ill due to COVID-19?”' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' For simplicity, we reversed the responses so that higher values represent higher levels of vaccine skepticism in each of the four outcomes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The baseline mean responses are 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8, 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0, 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1, and 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4, respectively, which suggest that at 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', the study population was relatively prepared to take the vaccine but was more skeptical of its illness and transmission prevention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' These outcomes are moderately correlated, with correlations ranging from 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='53 to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' To circumvent the multiple comparisons problem, we also derived an overall skepticism outcome by conducting a factor analysis on the four reversed individual skepticism variables for both outcomes at 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (a = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='87) and 𝑇" (a = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='89);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' see Tables A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4 in the appendix for the estimates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' All skepticism variables were standardized to have a mean of 0 and a standard deviation of 1 and were included in analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' At 𝑇", 5 days after 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', participants were randomised into one of three groups: control (no article), individual, and community responsibility treatment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In both treatments, participants were asked to 11 read a news-style article.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The articles, Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 in the appendix, are identical in the first paragraphs, which discuss the context of the pandemic and vaccine development at the time of writing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' They deviate towards the end by treatment group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The individual responsibility article explains that the vaccine reduces the risk of severe COVID-19 illness to vaccine recipients, and the prosocial article explains that to combat the virus, individuals must accept the vaccine to reduce community transmission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 Attention and manipulation check Participants in the treatment groups were asked two fact-based questions from the article, as well as whether taking the recommended steps during the pandemic will mainly protect them, or mainly protect others, from COVID-19 illness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Amongst the final sample of participants who passed all three attention checks, we find a significant difference between the two treatments on the manipulation-check item, t(352) = 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='64, p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000 for indicating that the vaccine protects yourself, and t(352) = -13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='07, p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000 for indicating that the vaccine protects others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4 Sociodemographic controls Participants also completed a post-experiment questionnaire, which elicited their ethnicity, education level, region, income, political views, optimism, risk attitudes, COVID-19 history, and adhesion to government guidelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Age and gender were collected automatically by Prolific.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 shows the procedural flow of the experiment and consort diagram, and Figures A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 present screenshots of the materials used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 12 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 Analysis We conduct all analyses of vaccine skepticism using Ordinary Least Squares (OLS) regression with robust standard errors clustered on the participant-level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Our primary analysis examines the treatment effects on the overall vaccine skepticism factor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We regress Equation (1) and present the results in column 3 of Table 1: ∆𝑆#$ = 𝑎 + 𝛽"𝑇# + 𝑋# %𝛾 + 𝛽&𝐵𝐽𝑊# + 𝛽\'(𝑇# × 𝐵𝐽𝑊#) + 𝑒, (1) where 𝑖 = 1, … , 𝑁;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 𝑡 = 1, … ,2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' ∆𝑆#$ represents the change in the overall vaccine skepticism factor from t=0 to t=1, where a higher value represents greater vaccine skepticism;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Ti represents the treatment condition (control, individual, or community message) and 𝛽" is the effect of this condition on skepticism;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 𝑋# % represents the matrix of covariates, including a standardized optimism factor (a=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8134), age, age-squared, gender dummy, £30,000+ annual income (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' below £30,000 annual income) dummy, London (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' non-London) dummy, undergraduate education (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' non- undergraduate education) dummy, white (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' non-white) dummy, Labour party (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' non-Labour) dummy;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 𝛽& is the effect of holding a strong (vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' weak) BJW;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=" 𝛽' represents the interaction of treatment and BJW, i." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' the differential effect of the treatment when participants have either a stronger or a weaker BJW;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' and e is the error term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Columns 1 and 2 model the parsimonious specifications of Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1), with covariates excluding and including BJW, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Table 2 models the effects of the interaction between treatment and BJW on each of the four skepticism outcomes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Their forms are identical to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1), with the exception that the outcome variable is replaced by each of the four vaccine skepticism subscales, standardized to mean of 0 and standard deviation of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 13 ∆𝑆()$ ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' = 𝑎 + 𝛽"𝑇# + 𝑋# %𝛾 + 𝛽&𝐵𝐽𝑊# + 𝛽\'(𝑇# × 𝐵𝐽𝑊#) + 𝑒, (2) where 𝑖 = 1, … , 𝑁;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 𝑗 = 1, … ,4;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 𝑡 = 1, … ,2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Here, ∆𝑆() ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' = 𝑆()$ ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' − 𝑆()$*" ?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' , where the notation j represents different domains of beliefs, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 𝑆"# represents the belief that the vaccine is not useful;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 𝑆 represents the likelihood of not accepting the vaccine;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=" 𝑆'# represents the belief that the vaccine will not reduce transmission;" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' and 𝑆+# represents the belief that the vaccine will not prevent serious illness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The rest of the specification is identical to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Note that we deviate from the pre-registered document in two respects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' First, we include an overall skepticism factor as an outcome variable in our primary analysis, circumventing the multiple comparisons problem in our primary analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Second, we run OLS regressions with standard errors clustered at the participant level as the primary analysis rather than using analysis of variance (ANOVA).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This change is made due to the inclusion of continuous independent variables in the regression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Results 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 Message effectiveness We begin by examining the within-person changes in vaccine skepticism by treatment group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' As predicted, Figure 1 shows that the individual message significantly reduces overall skepticism by 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='04 standard deviation, compared to the control group which increases overall skepticism by 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='07 standard deviation (Wilcoxon signed-rank test, p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='030).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' There is weaker evidence that the community message also reduces overall skepticism, which decreased by 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='02 standard deviation (Wilcoxon signed-rank test, p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='103).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Figure 1 thus provides raw data evidence that individual- focused public health message is most effective at reducing overall vaccine skepticism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 14 [Figure 1 here] To understand this result more thoroughly, Table 1 estimates regression equations that adjust for other covariates, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' We find the regression results to be consistent with Figure 1’s findings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The individual-focused message decreases overall skepticism more robustly than the community message, b = -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='11, [95% C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' : -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='20, -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='02], p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='014, versus b = -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='09, [95% C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' : - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='19, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='01], p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='083, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' [Table 1 here] 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 BJW as a moderator of pro-vaccine message impacts To formally test for the heterogeneous effect of public health messages by BJW, Tables 1 and 2 include the interaction terms between treatment and a high BJW dummy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Column 3 of Table 1 shows that for people with a low BJW, the individual message is extremely effective at lowering their overall skepticism factor, b = - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='19, [95% C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' : - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='32, -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='06], p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' As discussed earlier, columns 1 and 2 of Table 1 demonstrate a greater effectiveness of the individual message on average.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The results of column 3 suggest that the effectiveness of this individualistic message is more robust for people with a low BJW, whereas we see no such differential effect for the collective message.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Figures 2 and 3 show this distinction visually, with the predictive margins plots of the control and individual treatment overlapping (Figure 2), and the predictive margins plots of the control and collective treatment (Figure 3) not overlapping.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' When examining the interaction regressions for each sub-scale of vaccine skepticism (Table 2), we find that the strong effect of the individual treatment on overall skepticism for people with a low BJW is driven by a reduction in skepticism around the belief that the vaccine will not prevent illness, b = - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='32, [95% C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' : - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='50, -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='14], p < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This suggests that people with a low BJW, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' those who do not believe that there is a justice system which ensures that overall good things happen to good 15 people and bad things happen to bad people, are extremely reactive to the individualistic message.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' It increases their confidence in the vaccine being able to protect them from serious illness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In other words, receiving the individualistic message, which accurately highlights that receiving the vaccine can prevent serious illness, correctly updates the beliefs around this issue for those with a low BJW, but not for those with a strong BJW.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This suggests that for someone with a strong BJW, the belief in this just world order overpowers the belief in the science of the vaccine, as perhaps the deservingness of a person to fall ill would govern their likelihood of sickness moreso than the vaccine’s protective properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' [Figures 2 and 3 here] Furthermore, we do not find evidence that people with a strong BJW react particularly poorly to the community message, b = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='03, [95% C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' : - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='16, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='22], p = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='778.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This suggests that a message which urges the public to take care of its community does not come into strong conflict with believers of a just world who may not feel responsible for the pandemic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This lack of resistance is consistent with just world believers’ willingness to engage in other COVID-19 preventative measures (Devereux et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' [Table 2 here] 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Discussions Our findings that the individual and community messages concerning the COVID-19 vaccine can shift beliefs around the vaccine’s various protective functions demonstrates an unsurprising link between the presentation of fact and its influence on a corresponding attitude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Nevertheless, in their desperate attempts to convince the public to get vaccinated, policymakers have sometimes turned to extreme measures, such as million-dollar lotteries, rifle giveaways, and free beer and 16 donuts (Lewis 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' However, while policymakers may have expected a clear increase in uptake, emerging evidence suggests that there is limited evidence in favor of these creative incentivizing strategies (Walkey et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Acharya & Dhakal, 2021), perhaps due to newfound suspicion of such gimmicky programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Instead, policymakers should provide truthful information about the capacities of the COVID-19 vaccine, relying on existing evidence that these strategies effectively lower vaccine skepticism (Pennycook et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Yuan & Chu, 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Our messages do not easily shift the belief that the vaccine reduces transmission of the virus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This is especially important as new evidence emerges around the limited effectiveness of the vaccines against mutations of the coronavirus and in preventing transmission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Early studies suggest that the COVID-19 vaccines may not be as effective in preventing transmission as previously thought (Reuters, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While policymakers should highlight the protective benefits of the vaccine, they must be cautious in not overstating the vaccine’s effectiveness around transmission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Doing so could give vaccinated individuals a false sense of security, and ultimately reduced trust in public health authorities, resulting in less social distancing and respect for COVID-19 guidelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' As new scientific evidence about the vaccine emerges, officials must update their messaging content accordingly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The literature shows that prosocial messages play an important role in motivating COVID-19 preventative actions, like signing up for contact-tracing apps (Jordan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In contrast, vaccine skepticism responds differently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Consistent with previous findings (Yuan & Chu, 2022), we show that individual responsibility messages work as well, and sometimes better, than the community messages in reducing vaccine skepticism, depending on the dimension of skepticism 17 in question.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This discrepancy between non-vaccine COVID-19 prevention and vaccine messages could be because general preventative measures are perceived to be less risky than taking the vaccine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Riskier behaviors require more self-gain, which explains why the individual message is more successful.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Furthermore, the pro-vaccine messages used in this experiment affect different domains of vaccine skepticism differently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' More specifically, they do not convince the population that the vaccine is useful to ending the pandemic, nor do they influence vaccination intentions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In the urgent pandemic context, while attitudes matter, vaccination behaviors are even more critical.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Alternative strategies to motivate behavior must not be overlooked or confounded with strategies that target attitudes in future research.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' When further examining heterogeneous treatment affects by intensity of BJW, we find that the overall success of the individual message is more robust among individuals with a low BJW, compared to those with a high BJW.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The individual message, which focusses on the primary effect of the vaccine, may speak more particularly to people with a weak BJW because they see the world in a more factual, cartesian way.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Someone with a strong BJW, on the other hand, may consider competing justice-related reasonings for the spread of or protection against COVID.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The same is not true of the effects of the community message.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Individuals with a strong BJW were found to be unmoved by the community message, possibly because this prosocial message sets an expectation that challenges the distribution of responsibility in a just world, as previously discussed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' While individuals who see public health as a moral issue are more persuaded by other-focused (rather than self-focused) social distancing messages (Luttrell & Petty, 2020), BJW is not a worldview 18 based on altruistic morals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Rather, where others may fall ill due to COVID-19, strong believers of a just world would blame the patients for their own misfortune, rather than assuming responsibility over the pandemic via mass collective vaccination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Our results suggest that evidence-based messages (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' : the vaccine will protect you) have heterogeneous effects according to worldview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This heterogeneity replicates the findings of Yuan & Chu, who recently demonstrate that the individual-centered COVID-19 vaccine message is more impactful than a community-centered one, largely due to people whose worldview aligns with a more individualistic outlook (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Our studies differ in that we examine BJW, rather than individualism/communitarianism, and our sample was based in the UK, rather than the US.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' However, broadly speaking, the results confirm one another’s findings, which is that the individual-centered message works best overall, but that this effect is driven largely by people with a worldview that places themselves, the individual, independent of a larger community or justice structure, at the center.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Authorities ought to take into consideration the extent to which their vaccine messaging can have heterogeneous effects according to the worldviews of their population, especially as they encourage vaccine take-up amongst people with more extreme worldviews.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Conclusions Simple messages that promote the COVID-19 vaccine effectively reduce vaccine skepticism of the corresponding beliefs around the vaccine’s effectiveness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This reassuringly highlights the importance for policymakers to focus the information of their vaccination campaigns on the specific concerns of the public.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The differences we find in effectiveness by psychological outlook 19 are important for policymakers to consider, especially as the remaining unvaccinated likely hold more extreme world views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Messages that work well for people with low-level BJW evidently work less well for those with a more extreme worldview, suggesting that policymakers must reconsider how to motivate those harder-to-reach populations to get vaccinated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Custom messages that directly target people with such views could be an interesting line of research to follow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' This research is not without limitations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' First, the data is restricted to a specific age-group in the United Kingdom and therefore has not been tested in other contexts, where just-world beliefs and vaccine skepticism differ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' For example, in the United States, conservatism links with both BJW (Furnham, 2003) and COVID-19 vaccine skepticism (Latkin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2021), suggesting that BJW might be negatively correlated with pro-vaccine attitudes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Second, the sample in our study is not quota matched to the U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' population, nor was it obtained using probability sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Hence, the results cannot be considered nationally representative, and there is likely a degree of selection bias amongst users of Prolific.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Third, our dataset does not capture whether participants ultimately took up the vaccination, as it only captures attitudes and intentions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' As previously discussed, behaviors in this context are more important than attitudes, and would be valuable to follow up on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The authors declare no conflicts of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 20 References Acharya, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Dhakal, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Implementation of state vaccine incentive lottery programs and uptake of COVID-19 vaccinations in the United States.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' JAMA Network Open, 4(12), e2138238-e2138238.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Betsch, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Böhm, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Moral values do not affect prosocial vaccination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Nature Human Behaviour, 2(12), 881-882.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1038/s41562-018-0478-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Betsch, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Böhm, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Korn, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Holtmann, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' On the benefits of explaining herd immunity in vaccine advocacy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Nature Human Behaviour, 1(3), 1-6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1038/s41562-017-0056.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Browne, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Thomson, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Rockloff, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Pennycook, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Going against the herd: psychological and cultural factors underlying the ‘vaccination confidence gap’.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' PLoS one, 10(9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1371/journal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='pone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0132562.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Correia, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Batista, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Lima, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2009).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Does the belief in a just world bring happiness?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Causal relationships among belief in a just world, life satisfaction and mood.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Australian Journal of Psychology, 61(4), 220-227.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1080/00049530802579515.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Dalbert, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2009).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Belief in a just world.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Handbook of individual differences in social behavior, 288-297.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Dalbert, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=" The world is more just for me than generally: About the personal belief in a just world scale's validity." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Social justice research, 12(2), 79-98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1023/A:1022091609047 Dalbert, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Montada, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Schmitt, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1987).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Glaube an eine gerechte Welt als Motiv: Validierungskorrelate zweier Skalen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Psychologische Beitrage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Devereux, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Miller, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Kirshenbaum, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Moral disengagement, locus of 21 control, and belief in a just world: Individual differences relate to adherence to COVID- 19 guidelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Personality and Individual Differences, 182, 111069.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='paid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='111069 Feinberg, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Willer, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Apocalypse soon?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Dire messages reduce belief in global warming by contradicting just-world beliefs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Psychological science, 22(1), 34-38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1177/0956797610391911.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Furnham, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2003).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' ‘Belief in a just world: Research progress over the past decade’, Personality and Individual Differences, 34(5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/S0191-8869(02)00072-7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Gallagher, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Updegraff, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2012).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Health message framing effects on attitudes, intentions, and behavior: a meta-analytic review.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Annals of Behavioral Medicine, 43(1), 101-116.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1007/s12160-012-9446-6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Jiang, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Yue, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Lu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Yu, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Zhu, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' How belief in a just world benefits mental health: The effects of optimism and gratitude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Social Indicators Research, 126(1), 411- 423.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1007/s11205-015-0877-x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Jolley, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Douglas, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The effects of anti-vaccine conspiracy theories on vaccination intentions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' PloS one, 9(2), e89177.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1371/journal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='pone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0089177 Jordan, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Yoeli, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Rand, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Don’t get it or don’t spread it?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Comparing self- interested versus prosocially framed COVID-19 prevention messaging.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' PsyArXiv, 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Karlsson, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Soveri, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Lewandowsky, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Karlsson, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Karlsson, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Nolvi, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' & Antfolk, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Fearing the disease or the vaccine: The case of COVID- 19.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Personality and individual differences, 172, 110590.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='paid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='110590.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 22 Katella, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Comparing the COVID-19 Vaccines: How are They Different?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='. Yale Medicine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Available at: http://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='yalemedicine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/news/covid-19-vaccine-comparison (Accessed: 30 April, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Kelly, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' and Hornik, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Effects of framing health messages in terms of benefits to loved ones or others: An experimental study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Health Communication, 31(10), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1284- 1290.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1080/10410236.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1062976 Khubchandani, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Sharma, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Price, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Wiblishauser, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Sharma, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Webb, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' COVID-19 vaccination hesitancy in the United States: a rapid national assessment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Journal of Community Health, 46(2), 270-277.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1007/s10900-020-00958-x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Latkin, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Dayton, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Yi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Colon, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Kong, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Mask usage, social distancing, racial, and gender correlates of COVID-19 vaccine intentions among adults in the US.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' PloS one, 16(2), e0246970.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1371/journal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='pone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0246970.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Lerner, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Simmons, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1966).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Observer\'s reaction to the" innocent victim": Compassion or rejection?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='. Journal of Personality and social Psychology, 4(2), 203.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1037/h0023562.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Lewis, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' From $1-Million Lotteries to Free Beer: Do COVID Vaccination Incentives Work?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Scientific American.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='scientificamerican.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='com/article/from-1-million- lotteries-to-free-beer-do-covid-vaccination-incentives-work1/.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Lucas, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Alexander, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Firestone, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Lebreton, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2009).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Belief in a just world, social influence and illness attributions: Evidence of a just world boomerang effect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Journal of Health Psychology, 14(2), 258-266.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1177/1359105308100210.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Luttrell, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Petty, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Evaluations of self-focused versus other-focused arguments 23 for social distancing: An extension of moral matching effects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Social Psychological and Personality Science, 12(6), 946-954.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1177/1948550620947853.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' McPhee, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Nguyen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Euler, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Mock, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Wong, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Lam, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Nguyen, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Nguyen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Ha, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Do, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' and Buu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Successful promotion of hepatitis B vaccinations among Vietnamese-American children ages 3 to 18: results of a controlled trial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Pediatrics, 111(6), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1278-1288.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1542/peds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1278.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Motta, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Callaghan, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Sylvester, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Lunz-Trujillo, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Identifying the prevalence, correlates, and policy consequences of anti-vaccine social identity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Politics, Groups, and Identities, 1-15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1080/21565503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1932528.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Nestik, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Deyneka, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Socio-psychological predictors of belief in conspiracy theories of the origin of COVID-19 and involvement in social media.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Social Psychology and Society, 11(4), 87-104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='17759/sps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' NHS UK (2021) 25 April.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Available at https://twitter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='com/NHSuk (Accessed: 18 May 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Pennycook, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', McPhetres, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Lu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Rand, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Fighting COVID-19 misinformation on social media: Experimental evidence for a scalable accuracy-nudge intervention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Psychological Science, 31(7), 770-780.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1177/0956797620939054.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Peretti-Watel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Seror, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Cortaredona, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Launay, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Raude, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Verger, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' & Ward, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A future vaccination campaign against COVID-19 at risk of vaccine hesitancy and politicisation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The Lancet Infectious Diseases, 20(7), 769-770.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/S1473-3099(20)30426-6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Pogue, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Jensen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Stancil, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Ferguson, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Hughes, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Mello, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' & Poole, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Influences on attitudes regarding potential COVID-19 vaccination in the 24 United States.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Vaccines, 8(4), 582.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3390/vaccines8040582.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Porter, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Amin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Bednarczyk, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Omer, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Cancer-salient messaging for human papillomavirus vaccine uptake: A randomized controlled trial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Vaccine, 36(18), 2494-2500.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='vaccine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='01.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='040.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Pritchard, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Matthews, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Stoesser, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Eyre, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Gethings, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Vihta, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' & Pouwels, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Impact of vaccination on new SARS-CoV-2 infections in the UK.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Nature Medicine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1038/s41591-021-01410-w Reuters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021, August 6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Early signs COVID-19 vaccines may not stop Delta transmission, England says.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='reuters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='com/world/uk/england-says-delta-infections-produce- similar-virus-levels-regardless-vaccine-2021-08-06/.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Robertson, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Reeve, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Niedzwiedz, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Moore, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Blake, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Green, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' & Benzeval, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Predictors of COVID-19 vaccine hesitancy in the UK household longitudinal study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Brain, Behavior, and Immunity, 94, 41-50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='bbi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='03.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Ross, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1977).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The intuitive psychologist and his shortcomings: Distortions in the attribution process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In Advances in experimental social psychology (Vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 10, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 173-220).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Academic Press.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1016/S0065-2601(08)60357-3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Vaccinations in United Kingdom (30 April, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Gov.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='UK Coronavirus (COVID-10) in the UK.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Available at: https://coronavirus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='gov.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='uk/details/vaccinations (Accessed: 30 April, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Walkey, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Law, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Bosch, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Lottery-based incentive in Ohio and COVID-19 vaccination rates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' JAMA, 326(8), 766-767.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Wenzel, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Schindler, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Reinhard, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' General belief in a just world is positively 25 associated with dishonest behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Frontiers in psychology, 8, 1770.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3389/fpsyg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='01770.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' White, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', Norenzayan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Schaller, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' The content and correlates of belief in Karma across cultures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Personality and Social Psychology Bulletin, 45(8), 1184-1201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1177/0146167218808502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Yuan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=', & Chu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Vaccine for yourself, your community, or your country?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Examining audiences’ response to distance framing of COVID-19 vaccine messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Patient Education and Counseling, 105(2), 284-289.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Zizzo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Experimenter demand effects in economic experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Experimental Economics, 13(1), 75-98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1007/s10683-009-9230-z 26 1 How Effective are COVID-19 Vaccine Health Messages in Reducing Vaccine Skepticism?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Heterogeneity in Messages’ Effectiveness by Just-World Beliefs Tables and Figures Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Proportions of overall skepticism changes across control, individual, and community messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Change in standardized vaccine skepticism Change in skepticism (SD) from T=0 to T=1 5 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='05 0 Overall skepticism Control Individual Community 95% CI n = 526Table 1: The effects of public health messages on overall vaccine skepticism factor outcome: OLS regressions (1) (2) (3) ∆ Skepticism factor (std) ∆ Skepticism factor (std) ∆ Skepticism factor (std) Individual 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='113* 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='113* 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='189*** (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0453) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0457) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0657) Community 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0872 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0872 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='101 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0502) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0502) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0639) High BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000134 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0569 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0434) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0658) Individual x High BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='147 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0911) Community x High BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0274 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0969) Optimism (std) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00429 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00432 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00569 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0201) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0215) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0215) Age 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00640 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00640 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00625 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0190) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0191) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0188) Age squared 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='56e 05 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='56e 05 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='20e 05 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000293) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000294) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000290) Female 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000917 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000910 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00305 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0438) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0440) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0436) £ 30k+ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0179 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0179 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0153 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0410) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0414) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0411) London 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0313 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0313 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0292 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0635) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0636) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0633) University+ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0467 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0467 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0431 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0430) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0434) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0429) White 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0210 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0210 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0254 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0703) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0705) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0708) Labour 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00769 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00768 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00469 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0375) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0372) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0378) Constant 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0700 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0701 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0499 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='303) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='307) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='300) Cluster individuals 526 526 526 R-squared 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='024 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='024 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='029 Note: *** p<0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='001, ** p<0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='05.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Robust standard errors clustered at the individual level and are in parentheses.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Dependent variables represent the change from #!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' to #" and are standardized to have a mean of 0 and a standard deviation of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Table 2: The effects of public health messages on individual skepticism outcomes: OLS regressions with BJW interactions (1) (2) (3) (4) ∆ Vaccine not useful (std) ∆ Not accept vaccine (std) ∆ Not reduce transmission (std) ∆ Not prevent illness (std) Individual 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='166 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0212 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0458 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='323*** (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='102) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0466) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='107) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0917) Community 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='101 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0175 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0904 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='139 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0951) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0545) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='110) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0972) High BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0605 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00896 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0244 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='103 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0951) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0543) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='116) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='113) Individual x High BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='147 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0565 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0416 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='214 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='135) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0790) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='158) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='146) Community x High BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0651 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0288 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='189 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0998 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='134) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0827) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='170) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='155) Optimism (std) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0281 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='72e 05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00949 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0194 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0253) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0188) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0378) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0344) Age 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00199 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0131 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0512 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00812 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0276) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0156) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0311) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0281) Age squared 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='24e 05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000208 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000845 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000158 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000424) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000227) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000481) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='000433) Female 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0213 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0378 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0621 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0174 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0616) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0397) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0760) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0696) £ 30k+ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00347 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0365 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0707 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0745 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0606) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0381) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0706) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0662) London 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00687 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0415 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='00830 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0467 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0989) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0448) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0875) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0791) University+ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0975 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0117 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0588 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0427 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0601) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0326) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0678) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0667) White 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0705 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0778 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='122 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0191 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='108) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0591) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='118) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='115) Labour 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0578 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0123 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0130 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0311 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0557) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0321) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0687) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0605) Constant 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0239 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='114 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='716 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='284 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='458) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='259) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='498) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='444) Cluster individuals 526 526 526 526 R-squared 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='023 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='021 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='032 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='034 Note: *** p<0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='001, ** p<0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='05.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Robust standard errors clustered at the individual level and are in parentheses.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Dependent variables represent the change from #!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' to #" and are standardized to have a mean of 0 and a standard deviation of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Figure 2: Predictive margins of the individual treatment and control group,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' over the standardized BJW factor Predictive Margins of treat with 95% Cls 2 Linear Prediction 0 2 -4 -2 0 2 4 Standardized BJW Factor Control IndividualFigure 3: Predictive margins of the community treatment and control group,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' over the standardized BJW factor Predictive Margins of treat with 95% Cls Linear Prediction 2 0 2 -4 -2 0 2 4 Standardized BJW Factor Control Community 1 How Effective are COVID-19 Vaccine Health Messages in Reducing Vaccine Skepticism?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Heterogeneity in Messages’ Effectiveness by Just-World Beliefs Appendix 2 Table A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1: Balance checks on all observable characteristics amongst the final analysis sample.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Control (0) Individual (1) Community (2) (0) vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (1), p value (0) vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2), p value (1) vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (2), p value 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' (baseline) Not vaccine useful 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='704 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='904 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='616 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6) 172 180 174 Not accept vaccine 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='988 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='628 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='625 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 172 180 174 Not reduce transmission 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='645 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='670 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='983 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9) (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 172 180 174 Not prevent illness 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='895 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='741 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='839 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8) 172 180 174 𝑇" (endline) Not vaccine useful 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='253 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='615 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='566 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5) 172 180 174 Not accept vaccine 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='892 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='602 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='684 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8) 172 180 174 Not reduce transmission 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='548 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='039 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='124 (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8) 172 180 174 Not prevent illness 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='054 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='581 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='187 (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5) (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) 172 180 174 Quartile 1 BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='626 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='671 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='358 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 172 180 174 Quartile 4 BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='529 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='352 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='756 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 172 180 174 Optimism factor (Std) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='928 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='779 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='850 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1) 172 180 174 Age 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='479 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='231 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='600 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7) 172 180 174 Female 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='309 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='459 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='781 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 166 171 170 £30,000+ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='987 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='455 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='423 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1) 172 180 174 London 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='652 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='248 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='472 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 172 180 174 Undergraduate+ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='778 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='808 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='597 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 171 179 174 White 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='525 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='724 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='779 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 172 180 174 Labour party 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='825 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='438 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='578 3 (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='0) 169 172 172 Note: standard deviations in parenthesis, sample size of respondents in italics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 4 Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1: Experimental process and consort diagram ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='T0 600participants Completed BJW scale and baseline skepticism outcomes T1 600 participants Invited to return for part 2 Control Individual treatmentCollective treatment 172 participants 181 participants 174 participants Read article about Read article about individual benefits communitybenefits to vaccination tovaccination Passed attention Passed attention checks checks 180 participants 174 participants 526 participants Completed endline skepticism outcomes and demographic questions 5 Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2: Survey design: questions at 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='. Please read each statement carefully and indicate the extent to which you personally agree or disagree with it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Very Very strongly Slightly Slightly strongly disagree Disagree disagree agree Agree agree I think basically the 0 0 0 0 0 0 world is a justplace.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I believe that, by and large, people get what 0 0 0 0 0 0 they deserve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Iam confidentthat justice always prevails 0 0 0 0 0 over injustice.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I am convinced that in the long run, people 0 will be compensated for injustices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I firmly believe that injustices inallareas of life (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' professional, family, 0 0 0 politics) are the exception rather than the rule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I think people try to be fair when making 0 0 0 important decisions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 6 Please read each statement carefully and indicate the extent to which you personally agree ordisagree with it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Very Very strongly Slightly Slightly strongly disagree Disagree disagree agree Agree agree I believe that, by and large, I deserve what 0 0 0 0 0 0 happens to me.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I am usually treated 0 0 0 0 0 0 fairly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I believe that I usually 0 0 0 0 0 0 get what I deserve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Overall, events in my 0 0 0 0 0 0 life are just.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' In my life injustice is theexceptionrather 0 0 0 0 0 0 than the rule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I believe that most of 0 0 0 0 0 0 the things that happen in my life are fair.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I think that important decisionsthatare 0 made concerning me are usually just.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 7 How certainareyouthattheCOViD-19vaccinesare ausefultool infightingthepandemic?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Not at all certain Extremelycertain 0 10 20 30 40 50 60 70 80 90 100 How likelyareyouto accept the CovID-19vaccinewhen offered?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Not at all likely Extremely likely 0 10 20 30 40 50 60 70 80 90 100 HowcertainareyouthattheCOViD-19vaccinereducestransmissionbetweenindividuals?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 0 10 20 30 40 50 60 70 80 90 100 How certainareyouthattheCoviD-19vaccinewouldpreventyoupersonallyfromgettingveryill dueto COVID-19?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 0 10 20 30 40 50 60 70 80 90 100 8 Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='1: Survey design: treatment messages at 𝑇".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Control participants were asked to respond to the same four skepticism outcomes shown in Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Individual (left) and community (right) messages participants were first asked to read the following fictitious news articles and were then prompted to respond to the four skepticism outcomes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Belowisanewsstorysimilartoothernewsstoriesyoumighthavereadbefore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Please readthestoryandrespondtothequestionsthatfollow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' BOsTON --"Coronavirus disease (COVID-19)is a highly contagious illness, caused bythe transmission of the SARS-CoV-2 virus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' First identified in December 2019, the virus has causedapandemicthatresulted inshutdownsall aroundtheglobe.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='It wasfirst widely haswreakedhavocontheglobe.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Claimingmillionsoflives,thispandemichascreateda cleardemarcationintime:pre-covid,andpost-covid,"says ProfessorArthurMichali,a publichealthexpertfromaleadingresearchuniversity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Beforethispandemic,Tcouldhave attendedaconferenceinTokyo oneday,ledaresearchcollaboration inGenevathenext, andarrivedback inBostonthethirdday.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content="Thiskindoftravel issimplynolongerpossible under currentcircumstances,andit's likelythatthis sort ofbehaviourcontributedtothe rapidspreadofthediseaseworldwide." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='" ProfessorMichali,whohaswonnumerousawardsforhis researchoverthelasttwo decades,ispartoftheCOViD-19EmergencyCommitteeattheWorldHealthOrganisation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Amongstothertopics,thiscommitteeisworkingtobetterunderstandthevarious responses and interventions that can help curb the spread of the disease.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Michali is co-authoring a forthcoming pamphlet, entitled"The COViD-19Vaccine:what circulatingandforthcomingvaccines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='ThepamphletdescribesCOviD-19as"adangerous disease,particularlyforthe elderlyand clinicallyvulnerable,astheyare more likelyto suffersevere,andpossiblyfatal,respiratoryillness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Nevertheless,anyone,regardlessof ageormedical background,isatriskof sufferingaharshillness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Michaliwishesto emphasisethatthebestthingyoucandotoprotectyourselffromthisdiseaseisto takeupthevaccinewhenyouareofferedit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Some ofthevaccinesonthemarketare boasting95%efficacyrates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Thismeansthatreceivingthevaccinedramaticallyreduces yourriskofdevelopingseriousCOViD-19symptomsifyouareexposedtotheviruslater downtheline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Althoughexpertsarecontinuingtoemphasisetheimportanceofsocial distancingandwearingmasks,thesemeasuresare notperfect,andthereremainsariskof inadvertentlycatchingthediseasethatcouldleaveyoubed-riddenforweeks,even months.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Receivingthe vaccine isthesingle most important stepan individualcantaketo protect him orherself fromthe virus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Michali reflects intheconcludingthoughts of the pamphlet,"thereisnotmuchthatwecancontrolintimeslikethese,butyouneedto do whatyoucantoprotectyourself inthese uncertain times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Takingup thevaccine whenoffered isthebestactionyoucantaketokeepyourself safe!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Importantly, Michali wants individuals to rememberthat it is their personal responsibility to keep themselvesprotected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Below isanewsstorysimilartoothernewsstoriesyoumighthavereadbefore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Please readthestoryand respondtothequestionsthatfollow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' BOsTON--“Coronavirus disease (COVID-19)is a highly contagious illness, caused bythe transmission oftheSARS-CoV-2 virus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='First identified inDecember2019,the virushas caused apandemic that resulted in shutdowns all aroundtheglobe.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Itwasfirst widely spreadbetweenhumansatawholesaleseafoodmarket inWuhan,China.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Thisdisease haswreakedhavocontheglobe.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Claimingmillionsof lives,thispandemichascreateda cleardemarcation intime:pre-covid,and post-covid,says ProfessorArthurMichali,a publichealthexpertfromaleadingresearchuniversity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Beforethispandemic,Icouldhave attendedaconference inTokyooneday,ledaresearchcollaboration inGenevathenext, andarrivedbackinBostonthethirdday.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content="Thiskindoftravelissimplynolongerpossible undercurrentcircumstances,andit's likelythatthissortof behaviourcontributedtothe rapidspreadofthediseaseworldwide." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='" ProfessorMichali,whohaswonnumerousawardsforhisresearchoverthelasttwo decades, is part of the COviD-19 Emergency Committee at the World Health Organisation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Amongstothertopics,thiscommitteeisworkingtobetterunderstandthevarious Michali is co-authoring a forthcoming pamphlet, entitled “"The COviD-19 Vaccine: what circulatingandforthcomingvaccines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='ThepamphletdescribesCOviD-19as"adangerous disease,particularly forthe elderly and clinically vulnerable,as they are more likely to suffersevere,andpossiblyfatal,respiratoryillness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Nevertheless,anyone,regardlessof ageormedical background,is atrisk ofsufferingaharsh illness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Michali wishesto emphasisethatthebestthingyoucandotoprotectothersfromthisdiseaseisto takeupthevaccinewhenyouareoffered it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Someofthevaccinesonthemarketare boasting95%efficacyrates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Thismeansthatreceivingthevaccinedramaticallyreduces yourriskofdevelopingseriousCOviD-19symptomsifyouareexposedtotheviruslater downtheline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='CommunitytransmissionhasbeenshowntobelowerwhensevereCOviD 19symptomsdonotpresent,soyouareprotectingyourneighbours,parents, grandparents,andfriendsbyreceivingthevaccine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' "Althoughexpertsarecontinuingto emphasisetheimportanceofsocialdistancingandwearingmasks,thesemeasuresare notperfect,astheviruscanstillspreadbetweenpeople.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Theworryisnotsomuchabout individual cases, but rather, it is about reducing transmission in communities, as it is that type oftransmissionthat will preventus from everseeing anendtothispandemic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' thecommunityfromthevirus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Michali reflectsintheconcludingthoughtsofthepamphlet, "thereisnotmuchthatwe can control intimes likethese,butweneedtotake collectiveactiontofightthispandemic!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='Takingupthevaccinewhenofferedisthe bestactionyoucantakeforyourfamily,friends,andforyourcommunity!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Importantly,Michaliwantspeopletorememberthatitistheirresponsibilitytokeeppeople intheircommunity,especiallythosewhoarevulnerabletothedisease,protected 9 Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2: Survey design : manipulation check at 𝑇".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' According to the article, where was the COVID-19 virus first widely spread?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Geneva,Switzerland Boston, USA Wuhan, China Tokyo, Japan Accordingtothearticle,whatisProfessorArthurMichali currentlyworkingon?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' A strategy to liaise with journalists and media about COVID-19 Apamphletto informthe average citizenaboutthe currentandforthcoming COviD-19vaccines Atravel itineraryfromTokyotoGenevato Boston Asociologicalstudyonthespreadof COviD-19 According to the article, whom will you primarily protect by taking up a CoOVID-19 vaccine?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Yourself Healthworkersinothercountries Peoplewho have justdiedofCOvID-19-related illness Others in your community 10 Figure A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3: Survey design: demographic questions at 𝑇".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Areyou generally aperson who tries to avoid taking risks orare youfully prepared to take risks?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=" Won't take risks Ready to take risks 0 1 2 3 4 5 6 7 8 6 10 HaveyoubeendiagnosedwithCovID-19atanypoint?" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Howfrequentlydoyoufollowgovernmentguidelinesonfacecoveringswheninshops?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I neverwearafacecoveringbecauseIamexemptfromwearingone I neverwearafacecoveringand I amnotexemptfromwearingone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Most of the time I do not wear a face covering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' HalfofthetimeIwearafacecovering,halfI donot.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Most ofthe time I wearaface covering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Ialwayswearafacecovering How confident are you thatface coverings area useful tool in fighting the pandemic?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Not at all confident Extremelyconfident 0 10 20 30 40 50 60 70 80 90 100What isyourethnicity?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' What is the highest level of education that you have completed?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Inwhichregiondoyoucurrentlyreside?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' What is youryearlyhousehold incomebeforetax?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' V Whichpolitical party do you consideryourself to beclosest to?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Please indicateyourattitudesto eachof thefollowingstatements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' I neither I disagree a I disagree a agree nor Iagree a lot little disagree little I agree a lot In uncertain times, usuallyexpectthe 0 0 0 0 0 best.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=" I'm always optimistic 0 0 0 0 0 aboutmy future." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Overall,Iexpectmore good things to happen 0 0 0 0 0 to me than bad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' 11 Table A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='2: Factor analysis on the personal and general BJW factors, which produce the combined BJW factor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Factor analysis/correlation Factor Eigenvalue Difference Proportion Cumulative Factor1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='79 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='04 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='46 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='46 Factor loadings (pattern matrix) and unique variances Variable Factor1 Uniqueness General BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='63 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='61 Personal BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='63 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='61 Scoring coefficients Variable Factor1 General BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='41 Personal BJW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='41 Cronbach’s alpha a 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='69 Table A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='3: Factor analysis on the skepticism outcomes at 𝑇!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='. Factor analysis/correlation Factor Eigenvalue Difference Proportion Cumulative Factor1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='57 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='60 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='10 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='10 Factor loadings (pattern matrix) and unique variances Variable Factor1 Uniqueness Vaccine Useful 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='86 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='26 Accept Vaccine 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='83 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='31 Reduce Transmission 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='59 Prevent Illness 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='85 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='28 Scoring coefficients Variable Factor1 Vaccine Useful 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='35 Accept Vaccine 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='28 Reduce Transmission 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='12 Prevent Illness 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='31 12 Cronbach’s alpha a 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='88 Table A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='4: Factor analysis on the skepticism outcomes at 𝑇".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content=' Factor analysis/correlation Factor Eigenvalue Difference Proportion Cumulative Factor1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='69 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='70 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='09 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='09 Factor loadings (pattern matrix) and unique variances Variable Factor1 Uniqueness Vaccine Useful 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='88 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='22 Accept Vaccine 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='82 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='33 Reduce Transmission 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='51 Prevent Illness 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='87 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='25 Scoring coefficients Variable Factor1 Vaccine Useful 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='37 Accept Vaccine 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='23 Reduce Transmission 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='13 Prevent Illness 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='32 Cronbach’s alpha a 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
+page_content='89' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/6tE1T4oBgHgl3EQfnAQs/content/2301.03303v1.pdf'}
diff --git a/89AzT4oBgHgl3EQfFPox/vector_store/index.faiss b/89AzT4oBgHgl3EQfFPox/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..5713f215a3446fceb3821023cb6a94b6e359cb7f
--- /dev/null
+++ b/89AzT4oBgHgl3EQfFPox/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00f5e27416d3d1332fb32baf50133b4dbf0e87bb8ffba6ef09da15ae82546e14
+size 4063277
diff --git a/8tFAT4oBgHgl3EQfpB3g/content/tmp_files/2301.08638v1.pdf.txt b/8tFAT4oBgHgl3EQfpB3g/content/tmp_files/2301.08638v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..fddf6f12a9c700723bbd32df1a2d591aa5e8b41d
--- /dev/null
+++ b/8tFAT4oBgHgl3EQfpB3g/content/tmp_files/2301.08638v1.pdf.txt
@@ -0,0 +1,589 @@
+arXiv:2301.08638v1 [hep-th] 20 Jan 2023
+Enlarging the symmetry of pure R2 gravity, BRST invariance
+and its spontaneaous breaking
+Ariel Edery∗
+Department of Physics and Astronomy, Bishop’s University, 2600 College Street,
+Sherbrooke, Qu´ebec, Canada, J1M 1Z7.
+Abstract
+Pure R2 gravity was considered originally to possess only global scale symmetry. It was
+later shown to have the larger restricted Weyl symmetry where it is invariant under the
+Weyl transformation gµν → Ω2(x) gµν when the conformal factor Ω(x) obeys the harmonic
+condition □Ω(x) = 0. Restricted Weyl symmetry has an analog in gauge theory. Under a
+gauge transformation Aµ → Aµ + 1
+e∂µf(x), the gauge-fixing term (∂µAµ)2 has a residual
+gauge symmetry when □f = 0. In this paper, we consider scenarios where the symmetry
+of pure R2 gravity can be enlarged even further. In one scenario, we add a massless scalar
+field to the pure R2 gravity action and show that the action becomes on-shell Weyl invari-
+ant when the equations of motion are obeyed. We then enlarge the symmetry to a BRST
+symmetry where no on-shell or restricted Weyl condition is required. The BRST trans-
+formations here are not associated with gauge transformations (such as diffeomorphisms)
+but with Weyl (local scale) transformations where the conformal factor consists of a prod-
+uct of Grassmann variables. BRST invariance in this context is a generalization of Weyl
+invariance that is valid in the presence of the Weyl-breaking R2 term. In contrast to the
+BRST invariance of gauge theories like QCD, it is not preserved after quantization since
+renormalization introduces a scale (leading to the well-known Weyl (conformal) anomaly).
+We show that the spontaneous breaking of the BRST symmetry yields an Einstein action;
+this still has a symmetry which is also anomalous. This is in accord with previous work
+that shows that there is conformal anomaly matching between the unbroken and broken
+phases when conformal symmetry is spontaneously broken.
+∗aedery@ubishops.ca
+1
+
+1
+Introduction
+Pure R2 gravity (R2 alone with no additional R term) is unique among quadratic gravity the-
+ories as it is unitary and moreover has been shown to be conformally equivalent to Einstein
+gravity with non-zero cosmological constant and massless scalar field [1–5] (though in a Pala-
+tini formalism one can avoid having a massless scalar [6]). It has been known for a long time
+that it is invariant under the global scale transformation gµν → λ2 gµν where λ is a constant. It
+was later discovered to possess a larger symmetry than global scale symmetry called restricted
+Weyl symmetry [7] where it is invariant under the transformation gµν → Ω2(x) gµν when the
+conformal factor Ω(x) obeys the harmonic condition □Ω = gµν∇µ∇νΩ = 0. The conformal
+factor Ω(x) is therefore not limited to being a constant. The aforementioned equivalence be-
+tween pure R2 gravity and Einstein gravity with cosmological constant was then interpreted
+in a new light: it occurs when the restricted Weyl symmetry is spontaneously broken [3,5]. In
+the broken sector, the Ricci scalar of the background (vacuum) spacetime has R ̸= 0 which
+excludes a flat background. This is why the equivalence requires a non-zero cosmological con-
+stant on the the Einstein side. The unbroken sector which has an R = 0 vacuum (background)
+has no relation to Einstein gravity. In fact, it has been shown that a linearization of pure R2
+gravity about Minkowski spacetime does not yield gravitons but only a propagating scalar [4];
+simply put, pure R2 gravity does not gravitate about a flat background [4]. However, it was
+later shown that if one includes a non-minimally coupled scalar field in the restricted Weyl-
+invariant action and the field acquires a non-zero VEV, then the theory can gravitate about
+flat spacetime [5,8]. Various aspects of restricted Weyl symmetry, it spontaneous breaking as
+well as its role in critical gravity were then explored further in [5–7,9–11]
+Restricted Weyl symmetry has an analog in gauge theory. The gauge-fixing term (∂µAµ)2
+is invariant under the gauge transformation Aµ → Aµ + 1
+e∂µf(x) only when the arbitrary
+smooth function f(x) obeys the condition □f = 0 where □ here represents the flat space
+d’Alembertian. Therefore, the gauge-fixing term has a residual gauge symmetry when □f = 0
+is satisfied [12]. This is the analog to the restricted Weyl symmetry of pure R2 gravity when
+the conformal factor Ω(x) satisfies □Ω = 0. As we will see, this analogy is fruitful as it provides
+a bridge to the BRST symmetry of pure R2 gravity. Recent work on the BRST invariance of
+other gravitational theories can be found in [13,14,16].
+In this paper, we consider scenarios where the symmetry of pure R2 gravity is enlarged further.
+We show that when a massless scalar field is added to pure R2 gravity, the action becomes Weyl
+invariant when the equations of motion are satisfied. No separate external condition is required
+to be imposed on the conformal factor Ω(x) as this occurs naturally via the equations of motion.
+One passes from restricted Weyl invariance to on-shell Weyl invariance. One can then enlarge
+the symmetry further to include BRST symmetry. In analogy with the BRST invariance in
+gauge theories in the presence of a gauge-fixing term, we establish BRST invariance in the
+2
+
+presence of the Weyl-breaking pure R2 gravity term. The BRST transformations here are
+not associated with gauge transformations (such as diffeomorphisms) but are a generalization
+of Weyl (local scale) transformations where the conformal factor is composed of Grassmann
+variables. Therefore, in contrast to the BRST invariance in gauge theory, it is anomalous since
+renormalization introduces a scale (leading to the well-known Weyl (conformal) anomaly). We
+show that the spontaneous breaking of the BRST symmetry yields an Einstein action with its
+own symmetry that is also anomalous. This is in agreement with previous work where it was
+shown that when conformal symmetry is spontaneously broken there is conformal anomaly
+matching in the unbroken and and broken phases [18,19].
+The paper is organized as follows. In section 2, we obtain the on-shell Weyl invariance of pure
+R2 gravity when a massless scalar field is included in the action. In section 3, we obtain the
+BRST invariance of pure R2 gravity. In section 4, we show that the spontaneous breaking
+of the BRST symmetry yields an Einstein action and that there is a quantum anomaly in
+both the unbroken and broken sectors. We conclude with section 5 where we summarize our
+results, provide further physical insights and discuss directions for future work. We relegate
+to Appendix A some technical details on the symmetry of the Einstein action.
+2
+Pure R2 gravity plus a massless scalar: from restricted to
+on-shell Weyl invariance
+The action of pure R2 gravity is given by
+S =
+� √−g d4x α R2
+(1)
+where R is the Ricci scalar and α a dimensionless constant. This action is restricted Weyl
+invariant i.e. it is invariant under the Weyl transformation gµν → Ω2(x) gµν if the conformal
+factor Ω(x) obeys the condition □Ω(x) = 0. This invariance stems from the fact that R →
+R/Ω2 when □Ω(x) = 0. As already mentioned, this implies that pure R2 gravity has a greater
+symmetry than global scale symmetry (where Ω(x) would have to be a constant).
+We now show that pure R2 gravity can be Weyl-invariant on-shell when a minimally coupled
+real massless scalar field is added to the action. Here, the condition □Ω(x) = 0 is not imposed
+as an external condition but satisfied automatically by the equations of motion. The action of
+pure R2 gravity with a minimally coupled real massless scalar field φ is given by
+Sa =
+� √−g d4x
+�
+α R2 − 1
+2 gµν ∂µφ ∂νφ
+�
+(2)
+where φ(x) is a real scalar field. Under the Weyl transformation gµν → e−2 ǫ φ gµν, where ǫ is
+3
+
+a real constant, the Ricci scalar transforms as
+R → R e2ǫφ − 6 e3ǫ φ □(e−ǫ φ)
+(3)
+and √−g → e−4 ǫ φ √−g so that action (2) transforms to
+Sb =
+� √−g d4x
+�
+α
+�
+R2 − 12 R eǫ φ □(e−ǫ φ) + 36 e2ǫ φ (□(e−ǫ φ))2�
+− 1
+2 e−2 ǫ φ gµν ∂µφ ∂νφ
+�
+.
+(4)
+The equations of motion yield □(e−ǫ φ) = 0. Therefore, when the equations of motion are
+satisfied, the above action reduces to
+Sc =
+� √−g d4x
+�
+α R2 − 1
+2 gµν ∂µψ ∂νψ
+�
+(5)
+where ψ is a real massless scalar field (related to the old scalar φ via ψ = e−ǫ φ/ǫ). Note that
+the equation of motion for ψ is □ψ = 0 which is equivalent to □(e−ǫ φ) = 0 and consistent with
+what we previously obtained. We therefore recover pure R2 gravity with a minimally coupled
+real massless scalar field ψ. What happened here is that the restricted Weyl condition □ Ω = 0
+with Ω = e−ǫ φ did not have to be imposed as a separate condition because it was satisfied
+automatically by the equations of motion. In short, pure R2 gravity became Weyl invariant
+on-shell in the presence of a massless scalar field. It passed from restricted Weyl invariance to
+on-shell Weyl invariance.
+3
+BRST invariance of pure R2 gravity
+Before discussing BRST invariance in the case of pure R2 gravity, let us first recall how BRST
+invariance works in gauge theories in Minkowski spacetime. For illustrative purposes, we will
+consider the case of scalar QED. The Abelian version of the Faddeev-Popov Lagrangian is then
+given by [12]
+L = −1
+4 F 2
+µν − (Dµφ∗
+a)(Dµφa) − m2 φ∗
+a φa − 1
+2 ξ (∂µ Aµ)2 + ¯c □c
+(6)
+where c(x) and ¯c(x) are independent Grassmann-valued fields, φa are a set of complex scalar
+fields and Dµ is the usual covariant derivative. The gauge fixing term,
+1
+2 ξ(∂µ Aµ)2 breaks the
+gauge symmetry since it is not invariant under the transformation Aµ → Aµ + 1
+e ∂µf(x) where
+f(x) is an arbitrary function. However, it has a residual symmetry: it is invariant if f(x)
+obeys the condition □f = 0. As previously mentioned, this residual symmetry is the analog of
+restricted Weyl symmetry in pure R2 gravity.
+4
+
+The equation of motion for c(x) is □c = 0. Consider the gauge transformation with f(x) =
+θ c(x) for arbitrary Grassmann number θ. Then, if the equation of motion for c is satisfied,
+the scalar QED Lagrangian (6) is invariant under the following transformations
+Aµ → Aµ + 1
+e θ ∂µc(x)
+φa(x) → eiθ c(x) φa(x) = φa(x) + iθ c(x)φa(x) .
+(7)
+In other words, the equation □f = θ □c = 0 is automatically satisfied on-shell and does not
+have to be imposed as a separate condition. This is similar to what we saw in the previous
+section for pure R2 gravity which was invariant under gµν → Ω2 gµν with Ω = e−ǫφ when the
+equations of motion were satisfied.
+If the equation of motion for c is not used, the only term in the Lagrangian (6) which is not
+invariant under the transformation (7) is (∂µAµ)2 which transforms as
+(∂µAµ)2 → (∂µAµ)2 + 2
+e(∂µAµ)(θ□c)
+(8)
+where we used the fact that θ2 = 0 since θ is Grassmann. Now, if under (7) we also have ¯c
+transforming as
+¯c(x) → ¯c(x) − θ
+e ξ (∂µAµ)
+(9)
+then the scalar QED Lagrangian (6) is invariant without having to use the equation of motion
+for c. This is BRST invariance. The crucial point is that under the BRST transformations
+given by (7) and (9), the Lagrangian is invariant despite the presence of the gauge-fixing term
+(∂µAµ)2.
+We now turn to pure R2 gravity. Consider the action
+S =
+�
+d4x√−g (α R2 + ¯c □c)
+(10)
+where again c(x) and ¯c(x) are independent Grassmann-valued fields. This action is not Weyl-
+invariant i.e. it is not invariant under the transformation gµν → Ω2(x)gµν where Ω(x) is an
+arbitrary smooth function. Consider now the Weyl transformation
+gµν → e2 θ c(x)gµν = (1 + 2 θ c) gµν
+(11)
+where θ is again an arbitrary Grassmann number. Under this transformation we have
+√−g α R2 → √−g (α R2 − 12 α R θ □c )
+(12)
+5
+
+where the following transformations were used: √−g → (1+4 θ c) √−g and R → (1−2 θ c) R−
+6 θ □c. Again, we used that θ2 = 0. Under the transformation (11), □c transforms as
+□c → (1 − 2 θ c) □c
+(13)
+where gµν∂µc ∂νc = 0 was used (this stems from the fact that gµν is symmetric and c is
+Grassmann). The equation of motion for c is □c = 0 and we see from (12) that √−g α R2 is
+Weyl invariant on-shell. However, we can dispense with the on-shell condition if we also allow
+¯c to transform as
+¯c → (1 − 2 θ c) ¯c + 12 α R θ .
+(14)
+We then obtain
+√−g ¯c □c → √−g (¯c □c + 12 α R θ □c) .
+(15)
+The last term on the right hand side of (15) above cancels precisely the last term on the right
+hand side of (12). Therefore, the action (10) is invariant under the combined transformations
+of (11) and (14) (which we refer to to as BRST transformations).
+This is the BRST invariance
+of pure R2 gravity. Note that BRST invariance does not require any on-shell or restricted Weyl
+condition. It is a generalization of Weyl (conformal) invariance that is valid in the presence of
+the Weyl-breaking R2 term.
+Let us now take a closer look at what is common and what is different between the BRST
+invariance of pure R2 gravity and the BRST invariance in the gauge theories of particle physics
+(for concreteness and simplicity, we will consider scalar QED again but the main points apply
+also to QCD). The BRST invariance in scalar QED can be viewed as a generalization of gauge
+invariance in the presence of the gauge-fixing (and hence gauge-breaking) term (∂µ Aµ)2. The
+are two points in common between the scalar QED and R2 cases. First, the Ricci scalar R under
+a Weyl transformation and the term ∂µ Aµ under a gauge transformation both pick up an extra
+□Φ(x) term (where Φ(x) represents either a conformal factor Ω(x) in a Weyl transformation
+or a function f(x) in a gauge transformation). Recall that in a BRST transformation, Φ(x) is
+a product of a Grassmann number θ with a Grassmann field (the product yields a commuting
+(bosonic) quantity). The second point in common is that R and ∂µ Aµ are both squared. The
+squaring yields a (□Φ(x))2 term which is zero since θ2 = 0. The squaring still leaves one
+extra □Φ(x) term and this is cancelled out in both cases via the transformation property of a
+Grassmann field. These two common points render the analogy between the two cases quite
+strong. However, there is one important difference. In scalar QED (and in QCD) , the BRST
+transformations are associated with gauge transformations. The BRST invariance of pure R2
+gravity that we are considering here is not associated with gauge transformations (such as
+diffeomorphisms) but with Weyl (local scale) transformations. We will see that this difference
+plays an important role when the theory is quantized.
+6
+
+4
+Spontaneous breaking of BRST symmetry
+We now show that the BRST-invariant action
+S =
+�
+d4x√−g (αR2 + ¯c □c)
+(16)
+is conformally equivalent to an action that involves the Einstein-Hilbert term; this will involve
+the spontaneous breaking of BRST symmetry. The starting point is to introduce a auxiliary
+field σ(x) to rewrite the above action into the equivalent form
+S1 =
+�
+d4x√−g (−α(b σ + R)2 + αR2 + ¯c □c)
+�
+d4x√−g (−α b2 σ2 − 2 α b R σ + ¯c □c)
+(17)
+where b is a real non-zero constant with dimensions of mass squared and σ(x) is dimensionless.
+Action (17) is equivalent to the original action (16) since adding the squared term in the first
+line of (17) does not alter anything (classically, the equations of motion are unaffected and
+quantum mechanically, the path integral over σ is a Gaussian which yields a constant). The
+equivalent action (17) is also BRST invariant; it is invariant under the following transforma-
+tions:
+gµν → (1 + 2 θ c) gµν
+;
+¯c → (1 − 2 θ c) ¯c − 12 θ α b σ
+;
+σ → (1 − 2θ c) σ
+(18)
+where θ is again a Grassmann number. Note that the BRST invariance requires the auxiliary
+field σ to transform besides the fields gµν and ¯c. We now perform the following conformal
+(Weyl) transformation:
+gµν → σ−1 gµν
+¯c → σ ¯c
+(19)
+which leads to √−g → σ−2 √−g and R → σ R − 6 σ3/2□(σ−1/2). Under the above conformal
+transformation, action (17) becomes
+S2 =
+�
+d4x√−g (−α b2 − 2 α b R + 3α b
+σ2 ∂µσ ∂µσ + ¯c □c − 1
+σ ¯c ∂µc ∂µσ) .
+(20)
+The above action is no longer invariant under the BRST transformations given by (18). The
+BRST symmetry has been spontaneously broken. The factor σ−1 appearing in the confor-
+mal transformation (19) is valid only for non-zero σ so that the VEV (vacuum expectation
+value) of the field σ must be non-zero. The VEV is therefore not invariant under the BRST
+transformation σ → (1 − 2θ c) σ leading to the spontaneous breaking of the BRST symmetry.
+7
+
+We can identify −2 α b R as an Einstein-Hilbert term if we equate −2 α b with
+1
+16π G where G
+is Newton’s constant. The constant term −α b2 can then be associated with a cosmological
+constant Λ = −b/4. Note that though −2 α b is positive, the constant b can be either positive
+or negative (but not zero). This implies that the cosmological constant can be either positive
+corresponding to a de Sitter (dS) background or negative corresponding to an anti-de Sitter
+(AdS) background but it cannot be identically zero. We can then express (20) as the following
+Einstein action,
+SE =
+�
+d4x√−g
+�
+1
+16π G(R − 2 Λ) + 3α b
+σ2 ∂µσ ∂µσ + ¯c □c − 1
+σ ¯c ∂µc ∂µσ)
+�
+.
+(21)
+We have left the constant 3 α b in the action for simplicity but it is not an independent constant;
+it is equal to
+−3
+32 πG. We therefore obtain an Einstein-Hilbert action with non-zero cosmological
+constant, a kinetic term for the scalar σ (which we will express in canonical form later) and
+an interaction term.
+Recall that σ is non-zero so that divisions by σ pose no issue.
+It is
+well-known that in spontaneously broken theories, the vacuum breaks the symmetry but it is
+not actually broken in the Lagrangian but manifested or realized in a different way [12]. It
+can be directly verified (see Appendix A) that the Einstein action (21) is invariant under the
+following transformations:
+σ → (1 − 2θ c) σ , gµν → gµν and ¯c → ¯c − 12 θ α b .
+(22)
+The BRST symmetry of action (17) manifests itself in the Einstein action (21) via its symmetry
+under the above transformations (22). We now show how transformation (22) stems from the
+BRST transformations (18). In the Einstein action and transformation (22) label the metric
+and the barred Grassmann field with a subscript E i.e.
+gµνE and ¯cE.
+In action (17) and
+transformation (18) we leave gµν and ¯c as is. Then the conformal transformation (19) yields
+gµνE = σ gµν and ¯cE = σ−1 ¯c. Under the BRST transformations (18) we obtain gµνE = σ gµν →
+(1 − 2 θ c) σ (1 + 2 θ c) gµν = σ gµν = gµνE and ¯cE = σ−1 ¯c → (1 + 2 θ c) σ−1�
+(1 − 2 θ c) ¯c −
+12 θ α b σ
+�
+= σ−1¯c − 12 θ α b = ¯cE − 12 θ α b. We have therefore obtained the transformations
+gµνE → gµνE and ¯cE → ¯cE − 12 θ α b which correspond to those in (22). Note that we used
+σ → (1 − 2θ c) σ in (18) to derive this, so the transformation of σ is also part of (22).
+We can define a real massless scalar field ψ(x) =
+√
+−3α b ln σ(x) so that the kinetic term for
+σ is expressed in canonical form. The Einstein action (21) expressed in terms of the field ψ is
+S =
+�
+d4x√−g
+�
+1
+16π G(R − 2 Λ) − ∂µψ ∂µψ + ¯c □c −
+1
+√
+−3α b ¯c ∂µc ∂µψ
+�
+.
+(23)
+The massless scalar field ψ corresponds to the Nambu-Goldstone boson of the broken sector.
+Under transformation (22), the field ψ transforms as a shift ψ → ψ−
+√
+−3α b 2 θ c (whereas ¯c →
+¯c−12 θ α b and gµν → gµν). The above action (23) is invariant under those transformations (see
+8
+
+Appendix A). This is in accord with what we expect from spontaneously broken theories: the
+original symmetry in the Lagrangian manifests itself in the broken sector as a shift symmetry
+of the Goldstone bosons [12].
+4.1
+Quantum anomaly
+We saw that the action (17) is BRST invariant under the following transformations:
+gµν → (1 + 2 θ c) gµν , ¯c → (1 − 2 θ c) ¯c − 12 θ α b σ , σ → (1 − 2θ c) σ. Each transformation
+involves a Weyl transformation where the conformal factor is expressed in terms of of a prod-
+uct of two Grassmann variables The BRST symmetry is therefore a generalization of Weyl
+(conformal) symmetry. After quantization, renormalization introduces a scale which breaks
+the BRST symmetry since it automatically breaks Weyl symmetry (leading to the well-known
+Weyl (conformal) anomaly). So the BRST symmetry of pure R2 gravity is anomalous. This is
+in contrast to the BRST invariance of gauge theories like QCD which have no anomaly.
+After the BRST symmetry is spontaneously broken and we obtain the Einstein action (21),
+we saw that the BRST symmetry manifests itself now in the Einstein action as a symmetry
+under the transformations (22). This symmetry is also anomalous since the transformation of
+the field σ is a Weyl transformation and renormalization breaks this symmetry (leading again
+to the Weyl (conformal anomaly). Another way to see this is to note that the only fields that
+transform in (22) are ¯c and σ. The transformation for ¯c is simply a constant shift so that its
+path integral measure D¯c is invariant. However, σ undergoes a Weyl transformation and this
+introduces a non-trivial Jacobian J (i.e. J ̸= 1) to the measure Dσ. Since the measure is not
+invariant, this implies there is an anomaly [17]. So the symmetry in the unbroken phase and
+its associated symmetry in the broken phase are both anomalous. Our finding here is in accord
+with previous work that shows that when the Weyl or conformal symmetry is spontaneously
+broken there is conformal anomaly matching between the unbroken and broken phases [18,19].
+5
+Conclusion
+In the last six years or so, we have kept discovering new aspects of pure R2 gravity. A non-
+exhaustive list includes its unitarity among quadratic gravity theories [4], its conformal equiv-
+alence to Einstein gravity with non-zero cosmological constant and massless scalar field [1–5],
+its restricted Weyl symmetry [7,10,11], its spontaneous symmetry breaking to Einstein grav-
+ity [3,5] and the lack of a propagating graviton when the theory is linearized about a Minkowski
+background [4] (where there is no Einstein equivalence since the cosmological constant is zero).
+In this paper, we have gained further insights into this theory. We saw that pure R2 gravity
+has an analog with the gauge-fixing term (∂µAµ)2 in gauge theory. R2 is not invariant under
+9
+
+the Weyl transformation gµν → Ω2(x) gµν just like (∂µ Aµ)2 is not invariant under the gauge
+transformation Aµ → Aµ + 1
+e ∂µf(x). However, each have a residual symmetry (when □Ω = 0
+is satisfied in the gravity case and □f = 0 is satisfied in the gauge case). This analogy opened
+the door towards enlarging the symmetry of pure R2 gravity to include BRST symmetry.
+We first showed that when a massless scalars field was included in the pure R2 action, the
+condition □Ω = 0 could be met automatically when the equations of motion were satisfied
+i.e. we went from restricted Weyl to on-shell Weyl invariance. Finally, we obtained the BRST
+invariance of pure R2 gravity where no restricted Weyl or on-shell condition is required. The
+BRST transformations involve Weyl transformations where the conformal factor is composed of
+products of Grassmann variables (the conformal factor itself is commutative). The important
+point is that the BRST invariance exists despite the Weyl-breaking R2 term.
+There is one important difference between the BRST symmetry in gauge theories like QCD
+and the BRST symmetry that we have considered here for pure R2 gravity. Gauge invari-
+ance in particle physics is preserved after quantization. The BRST invariance of QCD is a
+generalization of gauge invariance so that it is also preserved after quantization; there is no
+anomaly. In contrast to gauge symmetry, global scale or Weyl (local scale) symmetry is broken
+after quantization since renormalization introduces a scale. The BRST symmetry of pure R2
+gravity is a generalization of Weyl (conformal) symmetry so that it is also broken after quan-
+tization leading to the well-known Weyl (conformal) anomaly. After the spontaneous breaking
+of the BRST symnmetry, we obtained an Einstein action. We showed that this action has
+its own symmetry and that it is also anomalous. This is in accord with previous work that
+shows that when the Weyl (conformal) symmetry is spontaneously broken there is conformal
+anomaly matching between the unbroken and broken sectors [18,19].
+The focus of this paper was pure R2 gravity because of its many special and attractive fea-
+tures that we previously mentioned. All other quadratic gravity theories (like Weyl-squared,
+Riemann-squared, etc.), apart from boundary terms, can be expressed as a linear combina-
+tion of R2 and RµνRµν. The latter term, the square of the Ricci tensor, appears in quantum
+corrections to General Relativity (GR) and even though it does not constitute a valid UV
+completion of GR due to its non-unitarity (yields a massive spin two ghost [4, 20]), it still
+makes a well-known calculable short-range correction to the Newtonian potential [12,21]. Like
+R2, the term RµνRµν is not Weyl-invariant so it would be of interest to see if it can be BRST
+invariant. It is not in the form of a scalar squared like (∂µAµ)2 or R2, so one may be inclined
+to think that the BRST formalism would not apply here. However, like pure R2, it was shown
+in [7] that RµνRµν is restricted Weyl invariant (up to a boundary term). This suggests that
+the procedure used to establish the BRST invariance of pure R2 gravity might in the end also
+work for this quadratic theory. It would therefore be worthwhile and interesting to investigate
+this further.
+10
+
+Acknowledgments
+A.E. acknowledges support from a discovery grant of the National Science and Engineering
+Research Council of Canada (NSERC).
+A
+Symmetry of Einsten Action
+In this appendix we show that the Einstein action (21) given by
+SE =
+�
+d4x√−g
+�
+1
+16π G(R − 2 Λ) + 3α b
+σ2 ∂µσ ∂µσ + ¯c □c − 1
+σ ¯c ∂µc ∂µσ)
+�
+(A.1)
+is invariant under the transformations (22) given by
+σ → (1 − 2θ c) σ , gµν → gµν and ¯c → ¯c − 12 θ α b .
+(A.2)
+Under the above transformation, the metric gµν does not change so that √−g as well as the
+term √−g
+1
+16π G(R − 2 Λ) does not change.
+The other terms in the above Einstein action
+transform as
+3α b
+σ2 ∂µσ ∂µσ → 3α b
+σ2 ∂µσ ∂µσ − 12 θ α b
+σ
+∂µc ∂µσ
+− 1
+σ ¯c ∂µc ∂µσ) → − 1
+σ ¯c ∂µc ∂µσ + 12 θ α b
+σ
+∂µc ∂µσ
+¯c □c → ¯c □c − 12 θ α b □c
+(A.3)
+where we used that θ2 = 0 (since θ is a Grassmann number) and that gµν ∂µc ∂νc = 0 since gµν
+is symmetric and c(x) and its derivatives are Grassmann fields. We see that the extra term
+− 12 θ α b
+σ
+∂µc ∂µσ in the first line of (A.3) is canceled exactly by the extra term in the second
+line. The extra term in the third line, −12 θ α b □c, where −12 θ α b is a constant, does not
+cancel out with any other extra term in (A.3). However, √−g □c is a total derivative that
+yields an inconsequential boundary term in the action. We have therefore shown that action
+(A.1) is invariant under transformations (A.2).
+We saw in section 4 that the Einstein action (A.1) can be expressed in terms of a real massless
+scalar field ψ(x) =
+√
+−3α b ln σ(x) as action (23):
+S =
+�
+d4x√−g
+�
+1
+16π G(R − 2 Λ) − ∂µψ ∂µψ + ¯c □c −
+1
+√
+−3α b ¯c ∂µc ∂µψ
+�
+(A.4)
+where ψ was identified as the Nambu-Goldstone boson of the broken sector. We stated in
+section 4 that the action (A.4) was invariant under the following transformations:
+ψ → ψ −
+√
+−3α b 2 θ c , ¯c → ¯c − 12 θ α b and gµν → gµν .
+(A.5)
+11
+
+We now verify this statement. Under (A.5) the last three terms in action (A.4) transform as:
+− ∂µψ ∂µψ → −∂µψ ∂µψ + 4 θ
+√
+−3 α b ∂µψ ∂µc
+−
+1
+√
+−3α b ¯c ∂µc ∂µψ → −
+1
+√
+−3α b ¯c ∂µc ∂µψ − 4 θ
+√
+−3 α b ∂µψ ∂µc
+¯c □c → ¯c □c − 12 θ α b □c .
+(A.6)
+We see that the extra term +4 θ
+√
+−3 α b ∂µψ ∂µc in the first line above is cancelled by the
+extra term on the second line which is equal to its negative.
+The only extra term that is
+not cancelled is the term −12 θ α b □c appearing in the last line. However, √−g □c is a total
+derivative which yields a boundary term with no physical consequence. We have therefore
+verified that the Einstein action (A.4) is indeed invariant under the transformations (A.5).
+References
+[1] C. Kounnas, D. L¨ust and N. Toumbas, R2 inflation from scale invariant supergravity and
+anomaly free superstrings with fluxes, Fortsch. Phys. 63, 12 (2015) [arXiv:1409.7076].
+[2] A. Kehagias, C. Kounnas, D. L¨ust and A. Riotto, Black Hole Solutions in R2 Gravity,
+JHEP 05, 143 (2015)[arXiv:1502.04192].
+[3] A. Edery and Y. Nakayama, Generating Einstein gravity, cosmological constant and
+Higgs mass from restricted Weyl invariance, Mod. Phys. Lett. A 30, 1550152 (2015)
+[arXiv:1502.05932].
+[4] L. Alvarez-Gaume, A. Kehagias, C. Kounnas, D. L¨ust and A. Riotto, Aspects of Quadratic
+Gravity, Fortsch. Phys. 64, 176 (2016) [arXiv:1505.07657].
+[5] A. Edery and Y. Nakayama, Gravitating magnetic monopole via the spontaneous symmetry
+breaking of pure R2 gravity, Phys. Rev. D 98 (2018) 064011 [arXiv:1807.07004].
+[6] A. Edery and Y. Nakayama, Palatini formulation of pure R2 gravity yields Einstein gravity
+with no massless scalar, Phys. Rev. D 99, 124018 (2019) [arXiv:1902.07876].
+[7] A. Edery and Y. Nakayama, Restricted Weyl invariance in four-dimensional curved space-
+time, Phys. Rev. D 90, 043007 (2014) [arXiv:1406.0060].
+[8] A. Edery, Pure R2 gravity can gravitate about a flat background, J. Phys. Conf. Ser. 1956,
+012005 (2021).
+[9] A. Edery and Y. Nakayama, Critical gravity from four dimensional scale invariant gravity,
+JHEP 11, 169 (2019) [arXiv:1908.08778].
+12
+
+[10] I. Oda, Restricted Weyl symmetry, Phys. Rev. D 102, 045008 (2020)[arXiv:2005.04771].
+[11] I. Oda, Restricted Weyl Symmetry and Spontaneous Symmetry Breakdown of Conformal
+Symmetry, Mod. Phys. Lett. A 36, 2150203 (2021) [arXiv:2104.04694].
+[12] M. Schwartz, Quantum Field Theory and the Standard Model, (Cambridge University
+Press, Cambridge, UK, 2014).
+[13] I. Oda and P. Saake, BRST formalism of Weyl conformal gravity, Phys. Rev. D 106,
+106007 (2022) [arXiv:2209.14533].
+[14] D. Prinz, The BRST Double Complex for the Coupling of Gravity to Gauge Theories,
+[arXiv:2206.00780].
+[15] T. Kugo, R. Nakayama and N.Ohta, Covariant BRST quantization of unimodular grav-
+ity: Formulation with antisymmetric tensor ghosts, Phys. Rev. D 105, 086006 (2022),
+[arXiv:2202.03626].
+[16] L. Berezhiani, G. Dvali and O. Sakhelashvili, de Sitter space as a BRST invariant coherent
+state of gravitons Phys. Rev. D 105, 025022 (2022) [arXiv:2111.12022].
+[17] K. Fujikawa and H. Suzuki, Path Integrals and Quantum Anomalies, (Oxford University
+Press, Oxford, UK, 2004).
+[18] A. Schwimmer and S. Theisen, Spontaneous Breaking of Conformal Invariance and Trace
+Anomaly Matching,Nucl. Phys. B 847, 590 (2011) [arXiv:1011.0696.
+[19] A. Cabo-Bizet, E. Gava and K. S. Narain, Holography and Conformal Anomaly Matching,
+JHEP 11,044 (2013) [arXiv:1307.3784].
+[20] K. S. Stelle, Renormalization of Higher Derivative Quantum Gravity, Phys. Rev. D 16,
+953 (1977).
+[21] J.F. Donoghue, Leading quantum correction to the Newtonian potential, Phys. Rev. Lett.
+72, 2996 (1994).
+13
+
diff --git a/8tFAT4oBgHgl3EQfpB3g/content/tmp_files/load_file.txt b/8tFAT4oBgHgl3EQfpB3g/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0d81c4da6908d11af36e8961a731b0a375644416
--- /dev/null
+++ b/8tFAT4oBgHgl3EQfpB3g/content/tmp_files/load_file.txt
@@ -0,0 +1,355 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf,len=354
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='08638v1 [hep-th] 20 Jan 2023 Enlarging the symmetry of pure R2 gravity, BRST invariance and its spontaneaous breaking Ariel Edery∗ Department of Physics and Astronomy, Bishop’s University, 2600 College Street, Sherbrooke, Qu´ebec, Canada, J1M 1Z7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Abstract Pure R2 gravity was considered originally to possess only global scale symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It was later shown to have the larger restricted Weyl symmetry where it is invariant under the Weyl transformation gµν → Ω2(x) gµν when the conformal factor Ω(x) obeys the harmonic condition □Ω(x) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Restricted Weyl symmetry has an analog in gauge theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under a gauge transformation Aµ → Aµ + 1 e∂µf(x), the gauge-fixing term (∂µAµ)2 has a residual gauge symmetry when □f = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In this paper, we consider scenarios where the symmetry of pure R2 gravity can be enlarged even further.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In one scenario, we add a massless scalar field to the pure R2 gravity action and show that the action becomes on-shell Weyl invari- ant when the equations of motion are obeyed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We then enlarge the symmetry to a BRST symmetry where no on-shell or restricted Weyl condition is required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST trans- formations here are not associated with gauge transformations (such as diffeomorphisms) but with Weyl (local scale) transformations where the conformal factor consists of a prod- uct of Grassmann variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' BRST invariance in this context is a generalization of Weyl invariance that is valid in the presence of the Weyl-breaking R2 term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In contrast to the BRST invariance of gauge theories like QCD, it is not preserved after quantization since renormalization introduces a scale (leading to the well-known Weyl (conformal) anomaly).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We show that the spontaneous breaking of the BRST symmetry yields an Einstein action;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' this still has a symmetry which is also anomalous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is in accord with previous work that shows that there is conformal anomaly matching between the unbroken and broken phases when conformal symmetry is spontaneously broken.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' ∗aedery@ubishops.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='ca 1 1 Introduction Pure R2 gravity (R2 alone with no additional R term) is unique among quadratic gravity the- ories as it is unitary and moreover has been shown to be conformally equivalent to Einstein gravity with non-zero cosmological constant and massless scalar field [1–5] (though in a Pala- tini formalism one can avoid having a massless scalar [6]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It has been known for a long time that it is invariant under the global scale transformation gµν → λ2 gµν where λ is a constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It was later discovered to possess a larger symmetry than global scale symmetry called restricted Weyl symmetry [7] where it is invariant under the transformation gµν → Ω2(x) gµν when the conformal factor Ω(x) obeys the harmonic condition □Ω = gµν∇µ∇νΩ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The conformal factor Ω(x) is therefore not limited to being a constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The aforementioned equivalence be- tween pure R2 gravity and Einstein gravity with cosmological constant was then interpreted in a new light: it occurs when the restricted Weyl symmetry is spontaneously broken [3,5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In the broken sector, the Ricci scalar of the background (vacuum) spacetime has R ̸= 0 which excludes a flat background.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is why the equivalence requires a non-zero cosmological con- stant on the the Einstein side.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The unbroken sector which has an R = 0 vacuum (background) has no relation to Einstein gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In fact, it has been shown that a linearization of pure R2 gravity about Minkowski spacetime does not yield gravitons but only a propagating scalar [4];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' simply put, pure R2 gravity does not gravitate about a flat background [4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, it was later shown that if one includes a non-minimally coupled scalar field in the restricted Weyl- invariant action and the field acquires a non-zero VEV, then the theory can gravitate about flat spacetime [5,8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Various aspects of restricted Weyl symmetry, it spontaneous breaking as well as its role in critical gravity were then explored further in [5–7,9–11] Restricted Weyl symmetry has an analog in gauge theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The gauge-fixing term (∂µAµ)2 is invariant under the gauge transformation Aµ → Aµ + 1 e∂µf(x) only when the arbitrary smooth function f(x) obeys the condition □f = 0 where □ here represents the flat space d’Alembertian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Therefore, the gauge-fixing term has a residual gauge symmetry when □f = 0 is satisfied [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is the analog to the restricted Weyl symmetry of pure R2 gravity when the conformal factor Ω(x) satisfies □Ω = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' As we will see, this analogy is fruitful as it provides a bridge to the BRST symmetry of pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Recent work on the BRST invariance of other gravitational theories can be found in [13,14,16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In this paper, we consider scenarios where the symmetry of pure R2 gravity is enlarged further.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We show that when a massless scalar field is added to pure R2 gravity, the action becomes Weyl invariant when the equations of motion are satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' No separate external condition is required to be imposed on the conformal factor Ω(x) as this occurs naturally via the equations of motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' One passes from restricted Weyl invariance to on-shell Weyl invariance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' One can then enlarge the symmetry further to include BRST symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In analogy with the BRST invariance in gauge theories in the presence of a gauge-fixing term, we establish BRST invariance in the 2 presence of the Weyl-breaking pure R2 gravity term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST transformations here are not associated with gauge transformations (such as diffeomorphisms) but are a generalization of Weyl (local scale) transformations where the conformal factor is composed of Grassmann variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Therefore, in contrast to the BRST invariance in gauge theory, it is anomalous since renormalization introduces a scale (leading to the well-known Weyl (conformal) anomaly).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We show that the spontaneous breaking of the BRST symmetry yields an Einstein action with its own symmetry that is also anomalous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is in agreement with previous work where it was shown that when conformal symmetry is spontaneously broken there is conformal anomaly matching in the unbroken and and broken phases [18,19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The paper is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In section 2, we obtain the on-shell Weyl invariance of pure R2 gravity when a massless scalar field is included in the action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In section 3, we obtain the BRST invariance of pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In section 4, we show that the spontaneous breaking of the BRST symmetry yields an Einstein action and that there is a quantum anomaly in both the unbroken and broken sectors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We conclude with section 5 where we summarize our results, provide further physical insights and discuss directions for future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We relegate to Appendix A some technical details on the symmetry of the Einstein action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 2 Pure R2 gravity plus a massless scalar: from restricted to on-shell Weyl invariance The action of pure R2 gravity is given by S = � √−g d4x α R2 (1) where R is the Ricci scalar and α a dimensionless constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This action is restricted Weyl invariant i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' it is invariant under the Weyl transformation gµν → Ω2(x) gµν if the conformal factor Ω(x) obeys the condition □Ω(x) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This invariance stems from the fact that R → R/Ω2 when □Ω(x) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' As already mentioned, this implies that pure R2 gravity has a greater symmetry than global scale symmetry (where Ω(x) would have to be a constant).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We now show that pure R2 gravity can be Weyl-invariant on-shell when a minimally coupled real massless scalar field is added to the action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Here, the condition □Ω(x) = 0 is not imposed as an external condition but satisfied automatically by the equations of motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The action of pure R2 gravity with a minimally coupled real massless scalar field φ is given by Sa = � √−g d4x � α R2 − 1 2 gµν ∂µφ ∂νφ � (2) where φ(x) is a real scalar field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under the Weyl transformation gµν → e−2 ǫ φ gµν, where ǫ is 3 a real constant, the Ricci scalar transforms as R → R e2ǫφ − 6 e3ǫ φ □(e−ǫ φ) (3) and √−g → e−4 ǫ φ √−g so that action (2) transforms to Sb = � √−g d4x � α � R2 − 12 R eǫ φ □(e−ǫ φ) + 36 e2ǫ φ (□(e−ǫ φ))2� − 1 2 e−2 ǫ φ gµν ∂µφ ∂νφ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (4) The equations of motion yield □(e−ǫ φ) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Therefore, when the equations of motion are satisfied, the above action reduces to Sc = � √−g d4x � α R2 − 1 2 gµν ∂µψ ∂νψ � (5) where ψ is a real massless scalar field (related to the old scalar φ via ψ = e−ǫ φ/ǫ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Note that the equation of motion for ψ is □ψ = 0 which is equivalent to □(e−ǫ φ) = 0 and consistent with what we previously obtained.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We therefore recover pure R2 gravity with a minimally coupled real massless scalar field ψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' What happened here is that the restricted Weyl condition □ Ω = 0 with Ω = e−ǫ φ did not have to be imposed as a separate condition because it was satisfied automatically by the equations of motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In short, pure R2 gravity became Weyl invariant on-shell in the presence of a massless scalar field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It passed from restricted Weyl invariance to on-shell Weyl invariance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 3 BRST invariance of pure R2 gravity Before discussing BRST invariance in the case of pure R2 gravity, let us first recall how BRST invariance works in gauge theories in Minkowski spacetime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' For illustrative purposes, we will consider the case of scalar QED.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The Abelian version of the Faddeev-Popov Lagrangian is then given by [12] L = −1 4 F 2 µν − (Dµφ∗ a)(Dµφa) − m2 φ∗ a φa − 1 2 ξ (∂µ Aµ)2 + ¯c □c (6) where c(x) and ¯c(x) are independent Grassmann-valued fields, φa are a set of complex scalar fields and Dµ is the usual covariant derivative.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The gauge fixing term, 1 2 ξ(∂µ Aµ)2 breaks the gauge symmetry since it is not invariant under the transformation Aµ → Aµ + 1 e ∂µf(x) where f(x) is an arbitrary function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, it has a residual symmetry: it is invariant if f(x) obeys the condition □f = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' As previously mentioned, this residual symmetry is the analog of restricted Weyl symmetry in pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 4 The equation of motion for c(x) is □c = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Consider the gauge transformation with f(x) = θ c(x) for arbitrary Grassmann number θ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Then, if the equation of motion for c is satisfied, the scalar QED Lagrangian (6) is invariant under the following transformations Aµ → Aµ + 1 e θ ∂µc(x) φa(x) → eiθ c(x) φa(x) = φa(x) + iθ c(x)φa(x) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (7) In other words, the equation □f = θ □c = 0 is automatically satisfied on-shell and does not have to be imposed as a separate condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is similar to what we saw in the previous section for pure R2 gravity which was invariant under gµν → Ω2 gµν with Ω = e−ǫφ when the equations of motion were satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' If the equation of motion for c is not used, the only term in the Lagrangian (6) which is not invariant under the transformation (7) is (∂µAµ)2 which transforms as (∂µAµ)2 → (∂µAµ)2 + 2 e(∂µAµ)(θ□c) (8) where we used the fact that θ2 = 0 since θ is Grassmann.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Now, if under (7) we also have ¯c transforming as ¯c(x) → ¯c(x) − θ e ξ (∂µAµ) (9) then the scalar QED Lagrangian (6) is invariant without having to use the equation of motion for c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is BRST invariance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The crucial point is that under the BRST transformations given by (7) and (9), the Lagrangian is invariant despite the presence of the gauge-fixing term (∂µAµ)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We now turn to pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Consider the action S = � d4x√−g (α R2 + ¯c □c) (10) where again c(x) and ¯c(x) are independent Grassmann-valued fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This action is not Weyl- invariant i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' it is not invariant under the transformation gµν → Ω2(x)gµν where Ω(x) is an arbitrary smooth function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Consider now the Weyl transformation gµν → e2 θ c(x)gµν = (1 + 2 θ c) gµν (11) where θ is again an arbitrary Grassmann number.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under this transformation we have √−g α R2 → √−g (α R2 − 12 α R θ □c ) (12) 5 where the following transformations were used: √−g → (1+4 θ c) √−g and R → (1−2 θ c) R− 6 θ □c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Again, we used that θ2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under the transformation (11), □c transforms as □c → (1 − 2 θ c) □c (13) where gµν∂µc ∂νc = 0 was used (this stems from the fact that gµν is symmetric and c is Grassmann).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The equation of motion for c is □c = 0 and we see from (12) that √−g α R2 is Weyl invariant on-shell.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, we can dispense with the on-shell condition if we also allow ¯c to transform as ¯c → (1 − 2 θ c) ¯c + 12 α R θ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (14) We then obtain √−g ¯c □c → √−g (¯c □c + 12 α R θ □c) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (15) The last term on the right hand side of (15) above cancels precisely the last term on the right hand side of (12).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Therefore, the action (10) is invariant under the combined transformations of (11) and (14) (which we refer to to as BRST transformations).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is the BRST invariance of pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Note that BRST invariance does not require any on-shell or restricted Weyl condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It is a generalization of Weyl (conformal) invariance that is valid in the presence of the Weyl-breaking R2 term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Let us now take a closer look at what is common and what is different between the BRST invariance of pure R2 gravity and the BRST invariance in the gauge theories of particle physics (for concreteness and simplicity, we will consider scalar QED again but the main points apply also to QCD).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST invariance in scalar QED can be viewed as a generalization of gauge invariance in the presence of the gauge-fixing (and hence gauge-breaking) term (∂µ Aµ)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The are two points in common between the scalar QED and R2 cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' First, the Ricci scalar R under a Weyl transformation and the term ∂µ Aµ under a gauge transformation both pick up an extra □Φ(x) term (where Φ(x) represents either a conformal factor Ω(x) in a Weyl transformation or a function f(x) in a gauge transformation).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Recall that in a BRST transformation, Φ(x) is a product of a Grassmann number θ with a Grassmann field (the product yields a commuting (bosonic) quantity).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The second point in common is that R and ∂µ Aµ are both squared.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The squaring yields a (□Φ(x))2 term which is zero since θ2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The squaring still leaves one extra □Φ(x) term and this is cancelled out in both cases via the transformation property of a Grassmann field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' These two common points render the analogy between the two cases quite strong.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, there is one important difference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In scalar QED (and in QCD) , the BRST transformations are associated with gauge transformations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST invariance of pure R2 gravity that we are considering here is not associated with gauge transformations (such as diffeomorphisms) but with Weyl (local scale) transformations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We will see that this difference plays an important role when the theory is quantized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 6 4 Spontaneous breaking of BRST symmetry We now show that the BRST-invariant action S = � d4x√−g (αR2 + ¯c □c) (16) is conformally equivalent to an action that involves the Einstein-Hilbert term;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' this will involve the spontaneous breaking of BRST symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The starting point is to introduce a auxiliary field σ(x) to rewrite the above action into the equivalent form S1 = � d4x√−g (−α(b σ + R)2 + αR2 + ¯c □c) � d4x√−g (−α b2 σ2 − 2 α b R σ + ¯c □c) (17) where b is a real non-zero constant with dimensions of mass squared and σ(x) is dimensionless.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Action (17) is equivalent to the original action (16) since adding the squared term in the first line of (17) does not alter anything (classically, the equations of motion are unaffected and quantum mechanically, the path integral over σ is a Gaussian which yields a constant).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The equivalent action (17) is also BRST invariant;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' it is invariant under the following transforma- tions: gµν → (1 + 2 θ c) gµν ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' ¯c → (1 − 2 θ c) ¯c − 12 θ α b σ ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' σ → (1 − 2θ c) σ (18) where θ is again a Grassmann number.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Note that the BRST invariance requires the auxiliary field σ to transform besides the fields gµν and ¯c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We now perform the following conformal (Weyl) transformation: gµν → σ−1 gµν ¯c → σ ¯c (19) which leads to √−g → σ−2 √−g and R → σ R − 6 σ3/2□(σ−1/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under the above conformal transformation, action (17) becomes S2 = � d4x√−g (−α b2 − 2 α b R + 3α b σ2 ∂µσ ∂µσ + ¯c □c − 1 σ ¯c ∂µc ∂µσ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (20) The above action is no longer invariant under the BRST transformations given by (18).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST symmetry has been spontaneously broken.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The factor σ−1 appearing in the confor- mal transformation (19) is valid only for non-zero σ so that the VEV (vacuum expectation value) of the field σ must be non-zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The VEV is therefore not invariant under the BRST transformation σ → (1 − 2θ c) σ leading to the spontaneous breaking of the BRST symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 7 We can identify −2 α b R as an Einstein-Hilbert term if we equate −2 α b with 1 16π G where G is Newton’s constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The constant term −α b2 can then be associated with a cosmological constant Λ = −b/4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Note that though −2 α b is positive, the constant b can be either positive or negative (but not zero).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This implies that the cosmological constant can be either positive corresponding to a de Sitter (dS) background or negative corresponding to an anti-de Sitter (AdS) background but it cannot be identically zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We can then express (20) as the following Einstein action, SE = � d4x√−g � 1 16π G(R − 2 Λ) + 3α b σ2 ∂µσ ∂µσ + ¯c □c − 1 σ ¯c ∂µc ∂µσ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (21) We have left the constant 3 α b in the action for simplicity but it is not an independent constant;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' it is equal to −3 32 πG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We therefore obtain an Einstein-Hilbert action with non-zero cosmological constant, a kinetic term for the scalar σ (which we will express in canonical form later) and an interaction term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Recall that σ is non-zero so that divisions by σ pose no issue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It is well-known that in spontaneously broken theories, the vacuum breaks the symmetry but it is not actually broken in the Lagrangian but manifested or realized in a different way [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It can be directly verified (see Appendix A) that the Einstein action (21) is invariant under the following transformations: σ → (1 − 2θ c) σ , gµν → gµν and ¯c → ¯c − 12 θ α b .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (22) The BRST symmetry of action (17) manifests itself in the Einstein action (21) via its symmetry under the above transformations (22).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We now show how transformation (22) stems from the BRST transformations (18).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In the Einstein action and transformation (22) label the metric and the barred Grassmann field with a subscript E i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' gµνE and ¯cE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In action (17) and transformation (18) we leave gµν and ¯c as is.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Then the conformal transformation (19) yields gµνE = σ gµν and ¯cE = σ−1 ¯c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under the BRST transformations (18) we obtain gµνE = σ gµν → (1 − 2 θ c) σ (1 + 2 θ c) gµν = σ gµν = gµνE and ¯cE = σ−1 ¯c → (1 + 2 θ c) σ−1� (1 − 2 θ c) ¯c − 12 θ α b σ � = σ−1¯c − 12 θ α b = ¯cE − 12 θ α b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We have therefore obtained the transformations gµνE → gµνE and ¯cE → ¯cE − 12 θ α b which correspond to those in (22).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Note that we used σ → (1 − 2θ c) σ in (18) to derive this, so the transformation of σ is also part of (22).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We can define a real massless scalar field ψ(x) = √ −3α b ln σ(x) so that the kinetic term for σ is expressed in canonical form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The Einstein action (21) expressed in terms of the field ψ is S = � d4x√−g � 1 16π G(R − 2 Λ) − ∂µψ ∂µψ + ¯c □c − 1 √ −3α b ¯c ∂µc ∂µψ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (23) The massless scalar field ψ corresponds to the Nambu-Goldstone boson of the broken sector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under transformation (22), the field ψ transforms as a shift ψ → ψ− √ −3α b 2 θ c (whereas ¯c → ¯c−12 θ α b and gµν → gµν).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The above action (23) is invariant under those transformations (see 8 Appendix A).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is in accord with what we expect from spontaneously broken theories: the original symmetry in the Lagrangian manifests itself in the broken sector as a shift symmetry of the Goldstone bosons [12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='1 Quantum anomaly We saw that the action (17) is BRST invariant under the following transformations: gµν → (1 + 2 θ c) gµν , ¯c → (1 − 2 θ c) ¯c − 12 θ α b σ , σ → (1 − 2θ c) σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Each transformation involves a Weyl transformation where the conformal factor is expressed in terms of of a prod- uct of two Grassmann variables The BRST symmetry is therefore a generalization of Weyl (conformal) symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' After quantization, renormalization introduces a scale which breaks the BRST symmetry since it automatically breaks Weyl symmetry (leading to the well-known Weyl (conformal) anomaly).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' So the BRST symmetry of pure R2 gravity is anomalous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is in contrast to the BRST invariance of gauge theories like QCD which have no anomaly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' After the BRST symmetry is spontaneously broken and we obtain the Einstein action (21), we saw that the BRST symmetry manifests itself now in the Einstein action as a symmetry under the transformations (22).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This symmetry is also anomalous since the transformation of the field σ is a Weyl transformation and renormalization breaks this symmetry (leading again to the Weyl (conformal anomaly).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Another way to see this is to note that the only fields that transform in (22) are ¯c and σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The transformation for ¯c is simply a constant shift so that its path integral measure D¯c is invariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, σ undergoes a Weyl transformation and this introduces a non-trivial Jacobian J (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' J ̸= 1) to the measure Dσ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Since the measure is not invariant, this implies there is an anomaly [17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' So the symmetry in the unbroken phase and its associated symmetry in the broken phase are both anomalous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Our finding here is in accord with previous work that shows that when the Weyl or conformal symmetry is spontaneously broken there is conformal anomaly matching between the unbroken and broken phases [18,19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 5 Conclusion In the last six years or so, we have kept discovering new aspects of pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' A non- exhaustive list includes its unitarity among quadratic gravity theories [4], its conformal equiv- alence to Einstein gravity with non-zero cosmological constant and massless scalar field [1–5], its restricted Weyl symmetry [7,10,11], its spontaneous symmetry breaking to Einstein grav- ity [3,5] and the lack of a propagating graviton when the theory is linearized about a Minkowski background [4] (where there is no Einstein equivalence since the cosmological constant is zero).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In this paper, we have gained further insights into this theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We saw that pure R2 gravity has an analog with the gauge-fixing term (∂µAµ)2 in gauge theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' R2 is not invariant under 9 the Weyl transformation gµν → Ω2(x) gµν just like (∂µ Aµ)2 is not invariant under the gauge transformation Aµ → Aµ + 1 e ∂µf(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, each have a residual symmetry (when □Ω = 0 is satisfied in the gravity case and □f = 0 is satisfied in the gauge case).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This analogy opened the door towards enlarging the symmetry of pure R2 gravity to include BRST symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We first showed that when a massless scalars field was included in the pure R2 action, the condition □Ω = 0 could be met automatically when the equations of motion were satisfied i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' we went from restricted Weyl to on-shell Weyl invariance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Finally, we obtained the BRST invariance of pure R2 gravity where no restricted Weyl or on-shell condition is required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST transformations involve Weyl transformations where the conformal factor is composed of products of Grassmann variables (the conformal factor itself is commutative).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The important point is that the BRST invariance exists despite the Weyl-breaking R2 term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' There is one important difference between the BRST symmetry in gauge theories like QCD and the BRST symmetry that we have considered here for pure R2 gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Gauge invari- ance in particle physics is preserved after quantization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST invariance of QCD is a generalization of gauge invariance so that it is also preserved after quantization;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' there is no anomaly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' In contrast to gauge symmetry, global scale or Weyl (local scale) symmetry is broken after quantization since renormalization introduces a scale.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The BRST symmetry of pure R2 gravity is a generalization of Weyl (conformal) symmetry so that it is also broken after quan- tization leading to the well-known Weyl (conformal) anomaly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' After the spontaneous breaking of the BRST symnmetry, we obtained an Einstein action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We showed that this action has its own symmetry and that it is also anomalous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This is in accord with previous work that shows that when the Weyl (conformal) symmetry is spontaneously broken there is conformal anomaly matching between the unbroken and broken sectors [18,19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The focus of this paper was pure R2 gravity because of its many special and attractive fea- tures that we previously mentioned.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' All other quadratic gravity theories (like Weyl-squared, Riemann-squared, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' ), apart from boundary terms, can be expressed as a linear combina- tion of R2 and RµνRµν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The latter term, the square of the Ricci tensor, appears in quantum corrections to General Relativity (GR) and even though it does not constitute a valid UV completion of GR due to its non-unitarity (yields a massive spin two ghost [4, 20]), it still makes a well-known calculable short-range correction to the Newtonian potential [12,21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Like R2, the term RµνRµν is not Weyl-invariant so it would be of interest to see if it can be BRST invariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It is not in the form of a scalar squared like (∂µAµ)2 or R2, so one may be inclined to think that the BRST formalism would not apply here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, like pure R2, it was shown in [7] that RµνRµν is restricted Weyl invariant (up to a boundary term).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' This suggests that the procedure used to establish the BRST invariance of pure R2 gravity might in the end also work for this quadratic theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' It would therefore be worthwhile and interesting to investigate this further.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 10 Acknowledgments A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' acknowledges support from a discovery grant of the National Science and Engineering Research Council of Canada (NSERC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' A Symmetry of Einsten Action In this appendix we show that the Einstein action (21) given by SE = � d4x√−g � 1 16π G(R − 2 Λ) + 3α b σ2 ∂µσ ∂µσ + ¯c □c − 1 σ ¯c ∂µc ∂µσ) � (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='1) is invariant under the transformations (22) given by σ → (1 − 2θ c) σ , gµν → gµν and ¯c → ¯c − 12 θ α b .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='2) Under the above transformation, the metric gµν does not change so that √−g as well as the term √−g 1 16π G(R − 2 Λ) does not change.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The other terms in the above Einstein action transform as 3α b σ2 ∂µσ ∂µσ → 3α b σ2 ∂µσ ∂µσ − 12 θ α b σ ∂µc ∂µσ − 1 σ ¯c ∂µc ∂µσ) → − 1 σ ¯c ∂µc ∂µσ + 12 θ α b σ ∂µc ∂µσ ¯c □c → ¯c □c − 12 θ α b □c (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='3) where we used that θ2 = 0 (since θ is a Grassmann number) and that gµν ∂µc ∂νc = 0 since gµν is symmetric and c(x) and its derivatives are Grassmann fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We see that the extra term − 12 θ α b σ ∂µc ∂µσ in the first line of (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='3) is canceled exactly by the extra term in the second line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The extra term in the third line, −12 θ α b □c, where −12 θ α b is a constant, does not cancel out with any other extra term in (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, √−g □c is a total derivative that yields an inconsequential boundary term in the action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We have therefore shown that action (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='1) is invariant under transformations (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We saw in section 4 that the Einstein action (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='1) can be expressed in terms of a real massless scalar field ψ(x) = √ −3α b ln σ(x) as action (23): S = � d4x√−g � 1 16π G(R − 2 Λ) − ∂µψ ∂µψ + ¯c □c − 1 √ −3α b ¯c ∂µc ∂µψ � (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='4) where ψ was identified as the Nambu-Goldstone boson of the broken sector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We stated in section 4 that the action (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='4) was invariant under the following transformations: ψ → ψ − √ −3α b 2 θ c , ¯c → ¯c − 12 θ α b and gµν → gµν .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='5) 11 We now verify this statement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Under (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='5) the last three terms in action (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='4) transform as: − ∂µψ ∂µψ → −∂µψ ∂µψ + 4 θ √ −3 α b ∂µψ ∂µc − 1 √ −3α b ¯c ∂µc ∂µψ → − 1 √ −3α b ¯c ∂µc ∂µψ − 4 θ √ −3 α b ∂µψ ∂µc ¯c □c → ¯c □c − 12 θ α b □c .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='6) We see that the extra term +4 θ √ −3 α b ∂µψ ∂µc in the first line above is cancelled by the extra term on the second line which is equal to its negative.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' The only extra term that is not cancelled is the term −12 θ α b □c appearing in the last line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' However, √−g □c is a total derivative which yields a boundary term with no physical consequence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' We have therefore verified that the Einstein action (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='4) is indeed invariant under the transformations (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' References [1] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Kounnas, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' L¨ust and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Toumbas, R2 inflation from scale invariant supergravity and anomaly free superstrings with fluxes, Fortsch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 63, 12 (2015) [arXiv:1409.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='7076].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [2] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Kehagias, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Kounnas, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' L¨ust and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Riotto, Black Hole Solutions in R2 Gravity, JHEP 05, 143 (2015)[arXiv:1502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='04192].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [3] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Edery and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Nakayama, Generating Einstein gravity, cosmological constant and Higgs mass from restricted Weyl invariance, Mod.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' A 30, 1550152 (2015) [arXiv:1502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='05932].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [4] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Alvarez-Gaume, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Kehagias, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Kounnas, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' L¨ust and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Riotto, Aspects of Quadratic Gravity, Fortsch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 64, 176 (2016) [arXiv:1505.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='07657].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [5] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Edery and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Nakayama, Gravitating magnetic monopole via the spontaneous symmetry breaking of pure R2 gravity, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 98 (2018) 064011 [arXiv:1807.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='07004].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [6] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Edery and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Nakayama, Palatini formulation of pure R2 gravity yields Einstein gravity with no massless scalar, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 99, 124018 (2019) [arXiv:1902.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='07876].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [7] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Edery and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Nakayama, Restricted Weyl invariance in four-dimensional curved space- time, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 90, 043007 (2014) [arXiv:1406.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='0060].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [8] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Edery, Pure R2 gravity can gravitate about a flat background, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Conf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Ser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 1956, 012005 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [9] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Edery and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Nakayama, Critical gravity from four dimensional scale invariant gravity, JHEP 11, 169 (2019) [arXiv:1908.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='08778].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 12 [10] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Oda, Restricted Weyl symmetry, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 102, 045008 (2020)[arXiv:2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='04771].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [11] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Oda, Restricted Weyl Symmetry and Spontaneous Symmetry Breakdown of Conformal Symmetry, Mod.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' A 36, 2150203 (2021) [arXiv:2104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='04694].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [12] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Schwartz, Quantum Field Theory and the Standard Model, (Cambridge University Press, Cambridge, UK, 2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [13] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Oda and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Saake, BRST formalism of Weyl conformal gravity, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 106, 106007 (2022) [arXiv:2209.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='14533].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [14] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Prinz, The BRST Double Complex for the Coupling of Gravity to Gauge Theories, [arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='00780].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [15] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Kugo, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Nakayama and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='Ohta, Covariant BRST quantization of unimodular grav- ity: Formulation with antisymmetric tensor ghosts, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 105, 086006 (2022), [arXiv:2202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='03626].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [16] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Berezhiani, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Dvali and O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Sakhelashvili, de Sitter space as a BRST invariant coherent state of gravitons Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 105, 025022 (2022) [arXiv:2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='12022].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [17] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Fujikawa and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Suzuki, Path Integrals and Quantum Anomalies, (Oxford University Press, Oxford, UK, 2004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [18] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Schwimmer and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Theisen, Spontaneous Breaking of Conformal Invariance and Trace Anomaly Matching,Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' B 847, 590 (2011) [arXiv:1011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='0696.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [19] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Cabo-Bizet, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Gava and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Narain, Holography and Conformal Anomaly Matching, JHEP 11,044 (2013) [arXiv:1307.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='3784].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [20] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Stelle, Renormalization of Higher Derivative Quantum Gravity, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' D 16, 953 (1977).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' [21] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content='F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Donoghue, Leading quantum correction to the Newtonian potential, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 72, 2996 (1994).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
+page_content=' 13' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/8tFAT4oBgHgl3EQfpB3g/content/2301.08638v1.pdf'}
diff --git a/ANAzT4oBgHgl3EQfTPxG/content/2301.01245v1.pdf b/ANAzT4oBgHgl3EQfTPxG/content/2301.01245v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..a131ce3498dc6f54601cca3323ec0ebd2eb65057
--- /dev/null
+++ b/ANAzT4oBgHgl3EQfTPxG/content/2301.01245v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e4ce288148d18f1bd09ca6825176f71c9a8680fbcc1fb616f5569120d1e6b6
+size 3351789
diff --git a/ANAzT4oBgHgl3EQfTPxG/vector_store/index.faiss b/ANAzT4oBgHgl3EQfTPxG/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..6389365900964a415fd3309d4b0ccdb596cb0426
--- /dev/null
+++ b/ANAzT4oBgHgl3EQfTPxG/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6570720f44da280a375a564069c09602121a87ced5ae3be677e4d24faa28867f
+size 2162733
diff --git a/ANE3T4oBgHgl3EQfTQrd/content/tmp_files/2301.04441v1.pdf.txt b/ANE3T4oBgHgl3EQfTQrd/content/tmp_files/2301.04441v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d2cbb8ea05cc32a1fc9f90ad348c457412c73b8b
--- /dev/null
+++ b/ANE3T4oBgHgl3EQfTQrd/content/tmp_files/2301.04441v1.pdf.txt
@@ -0,0 +1,919 @@
+Wasserstein Gradient Flows of the Discrepancy
+with Distance Kernel on the Line⋆
+Johannes Hertrich, Robert Beinert, Manuel Gräf, and Gabriele Steidl
+TU Berlin, Institute of Mathematics, Straße des 17. Juni 136, 10623 Berlin, Germany
+{hertrich,beinert, graef,steidl}@math.tu-berlin.de
+https://tu.berlin/imageanalysis/
+Abstract. This paper provides results on Wasserstein gradient flows between measures on
+the real line. Utilizing the isometric embedding of the Wasserstein space P2(R) into the Hilbert
+space L2((0, 1)), Wasserstein gradient flows of functionals on P2(R) can be characterized as
+subgradient flows of associated functionals on L2((0, 1)). For the maximum mean discrepancy
+functional Fν := D2
+K(·, ν) with the non-smooth negative distance kernel K(x, y) = −|x − y|,
+we deduce a formula for the associated functional. This functional appears to be convex,
+and we show that Fν is convex along (generalized) geodesics. For the Dirac measure ν = δq,
+q ∈ R as end point of the flow, this enables us to determine the Wasserstein gradient flows
+analytically. Various examples of Wasserstein gradient flows are given for illustration.
+Keywords: Maximum Mean Discrepancy · Wasserstein gradient flows · Riesz kernel.
+1
+Introduction
+Gradient flows provide a powerful tool for computing the minimizers of modeling functionals in
+certain applications. In particular, gradient flows on the Wasserstein space are an interesting field
+of research that combines optimization with (stochastic) dynamical systems and differential geom-
+etry. For a good overview on the theory, we refer to the books of Ambrosio, Gigli and Savaré [3],
+and Santambrogio [31]. Besides Wasserstein gradient flows of the Kullback–Leibler (KL) functional
+KL(·, ν) and the associated Fokker–Planck equation related to the overdamped Langevin dynamics,
+which were extensively examined in the literature, see, e.g., [19,26,28], flows of maximum mean
+discrepancy (MMD) functionals Fν := D2
+K(·, ν) became popular in machine learning [4] and image
+processing [14]. On the other hand, MMDs were used as loss functions in generative adversarial
+networks [6,13,22]. Wasserstein gradient flows of MMDs are not restricted to absolutely continuous
+measures and have a rich structure depending on the kernel. So the authors of [4] showed that for
+smooth kernels K, particle flows are indeed Wasserstein gradient flows meaning that Wasserstein
+flows starting at an empirical measure remain empirical measures and coincide with usual gradi-
+ent descent flows in Rd. The situation changes for non-smooth kernels like the negative distance,
+where empirical measures can become absolutely continuous ones and conversely, i.e. particles may
+explode. The concrete behavior of the flow depends also on the dimension, see [11,12,17,18]. The
+crucial part is the treatment of the so-called interaction energy within the discrepancy, which is
+repulsive and responsible for the proper spread of the measure. This nicely links to another field of
+mathematics, namely potential theory [21,30].
+⋆ Supported by the German Research Foundation (DFG) [grant numbers STE571/14-1, STE 571/16-1]
+and the Federal Ministry of Education and Research (BMBF, Germany) [grant number 13N15754].
+arXiv:2301.04441v1 [math.OC] 11 Jan 2023
+
+2
+J. Hertrich et al.
+In this paper, we are just concerned with Wasserstein gradient flows on the real line. Optimal
+transport techniques that reduce the original transport to those on the line were successfully used
+in several applications [1,5,9,10,20,27]. When working on R, we can exploit quantile functions of
+measures to embed the Wasserstein space P2(R) into the Hilbert space of (equivalence classes) of
+square integrable functions L2((0, 1)). Then, instead of dealing with functionals on P2(R), we can
+just work with associated functionals which are uniquely defined on a cone of L2((0, 1)). If the asso-
+ciated functional is convex, we will see that the original one is convex along (generalized) geodesics,
+which is a crucial property for the uniqueness of the Wasserstein gradient flow. Furthermore, we
+can characterize Wasserstein gradient flows using regular subdifferentials in L2((0, 1)). Note that
+the special case of Wasserstein gradient flows of the interaction energy was already considered in
+[7]. We will have a special look at the Wasserstein gradient flow of Fδq := D2
+K(·, δq) for the negative
+distance kernel, i.e. flows ending in δq. We will deduce an analytic formula for this flow and provide
+several examples to illustrate its behavior.
+Outline of the paper. In Section 2, we recall the basic notation on Wasserstein gradient flows in d
+dimensions. Then, in Section 3, we show how these flows can be simpler treated as gradient descent
+flows of an associated function on the Hilbert space L2((0, 1)). MMDs are introduced in Section 4.
+Then, in Section 5, we restrict our attention again to the real line and show how the associated
+functional looks for the MMD with negative distance kernel. In particular, this functional is convex.
+For the Dirac measure ν = δq, q ∈ R, we give an explicit formula for the Wasserstein gradient flow
+of the MMD functional. Examples illustrating the behavior of the Wasserstein flows are provided
+in Section 6. Finally, conclusions are drawn in Section 7.
+2
+Wasserstein Gradient Flows
+Let M(Rd) denote the space of σ-additive, signed measures and P(Rd) the set of probability
+measures. For µ ∈ M(Rd) and measurable T : Rd → Rn, the push-forward of µ via T is given by
+T#µ := µ ◦ T −1. We consider the Wasserstein space P2(Rd) := {µ ∈ P(Rd): �
+Rd ∥x∥2
+2 dµ(x) < ∞}
+equipped with the Wasserstein distance W2 : P2(Rd) × P2(Rd) → [0, ∞),
+W 2
+2 (µ, ν) :=
+min
+π∈Γ (µ,ν)
+�
+Rd×Rd ∥x − y∥2
+2 dπ(x, y),
+µ, ν ∈ P2(Rd),
+(1)
+where Γ(µ, ν) := {π ∈ P2(Rd × Rd) : (π1)#π = µ, (π2)#π = ν} and πi(x) := xi, i = 1, 2 for
+x = (x1, x2). The set of optimal transport plans π realizing the minimum in (1) is denoted by
+Γ opt(µ, ν). A curve γ : I → P2(Rd) on an interval I ⊂ R, is called a geodesic if there exists a
+constant C ≥ 0 such that
+W2(γ(t1), γ(t2)) = C|t2 − t1|,
+for all t1, t2 ∈ I.
+The Wasserstein space is a geodesic space, meaning that any two measures µ, ν ∈ P2(Rd) can be
+connected by a geodesic. The regular tangent space at µ ∈ P2(Rd) is given by
+TµP2(Rd) :=
+�
+λ(T − Id) : (Id, T)#µ ∈ Γ opt(µ, T#µ), λ > 0
+�L2,µ.
+Here L2,µ denotes the Bochner space of (equivalence classes of) functions ξ : Rd → Rd with
+finite ∥ξ∥2
+L2,µ :=
+�
+Rd ∥ξ(x)∥2
+2 dµ(x) < ∞. Note that TµP2(Rd) is not a “classical” tangent space, in
+
+Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line
+3
+particular it is an infinite dimensional subspace of L2,µ if µ is absolutely continuous and just Rd
+if µ = δx, x ∈ Rd. In particular, this means that the Wasserstein space has only a “manifold-like”
+structure.
+For λ ∈ R, a function F : P2(Rd) → (−∞, +∞] is called λ-convex along geodesics if, for every
+µ, ν ∈ dom F := {µ ∈ P2(Rd) : F(µ) < ∞}, there exists at least one geodesic γ : [0, 1] → P2(Rd)
+between µ and ν such that
+F(γ(t)) ≤ (1 − t) F(µ) + t F(ν) − λ
+2 t(1 − t) W 2
+2 (µ, ν),
+t ∈ [0, 1].
+In the case λ = 0, we just speak about convex functions. For a proper and lower semi-continuous
+(lsc) function F : P2(Rd) → (−∞, ∞] and µ ∈ P2(Rd), the reduced Fréchet subdifferential at µ is
+defined as
+∂F(µ) :=
+�
+ξ ∈ L2,µ : F(ν) − F(µ) ≥
+inf
+π∈Γ opt(µ,ν)
+�
+R2d
+⟨ξ(x), y − x⟩ dπ(x, y) + o(W2(µ, ν)) ∀ν ∈ P2(Rd)
+�
+. (2)
+A curve γ : I → P2(Rd) is absolutely continuous, if there exists a Borel velocity field vt : Rd → Rd
+with
+�
+I ∥vt∥L2,γ(t) dt < +∞ such that
+∂tγ(t) + ∇x · (vt γ(t)) = 0
+(3)
+on I × Rd in the distributive sense, i.e., for all ϕ ∈ C∞
+c (I × Rd) it holds
+�
+I
+�
+Rd ∂tϕ(t, x) + vt(x) · ∇x ϕ(t, x) dγ(t) dt = 0.
+A locally absolutely continuous curve γ : (0, +∞) → P2(Rd) with velocity field vt ∈ Tγ(t)P2(Rd) is
+called a Wasserstein gradient flow with respect to F : P2(Rd) → (−∞, +∞] if
+vt ∈ −∂F(γ(t)),
+for a.e. t > 0.
+(4)
+3
+Wasserstein Gradient Flows on the Line
+Now we restrict our attention to d = 1, i.e., we work on the real line. We will see that the above
+notation simplifies since there is an isometric embedding of P2(R) into L2((0, 1)). To this end, we
+consider the cumulative distribution function Rµ : R → [0, 1] of µ ∈ P2(R), which is defined by
+Rµ(x) := µ((−∞, x]), x ∈ R. It is non-decreasing and right-continuous with limx→−∞ Rµ(x) = 0 as
+well as limx→∞ Rµ(x) = 1. The quantile function Qµ : (0, 1) → R is the generalized inverse of Rµ
+given by
+Qµ(p) := min{x ∈ R: Rµ(x) ≥ p},
+p ∈ (0, 1).
+It is non-decreasing and left-continuous. The quantile functions form a convex cone C((0, 1)) :=
+{Q ∈ L2((0, 1)) : Q nondecreasing} in L2((0, 1)). Note that both the distribution and quantile
+functions are continuous except for at most countably many jumps. For a good overview see [29,
+§ 1.1]. By the following theorem, the mapping µ �→ Qµ is an isometric embedding of P2(R) into
+L2((0, 1)).
+Theorem 1 ([32, Thm 2.18]). For µ, ν ∈ P2(R), the quantile function Qµ ∈ C((0, 1)) satisfies
+µ = (Qµ)#λ(0,1) and
+W 2
+2 (µ, ν) =
+� 1
+0
+|Qµ(s) − Qν(s)|2ds.
+
+4
+J. Hertrich et al.
+Next we will see that instead of working with functionals F : P2(R) → (−∞, +∞], we can just
+deal with associated functionals F: L2((0, 1)) → (−∞, ∞] fulfilling F(Qµ) := F(µ). Note that F
+is defined in this way only on C((0, 1)), and there exist several continuous extensions to the whole
+linear space L2((0, 1)). Instead of the extended Fréchet subdifferential (2), we will use the regular
+subdifferential in L2((0, 1)) defined by
+∂G(f) :=
+�
+h ∈ L2((0, 1)) : G(g) ≥ G(f) + ⟨h, g − f⟩ + o(∥g − f∥L2) ∀g ∈ L2((0, 1))
+�
+.
+The following theorem characterizes Wasserstein gradient flows by this regular subdifferential and
+states a convexity relation between F : P2(R) → (−∞, +∞] and the associated functional F.
+Theorem 2. i) Let γ : (0, ∞) → P2(R) be a locally absolutely continuous curve and F: L2((0, 1)) →
+(−∞, ∞] such that the pointwise derivative ∂tQγ(t) exists and fulfills the L2 subgradient equation
+∂tQγ(t) ∈ −∂F(Qγ(t)),
+for almost every t ∈ (0, +∞).
+Then γ is a Wasserstein gradient flow with respect to the functional F : P2(R) → (−∞, +∞] defined
+by F(µ) := F(Qµ).
+ii) If F : C((0, 1)) → (−∞, ∞] is convex, then F(µ) := F(Qµ) is convex along geodesics.
+Proof. i) Since γ is (locally) absolute continuous, the velocity field vt from (3) fulfills by [3,
+Prop 8.4.6] for almost every t ∈ (0, ∞) the relation
+0 = lim
+h→0
+W2(γ(t + h), (Id + hvt)#γ(t))
+|h|
+= lim
+h→0
+W2((Qγ(t+h))#λ(0,1),
+�
+Qγ(t) + h(vt ◦ Qγ(t))
+�
+#λ(0,1))
+|h|
+= lim
+h→0
+���Qγ(t+h) − Qγ(t)
+h
+− vt ◦ Qγ(t)
+���
+L2 = ∥∂tQγ(t) − vt ◦ Qγ(t)∥L2.
+Thus, by assumption, vt ◦ Qγ(t) ∈ −∂F(Qγ(t)) a.e. In particular, for any µ ∈ P2(R), we obtain
+0 ≤ F(Qµ) − F(Qγ(t)) +
+� 1
+0
+vt(Qγ(t)(s)) (Qµ(s) − Qγ(t)(s)) ds + o(∥Qµ − Qγ(t)∥L2)
+= F(µ) − F(γ(t)) +
+�
+R×R
+vt(x) (y − x) dπ(x, y) + o
+�
+W2(µ, γ(t))
+�
+,
+where π := (Qγ(t), Qµ)#λ(0,1). Since π the unique optimal transport plan between γ(t) and µ, this
+yields by (2) that vt ∈ −∂F(γ(t)) showing the assertion by (4).
+ii) Let F: L2((0, 1)) → R be convex. For any geodesic γ : [0, 1] → P2(R), since µ �→ Qµ is an
+isometry, the curve t �→ Qγ(t) is a geodesic in L2((0, 1)) too. Since L2((0, 1)) is a linear space, the
+convexity of F: L2((0, 1)) → R yields that t �→ F(Qγ(t)) = F(γ(t)) is convex. Thus, F is convex
+along γ.
+⊓⊔
+Remark 1. If F : P2(R) → (−∞, +∞] is proper, lsc, coercive and λ-convex along so-called general-
+ized geodesics, then the Wasserstein gradient flow starting at any µ0 ∈ dom F is uniquely determined
+and is the uniform limit of the miminizing movement scheme of Jordan, Kinderlehrer and Otto [19]
+when the time step size goes to zero, see [3, Thm 11.2.1]. In R, but not in higher dimensions,
+λ-convex functions along geodesics fulfill also the stronger property that they are λ-convex along
+generalized geodesics, see [18].
+
+Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line
+5
+4
+Discrepancies
+We consider symmetric and conditionally positive definite kernels K : Rd × Rd → R of order one,
+i.e., for any n ∈ N, any pairwise different points x1, . . . , xn ∈ Rd and any a1, . . . , an ∈ R with
+�n
+i=1 ai = 0 the relation �n
+i,j=1 aiajK(xi, xj) ≥ 0 is satisfied. Typical examples are Riesz kernels
+K(x, y) := −∥x − y∥r,
+r ∈ (0, 2),
+where we have strict inequality except for all aj, j = 1, . . . , n being zero. The maximum mean
+discrepancy (MMD) D2
+K : P(Rd) × P(Rd) → R between two measures µ, ν ∈ P(Rd) is defined by
+D2
+K(µ, ν) := EK(µ − ν)
+with the so-called K-energy on signed measures
+EK(σ) := 1
+2
+�
+Rd
+�
+Rd K(x, y) dσ(x)dσ(y),
+σ ∈ M(Rd).
+The relation between discrepancies and Wasserstein distances is discussed in [15,24]. For fixed
+ν ∈ P(Rd), the MMD can be decomposed as
+Fν(µ) = D2
+K(µ, ν) = EK(µ) + VK,ν(µ) + EK(ν)
+� �� �
+const.
+with the interaction energy on probability measures
+EK(µ) = 1
+2
+�
+Rd
+�
+Rd K(x, y) dµ(x)dµ(y),
+µ ∈ P2(Rd)
+and the potential energy of µ with respect to the potential of ν,
+VK,ν(µ) :=
+�
+Rd VK,ν(y)dµ(x),
+VK,ν(x) := −
+�
+Rd K(x, y)dν(y).
+In dimensions d ≥ 2 neither EK nor D2
+K with the Riesz kernel are λ-convex along geodesics, see
+[18], so that certain properties of Wasserstein gradient flows do not apply. We will see that this is
+different on the real line.
+5
+MMD Flows on the Line
+In the rest of this paper, we restrict our attention to d = 1 and negative distance K(x, y) =
+−|x − y|, i.e. to Riesz kernels with r = 1. For fixed ν ∈ P2(R), we consider the MMD functional
+Fν := D2
+K(·, ν). Note that the unique minimizer of this functional is given by µ = ν.
+Lemma 1. Let Fν := D2
+K(·, ν) with the negative distance kernel. Then the convex functional
+Fν : L2((0, 1)) → R defined by
+Fν(f) :=
+� 1
+0
+�
+(1 − 2s)(f(s) + Qν(s)) +
+� 1
+0
+|f(s) − Qν(t)| dt
+�
+ds.
+(5)
+fulfills Fν(Qµ) = Fν(µ) for all µ ∈ P2(R). In particular, Fν is convex along (generalized) geodesics
+and there exists a unique Wasserstein gradient flow.
+
+6
+J. Hertrich et al.
+Proof. We reformulate Fν as
+Fν(µ) = −1
+2
+�
+R×R
+|x − y|(dµ(x) − dν(x))(dµ(y) − dν(y))
+= −1
+2
+� 1
+0
+� 1
+0
+|Qµ(s) − Qµ(t)| − 2|Qµ(s) − Qν(t)| + |Qν(s) − Qν(t)| ds dt
+=
+� 1
+0
+� 1
+t
+Qµ(t) − Qµ(s) + Qν(t) − Qν(s) ds dt +
+� 1
+0
+� 1
+0
+|Qµ(s) − Qν(t)| ds dt
+=
+� 1
+0
+� 1
+t
+Qµ(t) + Qν(t) ds dt −
+� 1
+0
+� s
+0
+Qµ(s) + Qν(s) dt ds +
+� 1
+0
+� 1
+0
+|Qµ(s) − Qν(t)| ds dt
+=
+� 1
+0
+�
+(1 − 2s)(Qµ(s) + Qν(s)) +
+� 1
+0
+|Qµ(s) − Qν(t)| dt
+�
+ds,
+which yields the first claim. The second one follows by Theorem 2ii) and Remark 1.
+⊓⊔
+Note that the lemma cannot immediately be generalized to Riesz kernels with r = (1, 2).
+Finally, we derive for the special choice ν = δq in D2
+K(·, ν) an analytic formula for its Wasserstein
+gradient flow.
+Proposition 1. Let Fδq := D2
+K(·, δq) with the negative distance kernel. Then the unique Wasser-
+stein gradient flow of Fδq starting at µ0 = γ(0) ∈ P2(R) is γ(t) = (gt)#λ(0,1), where the function
+gt : (0, 1) → R is given by
+gt(s) :=
+�
+�
+�
+�
+�
+�
+�
+min{Qµ0(s) + 2st, q},
+Qµ0(s) < q,
+q,
+Qµ0(s) = q,
+max{Qµ0(s) + 2st − 2t, q},
+Qµ0(s) > q.
+(6)
+Proof. First, note that gt ∈ C((0, 1)) such that it holds gt = Qγ(t). Since Qδq ≡ q, the subdifferential
+of Fδq in (5) at gt consists of all functions
+h(s) =
+�
+�
+�
+�
+�
+�
+�
+−2s,
+Qµ0(s) < q and t < q−Qµ0(s)
+2s
+,
+2 − 2s,
+Qµ0(s) > q and t < Qµ0(s)−q
+2−2s
+,
+1 − 2s + n(s),
+otherwise,
+with −1 ≤ n(s) ≤ 1 for s ∈ (0, 1). On the other hand, the pointwise derivative of gt in (6) can be
+written as
+∂tgt(s) =
+�
+�
+�
+�
+�
+�
+�
+2s,
+Qµ0(s) < q and t < q−Qµ0(s)
+2s
+,
+2s − 2,
+Qµ0(s) > q and t < Qµ0(s)−q
+2−2s
+,
+0,
+otherwise,
+such that we obtain ∂tQγ(t) = ∂tgt ∈ −∂Fν(gt) = −∂Fν(Qγ(t)). Thus, by Lemma 1 and Theorem 2,
+we obtain that γ is a Wasserstein gradient flow. It is unique since Fν is convex along geodesics by
+Theorem 2.ii, Lemma 1 and Remark 1.
+⊓⊔
+
+Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line
+7
+6
+Intuitive Examples
+Finally, we provide some intuitive examples of Wasserstein gradient flows of Fν := D2
+K(·, ν) with
+the negative distance kernel.
+6.1
+Flow between Dirac Measures
+We consider the flow of Fδ0 starting at the initial measure γ(0) = µ0 := δ−1. Due to Qδ0 ≡ 0,
+Proposition 1 yields the gradient flow γ(t) := (Qt)#λ(0,1) given by
+γ(t) =
+�
+�
+�
+�
+�
+�
+�
+δ−1,
+t = 0,
+1
+2tλ[−1,−1+2t],
+0 ≤ t ≤ 1
+2,
+1
+2tλ[−1,0] +
+�
+1 − 1
+2t
+�
+δ0,
+1
+2 < t.
+For t ∈ (0, 1
+2], the initial Dirac measure becomes a uniform measure with increasing support, and
+for t ∈ ( 1
+2, 1) it is the convex combination of a uniform measure and δ0. A visualization of the flow
+is given in Figure 1.
+□
+Fig. 1: Visualization of the Wasserstein gradient flow of Fδ0 from δ−1 to δ0. At various times t, the
+absolute continuous part is visualized by its density in blue (area equals mass) and the atomic part
+by the red dotted vertical line (height equals mass). The atomic part at the end point x = 0 starts
+to grow at time t = 1
+2, where the support of the density touches this point for the first time.
+6.2
+Flow on Restricted Sets
+Next, we are interested in the Wasserstein gradient flows on the subsets Si, i = 1, 2, given by
+(i) S1 := {δx : x ∈ R},
+(ii) S2 := {µm,σ =
+1
+2
+√
+3σλ[m−
+√
+3σ,m+
+√
+3σ] : m ∈ R, σ ∈ R≥0}.
+Note that S2 is a special instance of sets of scaled and translated measures µ ∈ P2(R) defined by
+{Ta,b#µ : a ∈ R≥0, b ∈ R}, where Ta,b(x) := ax + b. As mentioned in [16] the Wasserstein distance
+between measures µ1, µ2 from such sets has been already known to Fréchet:
+W 2
+2 (µ1, µ2) = |m1 − m2|2 + |σ1 − σ2|2,
+
+t= 0.00
+t= 0.25
+t= 0.29
+t= 0.33
+t=0.40
+t= 0.50
+t= 0.67
+t= 1.00
+t= 2.00
+t= 8
+2.0
+1.5
+1.0 -
+0.5
+0.0
+1
+0
+0
+0
+0
+0
+0
+08
+J. Hertrich et al.
+where mi and σi are the mean value and standard deviation of µi, i = 1, 2. This provides an
+isometric embedding of R×R≥0 into P2(R). The boundary of S2 is the set of Dirac measures S1 and
+is isometric to R. The sets are convex in the sense that for µ, ν ∈ Si all geodesics γ : [0, 1] → P(R)
+with γ(0) = µ and γ(1) = ν are in Si, i ∈ {1, 2}. For i = 1, 2, we consider
+Fi,ν(µ) :=
+�
+Fν
+µ ∈ Si,
++∞
+otherwise.
+Due to the convexity of Fν along geodesics and the convexity of the sets Si, we obtain that the
+functions Fi,ν are convex along geodesics.
+Flows of F1,ν We use the notation fx ≡ x for the constant function on (0, 1) with value x. It is
+straightforward to check that the function F: L2((0, 1)) → (−∞, ∞] given by
+F(f) =
+�
+F(x),
+if f = fx for some x ∈ R,
++∞,
+otherwise,
+with
+F(x) :=
+�
+R
+|x − y| dν(y) − 1
+2
+�
+R×R
+|y − z| dν(y)dν(z)
+fulfills F(Qµ) = F1,ν(µ). In the following, we aim to find x: [0, ∞) → R satisfying
+˙x(t) = −∂F(x(t)).
+Since the set {Qµ : µ ∈ S1} is a one-dimensional linear subspace of L2((0, 1)) spanned by the
+constant one-function f1, this yields fx(t) ∈ −∂F(fx(t)) such that the Wasserstein gradient flow is
+by Theorem 2 given by γ(t) = (fx(t))#λ(0,1) = δx(t).
+In the special case ν = δq for some q ∈ R, we have
+F(x) = |x − q|,
+∂F(x) =
+�
+�
+�
+�
+�
+�
+�
+{−1},
+x < q,
+[−1, 1],
+x = q,
+{1},
+x > q.
+Therefore, the Wasserstein gradient flow for x(0) = x0 ̸= 0 is given by
+γ(t) = δx(t),
+with
+x(t) =
+�
+x0 + t,
+x0 < q,
+x0 − t,
+x0 > q. ,
+0 ≤ t < |x0 − q|
+and γ(t) = δq for t ≥ |x0 − q|.
+For ν = 1
+2λ[−1,1] the gradient flow starting at x0 ∈ [−1, 1] is
+x(t) = x0e−t,
+t ≥ 0,
+and converges to the midpoint of the interval for t → ∞. If it starts at x0 ∈ R \ [−1, 1] the gradient
+flow is
+x(t) =
+�
+x0 + t,
+x0 < −1,
+x0 − t,
+x0 > 1.
+,
+0 ≤ t ≤ min |x0 − 1|, |x0 + 1|,
+where it reaches the nearest interval end point in finite time. In Figure 2, we plotted the x(t) for
+different initial values x(0).
+
+Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line
+9
+Fig. 2: Wasserstein gradient flow of F1,ν for ν = δ0 (left) and ν = 1
+2λ[−1,1] (right) from various
+initial points δx, x ∈ [−2, 2]. The support of the right measure ν is depicted by the blue region. The
+examples show that gradient flows may reach the optimal points in finite or infinite time.
+Flows of F2,ν We observe that Qµm,σ = fm,σ, where fm,σ(x) = m + 2
+√
+3σ(x − 1
+2). By Lemma 1
+we obtain that the function F: L2((0, 1)) → (−∞, ∞] given by
+F(f) =
+�
+F(m, σ),
+if f = fm,σ for (m, σ) ∈ R × R≥0,
++∞,
+otherwise,
+fulfills F(Qµ) = F2,ν(µ), where
+F(m, σ) :=
+�
+(0,1)
+(1 − 2s)(fm,σ(s) + Qν(s))ds +
+�
+(0,1)2 |fm,σ(s) − Qν(t)|dtds,
+The set {fm,σ : m, σ ∈ R} is a two dimensional linear subspace of L2((0, 1)) with orthonormal basis
+{f1,0, f0,1}. We aim to compute m: [0, ∞) → R and σ: [0, ∞) → R≥0 with
+( ˙m(t), ˙σ(t)) = −∂F(m(t), σ(t)),
+t ∈ I ⊂ R,
+(7)
+because this yields fm(t),σ(t) ∈ −∂F(fm(t),σ(t)) such that γ(t) = (fm(t),σ(t))#λ(0,1) = µm,σ is by
+Theorem 2 the Wasserstein gradient flow.
+In the following, we consider the special case ν = δ0 = µ0,0. Then, the function F reduces to
+F(m, σ) =
+�
+R
+(1 − 2s)(m + 2
+√
+3σ(s − 1
+2)) + |m + 2
+√
+3σ(s − 1
+2)|ds
+= − σ
+√
+3 +
+�
+�
+�
+|m|,
+if |m| ≥
+√
+3σ,
+m2+3σ2
+2
+√
+3σ2
+if |m| <
+√
+3σ,
+and the subdifferential is given by
+∂F(m, σ) =
+�
+�
+�
+sgn(m) × {− 1
+√
+3},
+if |m| ≥
+√
+3σ,
+{(
+m
+√
+3σ2 , −m2
+√
+3σ3 −
+1
+√
+3)},
+if |m| <
+√
+3σ,
+sgn(m) =
+�
+{ |m|
+m },
+if m ̸= 0,
+[−1, 1],
+if m = 0.
+We observe that F is differentiable for σ > 0. Thus, for any initial intial value (m(0), σ(0)) =
+(m0, σ0), we can compute the trajectory (m(t), σ(t)) solving (7) using an ODE solver. In Figure 3
+
+2.0
+1.5
+1.0
+0.5
+X
+0.0
+-0.5
+1.0
+1.5
+-2.0
+0.0
+0.5
+1.0
+1.5
+2.0
+2.5
+3.0
+3.5
+4.0
+t2.0
+1.5
+1.0
+0.5
+X
+0.0
+-0.5
+1.0
+1.5
+2.0
+0.0
+0.5
+1.0
+1.5
+2.0
+2.5
+3.0
+3.5
+4.0
+t10
+J. Hertrich et al.
+(left), we plotted the level sets of the function F(m, σ) as well as the solution trajectory (m(t), σ(t))
+for different initial values (m(0), σ(0)). For (m(0), σ(0)) = (−1, 0), the resulting flow is illustrated
+in Figure 3, right.
+Fig. 3: Wasserstein gradient flow F2,δ0 from (m(0), σ(0)) to δ0 (left) and from δ−1 to δ0 (right). In
+contrast Figure 1 it is a uniform measure for all t ∈ (0, 1).
+Flows for a Smooth Kernel For smooth, positive definite kernels K the MMD functional Fν :=
+D2
+K(·, ν) is in general not convex and leads to a more complex energy landscape than for the
+negative distance kernel. This may lead to problems for optimization algorithms. To illustrate this
+observation, we let ν := λ[−1,1] and compare the energy landscape of the restricted functional F2,ν
+for K(x, y) := −|x − y| and the kernel
+˜K(x, y) :=
+�
+(1 − 1
+2|x − y|)2(|x − y| + 1),
+|x − y| ≤ 2,
+0,
+else.
+(8)
+In contrast to the negative distance kernel K, the kernel ˜K is positive definite (without restrictions
+on the ai), cf. [33], and has a Lipschitz continuous gradient. The two energy landscapes of F2,ν are
+visualized in Figure 4. The non-convexity of Fν for ˜K is readily seen by the presence of a saddle
+point for F2,ν at µ = δ0 (equivalently to (m, σ) = (0, 0) in the mσ-plane). Note that any Wasserstein
+gradient flow of Fν starting at a Dirac measure δx converges to this saddle point µ = δ0.
+7
+Conclusions
+We provided insight into Wasserstein gradient flows of MMD functionals with negative distance
+kernels and characterized in particular flows ending in a Dirac measure. We have seen that such flows
+are not simple particle flows, e.g. starting in another Dirac measure the flow becomes immediately
+uniformly distributed and after a certain time a mixture of a uniform and a Dirac measure. In
+our future work, we want to extend our considerations to empirical measures and incorporate
+
+2.00
+1.75
+1.50
+1.25
+b 1.00
+0.75
+0.50
+0.25
+0.00
+-1.00
+-0.75
+-0.50
+-0.25
+0.00
+0.25
+0.50
+0.75
+1.00
+mt= 0.00
+t= 0.25
+t= 0.50
+t= 0.75
+t= 1.00
+t= 1.25
+t= 1.50
+t= 1.75
+t= 2.00
+t = 2.25
+2.0
+1.5
+1.0
+0.5
+0.0
+1
+0
+0
+0
+0
+0
+1
+0Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line
+11
+Fig. 4: Visualization of the energy landscapes of F2,λ[−1,1] for the convex negative distance kernel
+(left) and the non-convex, smooth kernel given in (8). The red dot is the global minimizer λ[−1,1]
+(left and right) and the blue point (right) is the saddle point δ0. The black lines depict selected
+gradient flows.
+deep learning techniques as in [2]. Also the treatment of other functionals which incorporate an
+interaction energy part appears to be interesting. Further, we may combine univariate techniques
+with multivariate settings using Radon transform like techniques as in [8,23,25].
+References
+1. Abraham, I., Abraham, R., Bergounioux, M., Carlier, G.: Tomographic reconstruction from a few views:
+A multi-marginal optimal transport approach. Applied Mathematics and Optimization 75(1), 55–73
+(2017)
+2. Altekrüger, F., Hertrich, J., Steidl, G.: Neural Wasserstein gradient flows for maximum mean discrep-
+ancies with Riesz kernels. arXiv:XXX (2023)
+3. Ambrosio, L., Gigli, N., Savare, G.: Gradient Flows. Lectures in Mathematics ETH Zürich, Birkhäuser,
+Basel (2005)
+4. Arbel, M., Korba, A., Salim, A., Gretton, A.: Maximum mean discrepancy gradient flow. In: Wallach,
+H., Larochelle, H., Beygelzimer, A., d Alché-Buc, F., Fox, E., Garnett, R. (eds.) Advances in Neural
+Information Processing Systems. vol. 32, pp. 1–11. Curran Associates Inc., New York, USA (2019)
+5. Beier, F., Beinert, R., Steidl, G.: On a linear Gromov–Wasserstein distance. IEEE Transactions on
+Image Processing 31, 7292–7305 (2022)
+6. Binkowski, M., Sutherland, D.J., Arbel, M., Gretton, A.: Demystifying MMD GANs. In: Proceedings
+ICLR 2018. OpenReview (2018)
+7. Bonaschi, G.A., Carrillo, J.A., Francesco, M.D., Peletier, M.A.: Equivalence of gradient flows and
+entropy solutions for singular nonlocal interaction equations in 1d. ESAIM Control Optimization and
+Calculus of Variation 21, 414–441 (2015)
+8. Bonet, C., Courty, N., Septier, F., Drumetz, L.: Efficient gradient flows in sliced-Wasserstein space.
+Transactions on Machine Learning Research (2022)
+9. Bonneel, N., Rabin, J., Peyré, G., Pfister, H.: Sliced and Radon Wasserstein barycenters of measures.
+Journal of Mathematical Imaging and Vision 1(51), 22–45 (2015)
+
+2.00
+1.75
+1.50
+1.25
+b6 1.00
+0.75
+0.50
+0.25
+0.00
+-1.00
+-0.75
+-0.50
+-0.25
+0.00
+0.25
+0.50
+0.75
+1.00
+m2.00
+1.75
+1.50
+1.25
+b 1.00
+0.75
+0.50
+0.25
+0.00
+1.00 -0.75-0.50 -0.25
+0.00
+0.25
+0.50
+0.75
+1.00
+m12
+J. Hertrich et al.
+10. Cai, T., Cheng, J., Schmitzer, B., Thorpe, M.: The linearized Hellinger-Kantorovich distance.
+arXiv:2102.08807 (2021)
+11. Carrillo, J.A., Huang, Y.: Explicit equilibrium solutions for the aggregation equation with power-law
+potentials. Kinetic and Related Models 10(1), 171–192 (2017)
+12. Chafaï, D., Saff, E.B., Womersley, R.S.: Threshold condensation to singular support for a Riesz equi-
+librium problem. arXiv:2206.04956v1 (2022)
+13. Dziugaite, G.K., Roy, D.M., Ghahramani, Z.: Training generative neural networks via maximum mean
+discrepancy optimization. In: Proceedings UAI 2015. UAI (2015)
+14. Ehler, M., Gräf, M., Neumayer, S., Steidl, G.: Curve based approximation of measures on manifolds by
+discrepancy minimization. Foundations of Computational Mathematics 21(6), 1595–1642 (2021)
+15. Feydy, J., Séjourné, T., Vialard, F.X., Amari, S., Trouvé, A., Peyré, G.: Interpolating between optimal
+transport and MMD using Sinkhorn divergences. In: Proc. of Machine Learning Research. vol. 89, pp.
+2681–2690. PMLR (2019)
+16. Gelbrich, M.: On a formula for the l2 Wasserstein metric between measures on Euclidean and Hilbert
+spaces. Mathematische Nachrichten 147(1), 185–203 (1990)
+17. Gutleb, T.S., Carrillo, J.A., Olver, S.: Computation of power law equilibrium measures on balls of
+arbitrary dimension. arXiv:2109.00843v1 (2021)
+18. Hertrich, J., Gräf, M., Beinert, R., Steidl, G.: Wasserstein steepest descent flows of disrepancies with
+Riesz kernels. arXiv:2211.01804 v1) (2022)
+19. Jordan, R., Kinderlehrer, D., Otto, F.: The variational formulation of the Fokker–Planck equation.
+SIAM Journal on Mathematical Analysis 29(1), 1–17 (1998)
+20. Kolouri, S., Park, S., Rohde, G.: The Radon cumulative distribution transform and its application to
+image classification. IEEE Transactions on Image Processing 25(2), 920–934 (2016)
+21. Landkof, N.: Foundations of Modern Potential Theory. Grundlehren der mathematischen Wis-
+senschaften, Springer, Berlin (1972)
+22. Li, C.L., Chang, W.C., Cheng, Y., Yang, Y., Póczos, B.: MMD GAN: Towards deeper understanding
+of moment matching network. arXiv:1705.08584 (2017)
+23. Liutkus, A., Simsekli, U., Majewski, S., Durmus, A., Stöter, F.R.: Sliced-wasserstein flows: Nonparamet-
+ric generative modeling via optimal transport and diffusions. In: Proc. of Machine Learning Research,
+vol. 97. PMLR (2019)
+24. Neumayer, S., Steidl, G.: From optimal transport to discrepancy. In: Chen, K., Schönlieb, C.B., Tai,
+X.C., Younes, L. (eds.) Handbook of Mathematical Models and Algorithms in Computer Vision and
+Imaging: Mathematical Imaging and Vision, pp. 1–36. Springer (2023)
+25. Nguyen, K., Ho, N., Pham, T., Bui, H.: Distributional sliced-wasserstein and applications to generative
+modeling. In: 9th International Conference on Learning Representations. IEEE (2021)
+26. Otto, F.: The geometry of dissipative evolution equations: the porous medium equation. Communica-
+tions in Partial Differential Equations 26, 101–174 (2001)
+27. Park, S., Kolouri, S., Kundu, S., Rohde, G.: The cumulative distribution transform and linear pattern
+classification. Applied and Computational Harmonic Analysis (2017)
+28. Pavliotis, G.A.: Stochastic processes and applications: Diffusion Processes, the Fokker-Planck and
+Langevin Equations. No. 60 in Texts in Applied Mathematics, Springer, New York (2014)
+29. Rockafellar, R.T., Royset, J.O.: Random variables, monotone relations, and convex analysis. Mathe-
+matical Programming 148, 297–331 (2014)
+30. Saff, E., Totik, V.: Logarithmic Potentials with External Fields. Grundlehren der mathematischen
+Wissenschaften, Springer, Berlin (1997)
+31. Santambrogio, F.: Optimal Transport for Applied Mathematicians, Progress in Nonlinear Differential
+Equations and their Applications, vol. 87. Birkhäuser, Basel (2015)
+32. Villani, C.: Topics in Optimal Transportation. No. 58 in Graduate Studies in Mathematics, American
+Mathematical Society, Providence (2003)
+33. Wendland, H.: Scattered Data Approximation. Cambridge University Press (2005)
+
diff --git a/ANE3T4oBgHgl3EQfTQrd/content/tmp_files/load_file.txt b/ANE3T4oBgHgl3EQfTQrd/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..620f2907078945421a4d5bf6e237bf8741fdc2cc
--- /dev/null
+++ b/ANE3T4oBgHgl3EQfTQrd/content/tmp_files/load_file.txt
@@ -0,0 +1,576 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf,len=575
+page_content='Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line⋆ Johannes Hertrich, Robert Beinert, Manuel Gräf, and Gabriele Steidl TU Berlin, Institute of Mathematics, Straße des 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Juni 136, 10623 Berlin, Germany {hertrich,beinert, graef,steidl}@math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='tu-berlin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='de https://tu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='berlin/imageanalysis/ Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' This paper provides results on Wasserstein gradient flows between measures on the real line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Utilizing the isometric embedding of the Wasserstein space P2(R) into the Hilbert space L2((0, 1)), Wasserstein gradient flows of functionals on P2(R) can be characterized as subgradient flows of associated functionals on L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For the maximum mean discrepancy functional Fν := D2 K(·, ν) with the non-smooth negative distance kernel K(x, y) = −|x − y|, we deduce a formula for the associated functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' This functional appears to be convex, and we show that Fν is convex along (generalized) geodesics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For the Dirac measure ν = δq, q ∈ R as end point of the flow, this enables us to determine the Wasserstein gradient flows analytically.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Various examples of Wasserstein gradient flows are given for illustration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Keywords: Maximum Mean Discrepancy · Wasserstein gradient flows · Riesz kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 1 Introduction Gradient flows provide a powerful tool for computing the minimizers of modeling functionals in certain applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In particular, gradient flows on the Wasserstein space are an interesting field of research that combines optimization with (stochastic) dynamical systems and differential geom- etry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For a good overview on the theory, we refer to the books of Ambrosio, Gigli and Savaré [3], and Santambrogio [31].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Besides Wasserstein gradient flows of the Kullback–Leibler (KL) functional KL(·, ν) and the associated Fokker–Planck equation related to the overdamped Langevin dynamics, which were extensively examined in the literature, see, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', [19,26,28], flows of maximum mean discrepancy (MMD) functionals Fν := D2 K(·, ν) became popular in machine learning [4] and image processing [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' On the other hand, MMDs were used as loss functions in generative adversarial networks [6,13,22].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Wasserstein gradient flows of MMDs are not restricted to absolutely continuous measures and have a rich structure depending on the kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' So the authors of [4] showed that for smooth kernels K, particle flows are indeed Wasserstein gradient flows meaning that Wasserstein flows starting at an empirical measure remain empirical measures and coincide with usual gradi- ent descent flows in Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The situation changes for non-smooth kernels like the negative distance, where empirical measures can become absolutely continuous ones and conversely, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' particles may explode.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The concrete behavior of the flow depends also on the dimension, see [11,12,17,18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The crucial part is the treatment of the so-called interaction energy within the discrepancy, which is repulsive and responsible for the proper spread of the measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' This nicely links to another field of mathematics, namely potential theory [21,30].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ⋆ Supported by the German Research Foundation (DFG) [grant numbers STE571/14-1, STE 571/16-1] and the Federal Ministry of Education and Research (BMBF, Germany) [grant number 13N15754].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='04441v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='OC] 11 Jan 2023 2 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In this paper, we are just concerned with Wasserstein gradient flows on the real line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Optimal transport techniques that reduce the original transport to those on the line were successfully used in several applications [1,5,9,10,20,27].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' When working on R, we can exploit quantile functions of measures to embed the Wasserstein space P2(R) into the Hilbert space of (equivalence classes) of square integrable functions L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then, instead of dealing with functionals on P2(R), we can just work with associated functionals which are uniquely defined on a cone of L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' If the asso- ciated functional is convex, we will see that the original one is convex along (generalized) geodesics, which is a crucial property for the uniqueness of the Wasserstein gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Furthermore, we can characterize Wasserstein gradient flows using regular subdifferentials in L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that the special case of Wasserstein gradient flows of the interaction energy was already considered in [7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We will have a special look at the Wasserstein gradient flow of Fδq := D2 K(·, δq) for the negative distance kernel, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' flows ending in δq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We will deduce an analytic formula for this flow and provide several examples to illustrate its behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Outline of the paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In Section 2, we recall the basic notation on Wasserstein gradient flows in d dimensions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then, in Section 3, we show how these flows can be simpler treated as gradient descent flows of an associated function on the Hilbert space L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' MMDs are introduced in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then, in Section 5, we restrict our attention again to the real line and show how the associated functional looks for the MMD with negative distance kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In particular, this functional is convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For the Dirac measure ν = δq, q ∈ R, we give an explicit formula for the Wasserstein gradient flow of the MMD functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Examples illustrating the behavior of the Wasserstein flows are provided in Section 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Finally, conclusions are drawn in Section 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 2 Wasserstein Gradient Flows Let M(Rd) denote the space of σ-additive, signed measures and P(Rd) the set of probability measures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For µ ∈ M(Rd) and measurable T : Rd → Rn, the push-forward of µ via T is given by T#µ := µ ◦ T −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We consider the Wasserstein space P2(Rd) := {µ ∈ P(Rd): � Rd ∥x∥2 2 dµ(x) < ∞} equipped with the Wasserstein distance W2 : P2(Rd) × P2(Rd) → [0, ∞), W 2 2 (µ, ν) := min π∈Γ (µ,ν) � Rd×Rd ∥x − y∥2 2 dπ(x, y), µ, ν ∈ P2(Rd), (1) where Γ(µ, ν) := {π ∈ P2(Rd × Rd) : (π1)#π = µ, (π2)#π = ν} and πi(x) := xi, i = 1, 2 for x = (x1, x2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The set of optimal transport plans π realizing the minimum in (1) is denoted by Γ opt(µ, ν).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' A curve γ : I → P2(Rd) on an interval I ⊂ R, is called a geodesic if there exists a constant C ≥ 0 such that W2(γ(t1), γ(t2)) = C|t2 − t1|, for all t1, t2 ∈ I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The Wasserstein space is a geodesic space, meaning that any two measures µ, ν ∈ P2(Rd) can be connected by a geodesic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The regular tangent space at µ ∈ P2(Rd) is given by TµP2(Rd) := � λ(T − Id) : (Id, T)#µ ∈ Γ opt(µ, T#µ), λ > 0 �L2,µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Here L2,µ denotes the Bochner space of (equivalence classes of) functions ξ : Rd → Rd with finite ∥ξ∥2 L2,µ := � Rd ∥ξ(x)∥2 2 dµ(x) < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that TµP2(Rd) is not a “classical” tangent space, in Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line 3 particular it is an infinite dimensional subspace of L2,µ if µ is absolutely continuous and just Rd if µ = δx, x ∈ Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In particular, this means that the Wasserstein space has only a “manifold-like” structure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For λ ∈ R, a function F : P2(Rd) → (−∞, +∞] is called λ-convex along geodesics if, for every µ, ν ∈ dom F := {µ ∈ P2(Rd) : F(µ) < ∞}, there exists at least one geodesic γ : [0, 1] → P2(Rd) between µ and ν such that F(γ(t)) ≤ (1 − t) F(µ) + t F(ν) − λ 2 t(1 − t) W 2 2 (µ, ν), t ∈ [0, 1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In the case λ = 0, we just speak about convex functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For a proper and lower semi-continuous (lsc) function F : P2(Rd) → (−∞, ∞] and µ ∈ P2(Rd), the reduced Fréchet subdifferential at µ is defined as ∂F(µ) := � ξ ∈ L2,µ : F(ν) − F(µ) ≥ inf π∈Γ opt(µ,ν) � R2d ⟨ξ(x), y − x⟩ dπ(x, y) + o(W2(µ, ν)) ∀ν ∈ P2(Rd) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (2) A curve γ : I → P2(Rd) is absolutely continuous, if there exists a Borel velocity field vt : Rd → Rd with � I ∥vt∥L2,γ(t) dt < +∞ such that ∂tγ(t) + ∇x · (vt γ(t)) = 0 (3) on I × Rd in the distributive sense, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', for all ϕ ∈ C∞ c (I × Rd) it holds � I � Rd ∂tϕ(t, x) + vt(x) · ∇x ϕ(t, x) dγ(t) dt = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' A locally absolutely continuous curve γ : (0, +∞) → P2(Rd) with velocity field vt ∈ Tγ(t)P2(Rd) is called a Wasserstein gradient flow with respect to F : P2(Rd) → (−∞, +∞] if vt ∈ −∂F(γ(t)), for a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' t > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (4) 3 Wasserstein Gradient Flows on the Line Now we restrict our attention to d = 1, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', we work on the real line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We will see that the above notation simplifies since there is an isometric embedding of P2(R) into L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' To this end, we consider the cumulative distribution function Rµ : R → [0, 1] of µ ∈ P2(R), which is defined by Rµ(x) := µ((−∞, x]), x ∈ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' It is non-decreasing and right-continuous with limx→−∞ Rµ(x) = 0 as well as limx→∞ Rµ(x) = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The quantile function Qµ : (0, 1) → R is the generalized inverse of Rµ given by Qµ(p) := min{x ∈ R: Rµ(x) ≥ p}, p ∈ (0, 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' It is non-decreasing and left-continuous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The quantile functions form a convex cone C((0, 1)) := {Q ∈ L2((0, 1)) : Q nondecreasing} in L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that both the distribution and quantile functions are continuous except for at most countably many jumps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For a good overview see [29, § 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' By the following theorem, the mapping µ �→ Qµ is an isometric embedding of P2(R) into L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Theorem 1 ([32, Thm 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='18]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For µ, ν ∈ P2(R), the quantile function Qµ ∈ C((0, 1)) satisfies µ = (Qµ)#λ(0,1) and W 2 2 (µ, ν) = � 1 0 |Qµ(s) − Qν(s)|2ds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 4 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Next we will see that instead of working with functionals F : P2(R) → (−∞, +∞], we can just deal with associated functionals F: L2((0, 1)) → (−∞, ∞] fulfilling F(Qµ) := F(µ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that F is defined in this way only on C((0, 1)), and there exist several continuous extensions to the whole linear space L2((0, 1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Instead of the extended Fréchet subdifferential (2), we will use the regular subdifferential in L2((0, 1)) defined by ∂G(f) := � h ∈ L2((0, 1)) : G(g) ≥ G(f) + ⟨h, g − f⟩ + o(∥g − f∥L2) ∀g ∈ L2((0, 1)) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The following theorem characterizes Wasserstein gradient flows by this regular subdifferential and states a convexity relation between F : P2(R) → (−∞, +∞] and the associated functional F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' i) Let γ : (0, ∞) → P2(R) be a locally absolutely continuous curve and F: L2((0, 1)) → (−∞, ∞] such that the pointwise derivative ∂tQγ(t) exists and fulfills the L2 subgradient equation ∂tQγ(t) ∈ −∂F(Qγ(t)), for almost every t ∈ (0, +∞).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then γ is a Wasserstein gradient flow with respect to the functional F : P2(R) → (−∞, +∞] defined by F(µ) := F(Qµ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ii) If F : C((0, 1)) → (−∞, ∞] is convex, then F(µ) := F(Qµ) is convex along geodesics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' i) Since γ is (locally) absolute continuous, the velocity field vt from (3) fulfills by [3, Prop 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='6] for almost every t ∈ (0, ∞) the relation 0 = lim h→0 W2(γ(t + h), (Id + hvt)#γ(t)) |h| = lim h→0 W2((Qγ(t+h))#λ(0,1), � Qγ(t) + h(vt ◦ Qγ(t)) � #λ(0,1)) |h| = lim h→0 ���Qγ(t+h) − Qγ(t) h − vt ◦ Qγ(t) ��� L2 = ∥∂tQγ(t) − vt ◦ Qγ(t)∥L2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Thus, by assumption, vt ◦ Qγ(t) ∈ −∂F(Qγ(t)) a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In particular, for any µ ∈ P2(R), we obtain 0 ≤ F(Qµ) − F(Qγ(t)) + � 1 0 vt(Qγ(t)(s)) (Qµ(s) − Qγ(t)(s)) ds + o(∥Qµ − Qγ(t)∥L2) = F(µ) − F(γ(t)) + � R×R vt(x) (y − x) dπ(x, y) + o � W2(µ, γ(t)) � , where π := (Qγ(t), Qµ)#λ(0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Since π the unique optimal transport plan between γ(t) and µ, this yields by (2) that vt ∈ −∂F(γ(t)) showing the assertion by (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ii) Let F: L2((0, 1)) → R be convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For any geodesic γ : [0, 1] → P2(R), since µ �→ Qµ is an isometry, the curve t �→ Qγ(t) is a geodesic in L2((0, 1)) too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Since L2((0, 1)) is a linear space, the convexity of F: L2((0, 1)) → R yields that t �→ F(Qγ(t)) = F(γ(t)) is convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Thus, F is convex along γ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ⊓⊔ Remark 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' If F : P2(R) → (−∞, +∞] is proper, lsc, coercive and λ-convex along so-called general- ized geodesics, then the Wasserstein gradient flow starting at any µ0 ∈ dom F is uniquely determined and is the uniform limit of the miminizing movement scheme of Jordan, Kinderlehrer and Otto [19] when the time step size goes to zero, see [3, Thm 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In R, but not in higher dimensions, λ-convex functions along geodesics fulfill also the stronger property that they are λ-convex along generalized geodesics, see [18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line 5 4 Discrepancies We consider symmetric and conditionally positive definite kernels K : Rd × Rd → R of order one, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', for any n ∈ N, any pairwise different points x1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' , xn ∈ Rd and any a1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' , an ∈ R with �n i=1 ai = 0 the relation �n i,j=1 aiajK(xi, xj) ≥ 0 is satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Typical examples are Riesz kernels K(x, y) := −∥x − y∥r, r ∈ (0, 2), where we have strict inequality except for all aj, j = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' , n being zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The maximum mean discrepancy (MMD) D2 K : P(Rd) × P(Rd) → R between two measures µ, ν ∈ P(Rd) is defined by D2 K(µ, ν) := EK(µ − ν) with the so-called K-energy on signed measures EK(σ) := 1 2 � Rd � Rd K(x, y) dσ(x)dσ(y), σ ∈ M(Rd).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The relation between discrepancies and Wasserstein distances is discussed in [15,24].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For fixed ν ∈ P(Rd), the MMD can be decomposed as Fν(µ) = D2 K(µ, ν) = EK(µ) + VK,ν(µ) + EK(ν) � �� � const.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' with the interaction energy on probability measures EK(µ) = 1 2 � Rd � Rd K(x, y) dµ(x)dµ(y), µ ∈ P2(Rd) and the potential energy of µ with respect to the potential of ν, VK,ν(µ) := � Rd VK,ν(y)dµ(x), VK,ν(x) := − � Rd K(x, y)dν(y).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In dimensions d ≥ 2 neither EK nor D2 K with the Riesz kernel are λ-convex along geodesics, see [18], so that certain properties of Wasserstein gradient flows do not apply.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We will see that this is different on the real line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 5 MMD Flows on the Line In the rest of this paper, we restrict our attention to d = 1 and negative distance K(x, y) = −|x − y|, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' to Riesz kernels with r = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For fixed ν ∈ P2(R), we consider the MMD functional Fν := D2 K(·, ν).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that the unique minimizer of this functional is given by µ = ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Let Fν := D2 K(·, ν) with the negative distance kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then the convex functional Fν : L2((0, 1)) → R defined by Fν(f) := � 1 0 � (1 − 2s)(f(s) + Qν(s)) + � 1 0 |f(s) − Qν(t)| dt � ds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (5) fulfills Fν(Qµ) = Fν(µ) for all µ ∈ P2(R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In particular, Fν is convex along (generalized) geodesics and there exists a unique Wasserstein gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 6 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We reformulate Fν as Fν(µ) = −1 2 � R×R |x − y|(dµ(x) − dν(x))(dµ(y) − dν(y)) = −1 2 � 1 0 � 1 0 |Qµ(s) − Qµ(t)| − 2|Qµ(s) − Qν(t)| + |Qν(s) − Qν(t)| ds dt = � 1 0 � 1 t Qµ(t) − Qµ(s) + Qν(t) − Qν(s) ds dt + � 1 0 � 1 0 |Qµ(s) − Qν(t)| ds dt = � 1 0 � 1 t Qµ(t) + Qν(t) ds dt − � 1 0 � s 0 Qµ(s) + Qν(s) dt ds + � 1 0 � 1 0 |Qµ(s) − Qν(t)| ds dt = � 1 0 � (1 − 2s)(Qµ(s) + Qν(s)) + � 1 0 |Qµ(s) − Qν(t)| dt � ds, which yields the first claim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The second one follows by Theorem 2ii) and Remark 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ⊓⊔ Note that the lemma cannot immediately be generalized to Riesz kernels with r = (1, 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Finally, we derive for the special choice ν = δq in D2 K(·, ν) an analytic formula for its Wasserstein gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Proposition 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Let Fδq := D2 K(·, δq) with the negative distance kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then the unique Wasser- stein gradient flow of Fδq starting at µ0 = γ(0) ∈ P2(R) is γ(t) = (gt)#λ(0,1), where the function gt : (0, 1) → R is given by gt(s) := � � � � � � � min{Qµ0(s) + 2st, q}, Qµ0(s) < q, q, Qµ0(s) = q, max{Qµ0(s) + 2st − 2t, q}, Qµ0(s) > q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (6) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' First, note that gt ∈ C((0, 1)) such that it holds gt = Qγ(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Since Qδq ≡ q, the subdifferential of Fδq in (5) at gt consists of all functions h(s) = � � � � � � � −2s, Qµ0(s) < q and t < q−Qµ0(s) 2s , 2 − 2s, Qµ0(s) > q and t < Qµ0(s)−q 2−2s , 1 − 2s + n(s), otherwise, with −1 ≤ n(s) ≤ 1 for s ∈ (0, 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' On the other hand, the pointwise derivative of gt in (6) can be written as ∂tgt(s) = � � � � � � � 2s, Qµ0(s) < q and t < q−Qµ0(s) 2s , 2s − 2, Qµ0(s) > q and t < Qµ0(s)−q 2−2s , 0, otherwise, such that we obtain ∂tQγ(t) = ∂tgt ∈ −∂Fν(gt) = −∂Fν(Qγ(t)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Thus, by Lemma 1 and Theorem 2, we obtain that γ is a Wasserstein gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' It is unique since Fν is convex along geodesics by Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='ii, Lemma 1 and Remark 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ⊓⊔ Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line 7 6 Intuitive Examples Finally, we provide some intuitive examples of Wasserstein gradient flows of Fν := D2 K(·, ν) with the negative distance kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='1 Flow between Dirac Measures We consider the flow of Fδ0 starting at the initial measure γ(0) = µ0 := δ−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Due to Qδ0 ≡ 0, Proposition 1 yields the gradient flow γ(t) := (Qt)#λ(0,1) given by γ(t) = � � � � � � � δ−1, t = 0, 1 2tλ[−1,−1+2t], 0 ≤ t ≤ 1 2, 1 2tλ[−1,0] + � 1 − 1 2t � δ0, 1 2 < t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For t ∈ (0, 1 2], the initial Dirac measure becomes a uniform measure with increasing support, and for t ∈ ( 1 2, 1) it is the convex combination of a uniform measure and δ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' A visualization of the flow is given in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' □ Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 1: Visualization of the Wasserstein gradient flow of Fδ0 from δ−1 to δ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' At various times t, the absolute continuous part is visualized by its density in blue (area equals mass) and the atomic part by the red dotted vertical line (height equals mass).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The atomic part at the end point x = 0 starts to grow at time t = 1 2, where the support of the density touches this point for the first time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='2 Flow on Restricted Sets Next, we are interested in the Wasserstein gradient flows on the subsets Si, i = 1, 2, given by (i) S1 := {δx : x ∈ R}, (ii) S2 := {µm,σ = 1 2 √ 3σλ[m− √ 3σ,m+ √ 3σ] : m ∈ R, σ ∈ R≥0}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that S2 is a special instance of sets of scaled and translated measures µ ∈ P2(R) defined by {Ta,b#µ : a ∈ R≥0, b ∈ R}, where Ta,b(x) := ax + b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' As mentioned in [16] the Wasserstein distance between measures µ1, µ2 from such sets has been already known to Fréchet: W 2 2 (µ1, µ2) = |m1 − m2|2 + |σ1 − σ2|2, t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='29 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='33 t=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='40 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='67 t= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 t= 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 t= 8 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1 0 0 0 0 0 0 08 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' where mi and σi are the mean value and standard deviation of µi, i = 1, 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' This provides an isometric embedding of R×R≥0 into P2(R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The boundary of S2 is the set of Dirac measures S1 and is isometric to R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The sets are convex in the sense that for µ, ν ∈ Si all geodesics γ : [0, 1] → P(R) with γ(0) = µ and γ(1) = ν are in Si, i ∈ {1, 2}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For i = 1, 2, we consider Fi,ν(µ) := � Fν µ ∈ Si, +∞ otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Due to the convexity of Fν along geodesics and the convexity of the sets Si, we obtain that the functions Fi,ν are convex along geodesics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Flows of F1,ν We use the notation fx ≡ x for the constant function on (0, 1) with value x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' It is straightforward to check that the function F: L2((0, 1)) → (−∞, ∞] given by F(f) = � F(x), if f = fx for some x ∈ R, +∞, otherwise, with F(x) := � R |x − y| dν(y) − 1 2 � R×R |y − z| dν(y)dν(z) fulfills F(Qµ) = F1,ν(µ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In the following, we aim to find x: [0, ∞) → R satisfying ˙x(t) = −∂F(x(t)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Since the set {Qµ : µ ∈ S1} is a one-dimensional linear subspace of L2((0, 1)) spanned by the constant one-function f1, this yields fx(t) ∈ −∂F(fx(t)) such that the Wasserstein gradient flow is by Theorem 2 given by γ(t) = (fx(t))#λ(0,1) = δx(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In the special case ν = δq for some q ∈ R, we have F(x) = |x − q|, ∂F(x) = � � � � � � � {−1}, x < q, [−1, 1], x = q, {1}, x > q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Therefore, the Wasserstein gradient flow for x(0) = x0 ̸= 0 is given by γ(t) = δx(t), with x(t) = � x0 + t, x0 < q, x0 − t, x0 > q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' , 0 ≤ t < |x0 − q| and γ(t) = δq for t ≥ |x0 − q|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For ν = 1 2λ[−1,1] the gradient flow starting at x0 ∈ [−1, 1] is x(t) = x0e−t, t ≥ 0, and converges to the midpoint of the interval for t → ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' If it starts at x0 ∈ R \\ [−1, 1] the gradient flow is x(t) = � x0 + t, x0 < −1, x0 − t, x0 > 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' , 0 ≤ t ≤ min |x0 − 1|, |x0 + 1|, where it reaches the nearest interval end point in finite time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In Figure 2, we plotted the x(t) for different initial values x(0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line 9 Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 2: Wasserstein gradient flow of F1,ν for ν = δ0 (left) and ν = 1 2λ[−1,1] (right) from various initial points δx, x ∈ [−2, 2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The support of the right measure ν is depicted by the blue region.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The examples show that gradient flows may reach the optimal points in finite or infinite time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Flows of F2,ν We observe that Qµm,σ = fm,σ, where fm,σ(x) = m + 2 √ 3σ(x − 1 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' By Lemma 1 we obtain that the function F: L2((0, 1)) → (−∞, ∞] given by F(f) = � F(m, σ), if f = fm,σ for (m, σ) ∈ R × R≥0, +∞, otherwise, fulfills F(Qµ) = F2,ν(µ), where F(m, σ) := � (0,1) (1 − 2s)(fm,σ(s) + Qν(s))ds + � (0,1)2 |fm,σ(s) − Qν(t)|dtds, The set {fm,σ : m, σ ∈ R} is a two dimensional linear subspace of L2((0, 1)) with orthonormal basis {f1,0, f0,1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We aim to compute m: [0, ∞) → R and σ: [0, ∞) → R≥0 with ( ˙m(t), ˙σ(t)) = −∂F(m(t), σ(t)), t ∈ I ⊂ R, (7) because this yields fm(t),σ(t) ∈ −∂F(fm(t),σ(t)) such that γ(t) = (fm(t),σ(t))#λ(0,1) = µm,σ is by Theorem 2 the Wasserstein gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In the following, we consider the special case ν = δ0 = µ0,0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Then, the function F reduces to F(m, σ) = � R (1 − 2s)(m + 2 √ 3σ(s − 1 2)) + |m + 2 √ 3σ(s − 1 2)|ds = − σ √ 3 + � � � |m|, if |m| ≥ √ 3σ, m2+3σ2 2 √ 3σ2 if |m| < √ 3σ, and the subdifferential is given by ∂F(m, σ) = � � � sgn(m) × {− 1 √ 3}, if |m| ≥ √ 3σ, {( m √ 3σ2 , −m2 √ 3σ3 − 1 √ 3)}, if |m| < √ 3σ, sgn(m) = � { |m| m }, if m ̸= 0, [−1, 1], if m = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We observe that F is differentiable for σ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Thus, for any initial intial value (m(0), σ(0)) = (m0, σ0), we can compute the trajectory (m(t), σ(t)) solving (7) using an ODE solver.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In Figure 3 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 X 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 t2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 X 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 t10 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (left), we plotted the level sets of the function F(m, σ) as well as the solution trajectory (m(t), σ(t)) for different initial values (m(0), σ(0)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' For (m(0), σ(0)) = (−1, 0), the resulting flow is illustrated in Figure 3, right.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 3: Wasserstein gradient flow F2,δ0 from (m(0), σ(0)) to δ0 (left) and from δ−1 to δ0 (right).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In contrast Figure 1 it is a uniform measure for all t ∈ (0, 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Flows for a Smooth Kernel For smooth, positive definite kernels K the MMD functional Fν := D2 K(·, ν) is in general not convex and leads to a more complex energy landscape than for the negative distance kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' This may lead to problems for optimization algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' To illustrate this observation, we let ν := λ[−1,1] and compare the energy landscape of the restricted functional F2,ν for K(x, y) := −|x − y| and the kernel ˜K(x, y) := � (1 − 1 2|x − y|)2(|x − y| + 1), |x − y| ≤ 2, 0, else.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (8) In contrast to the negative distance kernel K, the kernel ˜K is positive definite (without restrictions on the ai), cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' [33], and has a Lipschitz continuous gradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The two energy landscapes of F2,ν are visualized in Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The non-convexity of Fν for ˜K is readily seen by the presence of a saddle point for F2,ν at µ = δ0 (equivalently to (m, σ) = (0, 0) in the mσ-plane).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Note that any Wasserstein gradient flow of Fν starting at a Dirac measure δx converges to this saddle point µ = δ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 7 Conclusions We provided insight into Wasserstein gradient flows of MMD functionals with negative distance kernels and characterized in particular flows ending in a Dirac measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' We have seen that such flows are not simple particle flows, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' starting in another Dirac measure the flow becomes immediately uniformly distributed and after a certain time a mixture of a uniform and a Dirac measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In our future work, we want to extend our considerations to empirical measures and incorporate 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 b 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 mt= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 t= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 t= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 t= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 t= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 t= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 t= 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 t = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='0 1 0 0 0 0 0 1 0Wasserstein Gradient Flows of the Discrepancy with Distance Kernel on the Line 11 Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 4: Visualization of the energy landscapes of F2,λ[−1,1] for the convex negative distance kernel (left) and the non-convex, smooth kernel given in (8).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The red dot is the global minimizer λ[−1,1] (left and right) and the blue point (right) is the saddle point δ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' The black lines depict selected gradient flows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' deep learning techniques as in [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Also the treatment of other functionals which incorporate an interaction energy part appears to be interesting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Further, we may combine univariate techniques with multivariate settings using Radon transform like techniques as in [8,23,25].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' References 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Abraham, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Abraham, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Bergounioux, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Carlier, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Tomographic reconstruction from a few views: A multi-marginal optimal transport approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Applied Mathematics and Optimization 75(1), 55–73 (2017) 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Altekrüger, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Hertrich, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Steidl, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Neural Wasserstein gradient flows for maximum mean discrep- ancies with Riesz kernels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:XXX (2023) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Ambrosio, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Gigli, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Savare, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Gradient Flows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Lectures in Mathematics ETH Zürich, Birkhäuser, Basel (2005) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Arbel, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Korba, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Salim, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Gretton, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Maximum mean discrepancy gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: Wallach, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Larochelle, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Beygelzimer, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', d Alché-Buc, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Fox, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Garnett, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=') Advances in Neural Information Processing Systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 32, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 1–11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Curran Associates Inc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', New York, USA (2019) 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Beier, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Beinert, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Steidl, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': On a linear Gromov–Wasserstein distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' IEEE Transactions on Image Processing 31, 7292–7305 (2022) 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Binkowski, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Sutherland, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Arbel, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Gretton, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Demystifying MMD GANs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: Proceedings ICLR 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' OpenReview (2018) 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Bonaschi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Carrillo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Francesco, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Peletier, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' : Equivalence of gradient flows and entropy solutions for singular nonlocal interaction equations in 1d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' ESAIM Control Optimization and Calculus of Variation 21, 414–441 (2015) 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Bonet, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Courty, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Septier, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Drumetz, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Efficient gradient flows in sliced-Wasserstein space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Transactions on Machine Learning Research (2022) 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Bonneel, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Rabin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Peyré, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Pfister, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Sliced and Radon Wasserstein barycenters of measures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Journal of Mathematical Imaging and Vision 1(51), 22–45 (2015) 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 b6 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 m2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 b 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00 m12 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Cai, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Cheng, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Schmitzer, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Thorpe, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': The linearized Hellinger-Kantorovich distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:2102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='08807 (2021) 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Carrillo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Huang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Explicit equilibrium solutions for the aggregation equation with power-law potentials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Kinetic and Related Models 10(1), 171–192 (2017) 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Chafaï, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Saff, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Womersley, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' : Threshold condensation to singular support for a Riesz equi- librium problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='04956v1 (2022) 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Dziugaite, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Roy, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Ghahramani, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Training generative neural networks via maximum mean discrepancy optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: Proceedings UAI 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' UAI (2015) 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Ehler, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Gräf, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Neumayer, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Steidl, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Curve based approximation of measures on manifolds by discrepancy minimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Foundations of Computational Mathematics 21(6), 1595–1642 (2021) 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Feydy, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Séjourné, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Vialard, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Amari, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Trouvé, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Peyré, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Interpolating between optimal transport and MMD using Sinkhorn divergences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' of Machine Learning Research.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 89, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 2681–2690.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' PMLR (2019) 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Gelbrich, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': On a formula for the l2 Wasserstein metric between measures on Euclidean and Hilbert spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Mathematische Nachrichten 147(1), 185–203 (1990) 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Gutleb, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Carrillo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Olver, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Computation of power law equilibrium measures on balls of arbitrary dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:2109.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='00843v1 (2021) 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Hertrich, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Gräf, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Beinert, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Steidl, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Wasserstein steepest descent flows of disrepancies with Riesz kernels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:2211.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='01804 v1) (2022) 19.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Jordan, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Kinderlehrer, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Otto, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': The variational formulation of the Fokker–Planck equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' SIAM Journal on Mathematical Analysis 29(1), 1–17 (1998) 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Kolouri, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Park, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Rohde, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': The Radon cumulative distribution transform and its application to image classification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' IEEE Transactions on Image Processing 25(2), 920–934 (2016) 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Landkof, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Foundations of Modern Potential Theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Grundlehren der mathematischen Wis- senschaften, Springer, Berlin (1972) 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Chang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Cheng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Yang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Póczos, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': MMD GAN: Towards deeper understanding of moment matching network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' arXiv:1705.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='08584 (2017) 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Liutkus, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Simsekli, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Majewski, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Durmus, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Stöter, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' : Sliced-wasserstein flows: Nonparamet- ric generative modeling via optimal transport and diffusions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' of Machine Learning Research, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' PMLR (2019) 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Neumayer, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Steidl, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': From optimal transport to discrepancy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: Chen, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Schönlieb, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Tai, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Younes, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=') Handbook of Mathematical Models and Algorithms in Computer Vision and Imaging: Mathematical Imaging and Vision, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 1–36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Springer (2023) 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Nguyen, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Ho, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Pham, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Bui, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Distributional sliced-wasserstein and applications to generative modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' In: 9th International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' IEEE (2021) 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Otto, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': The geometry of dissipative evolution equations: the porous medium equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Communica- tions in Partial Differential Equations 26, 101–174 (2001) 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Park, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Kolouri, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Kundu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Rohde, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': The cumulative distribution transform and linear pattern classification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Applied and Computational Harmonic Analysis (2017) 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Pavliotis, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' : Stochastic processes and applications: Diffusion Processes, the Fokker-Planck and Langevin Equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 60 in Texts in Applied Mathematics, Springer, New York (2014) 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Rockafellar, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Royset, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content='O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' : Random variables, monotone relations, and convex analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Mathe- matical Programming 148, 297–331 (2014) 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Saff, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=', Totik, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Logarithmic Potentials with External Fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Grundlehren der mathematischen Wissenschaften, Springer, Berlin (1997) 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Santambrogio, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Optimal Transport for Applied Mathematicians, Progress in Nonlinear Differential Equations and their Applications, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Birkhäuser, Basel (2015) 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Villani, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Topics in Optimal Transportation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' 58 in Graduate Studies in Mathematics, American Mathematical Society, Providence (2003) 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Wendland, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=': Scattered Data Approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
+page_content=' Cambridge University Press (2005)' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/ANE3T4oBgHgl3EQfTQrd/content/2301.04441v1.pdf'}
diff --git a/BdE5T4oBgHgl3EQfTA_5/vector_store/index.faiss b/BdE5T4oBgHgl3EQfTA_5/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..4d1e4957496d167e41e566504fb1770fd951fcc9
--- /dev/null
+++ b/BdE5T4oBgHgl3EQfTA_5/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05f54f0e8d1420af202149feeb0607052a15e9b3552b0d063a402fdcbb399a2c
+size 7602221
diff --git a/C9E0T4oBgHgl3EQfyQLe/vector_store/index.faiss b/C9E0T4oBgHgl3EQfyQLe/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..44e4a199d59620f3a1bae1e7c01f2cdd473dd511
--- /dev/null
+++ b/C9E0T4oBgHgl3EQfyQLe/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6918807b1f7f5891763fc7c6b6751b3d3a92b8fb91c69d92caf37561500ce62
+size 589869
diff --git a/CdAzT4oBgHgl3EQfiP0X/content/2301.01495v1.pdf b/CdAzT4oBgHgl3EQfiP0X/content/2301.01495v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..5628f9cf33527aa301e737056c761b3d92557fe3
--- /dev/null
+++ b/CdAzT4oBgHgl3EQfiP0X/content/2301.01495v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:031f268e3ab4661fd80f035a7a4b4c6c4de30c88c0ffc15431163e35c6d07c16
+size 582953
diff --git a/CdAzT4oBgHgl3EQfiP0X/vector_store/index.faiss b/CdAzT4oBgHgl3EQfiP0X/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..2e19fb40d3d4bcc1da45866d0e0a561344237443
--- /dev/null
+++ b/CdAzT4oBgHgl3EQfiP0X/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f504e946307252ea22160694253c63efadeed2aa676ac2fc32a7336a8852d22d
+size 3407917
diff --git a/CdAzT4oBgHgl3EQfiP0X/vector_store/index.pkl b/CdAzT4oBgHgl3EQfiP0X/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..46da74810cc70052993a3a2fbc45723b0e2dd469
--- /dev/null
+++ b/CdAzT4oBgHgl3EQfiP0X/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f68befa56d6cf11f7354625b2b26c0e36b415ed94ad34e4d51003ac680acb0d1
+size 121463
diff --git a/CdE1T4oBgHgl3EQfpwUV/vector_store/index.faiss b/CdE1T4oBgHgl3EQfpwUV/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..eeaf57d3d4c10a31e240293e20874b9a85ad2646
--- /dev/null
+++ b/CdE1T4oBgHgl3EQfpwUV/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61bf4cce9f8f6939ad6be3f8aa471a1cf411db7e3308fc0bec6663e05fedf283
+size 2097197
diff --git a/DtAyT4oBgHgl3EQf4vpJ/content/2301.00791v1.pdf b/DtAyT4oBgHgl3EQf4vpJ/content/2301.00791v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..22db39e431f047fec3d7486f43e8bf6578d7cb5c
--- /dev/null
+++ b/DtAyT4oBgHgl3EQf4vpJ/content/2301.00791v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:006ba2a53e150fb3e65d3c826fac2340566936cf7c09cf04b9709c891d750032
+size 492988
diff --git a/DtAyT4oBgHgl3EQf4vpJ/vector_store/index.faiss b/DtAyT4oBgHgl3EQf4vpJ/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..fcb1638a2651adfab9684a6094f6e77df36909fd
--- /dev/null
+++ b/DtAyT4oBgHgl3EQf4vpJ/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7aef4c4148501d31fbb7c51bceabab33f99e40991868cd469dfcd17bc1349e5a
+size 5505069
diff --git a/DtAyT4oBgHgl3EQf4vpJ/vector_store/index.pkl b/DtAyT4oBgHgl3EQf4vpJ/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..0af114580ecaa31fa27f34c4894493ec27693b7b
--- /dev/null
+++ b/DtAyT4oBgHgl3EQf4vpJ/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5596f5844898b25eb82dc2fc236b3c18f43b2c3288b22960df08b618c5e62e2e
+size 226102
diff --git a/DtFRT4oBgHgl3EQfATdu/content/tmp_files/2301.13461v1.pdf.txt b/DtFRT4oBgHgl3EQfATdu/content/tmp_files/2301.13461v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..238afc53c119974e29deea40eed1468e2991eda7
--- /dev/null
+++ b/DtFRT4oBgHgl3EQfATdu/content/tmp_files/2301.13461v1.pdf.txt
@@ -0,0 +1,1353 @@
+Investigating fission dynamics of neutron shell closed nuclei 210Po, 212Rn and 213Fr
+within a stochastic dynamical approach
+Divya Arora, P. Sugathan,∗ and A. Chatterjee
+Inter-University Accelerator Centre, Aruna Asaf Ali Marg, New Delhi 110067, India
+Dissipative dynamics of nuclear fission is a well confirmed phenomenon described either by a
+Kramers-modified statistical model or by a dynamical model employing the Langevin equation.
+Though dynamical models as well as statistical models incorporating fission delay are found to
+explain the measured fission observables in many studies, it nonetheless shows conflicting results
+for shell closed nuclei in the mass region 200.
+Analysis of recent data for neutron shell closed
+nuclei in excitation energy range 40−80 MeV failed to arrive at a satisfactory description of the
+data and attributed the mismatch to shell effects and/or entrance channel effects, without reaching
+a definite conclusion. In the present work we show that a well established stochastic dynamical
+code simultaneously reproduces the available data of pre-scission neutron multiplicities, fission and
+evaporation residue excitation functions for neutron shell closed nuclei 210Po and 212Rn and their
+isotopes 206Po and 214,216Rn without the need for including any extra shell or entrance channel
+effects. The calculations are performed by using a phenomenological universal friction form factor
+with no ad-hoc adjustment of model parameters. However, we note significant deviation, beyond
+experimental errors, in some cases of Fr isotopes.
+I.
+INTRODUCTION
+Fission of atomic nuclei is considered to be one of the
+most complex physical phenomena in nuclear physics. It
+involves rapid re-arrangement of nuclear matter with a
+delicate interplay between the macroscopic bulk matter
+and the microscopic quantal properties [1, 2]. Though
+properties of fission have been studied exhaustively, many
+aspects of the dynamics are still not well-understood. For
+instance, discrepancies are reported between the mea-
+sured fission observables and the predictions of the clas-
+sical theory based on the standard Bohr-Wheeler statis-
+tical model of fission [3].
+Fission hindrance, enhanced
+pre-scission particle and giant dipole resonance (GDR)
+γ-ray multiplicities observed in hot nuclei suggested the
+effects of nuclear dissipation slowing down the fission pro-
+cess [4–10]. To account for frictional effects, Kramers dif-
+fusion model formalism with modified fission width [11],
+referred to as Kramers-modified statistical model was in-
+cluded in the standard statistical theory.
+Although nature and strength of the nuclear dissipa-
+tion have been studied quite extensively, a simultaneous
+description of the experimental observables, namely, pre-
+scission neutron multiplicities (νpre), fission excitation
+functions and evaporation residue (ER) cross-sections
+still remains challenging.
+Additionally, the dissipation
+coefficient is treated as an adjustable free parameter in
+the statistical model analysis. The pre-fission lifetime (or
+the dissipation strength), the level density parameter at
+ground state and saddle point deformation and fission
+barrier are empirically fitted to explain the νpre and/or
+fission and ER cross-section data [6, 12–15]. As a result,
+the conclusions reached are often system dependent and
+are inadequate to provide a consistent description of the
+∗ sugathan@gmail.com
+fission process.
+Inadequate modelling of fission in statistical model can
+drastically influence the understanding of the fission phe-
+nomenon [16, 17].
+This is especially observed in mass
+(A) ≈ 200 region that is explored here, to understand the
+role of N=126 neutron shell closure in the fissioning com-
+pound nucleus (CN). An anomalous increase in the exper-
+imental fission fragment angular anisotropy was reported
+for 210Po (N=126) as compared to 206Po (non-shell closed
+nuclei) across an excitation energy range (Eex) ≈ 40−60
+MeV and was conjectured to be a manifestation of shell
+effects at the unconditional saddle [18]. Further, a con-
+siderable amount of saddle shell correction was invoked
+to describe the experimental νpre data for 210Po nuclei
+[19]. However, a re-investigation of the experimental ex-
+citation functions and νpre data of 210Po ruled out any
+significant shell influence on the saddle [20] after corre-
+lated tuning of statistical-model parameters and inclu-
+sion of fission delay.
+Another interesting aspect is the contradictory inter-
+pretation for correlation between neutron shell structure
+and nuclear dissipation strength that was required to re-
+produce the measured ER and νpre excitation functions
+in N=126 shell closed nuclei, namely 212Rn and 213Fr.
+The theoretical analysis of νpre data of 212Rn [21] and
+213Fr [22] reported a low dissipation strength at Eex ≈
+50 MeV which was attributed to the influence of neu-
+tron shell closure. On the contrary, no discernible shell
+influence was reported from ER cross-section studies of
+212Rn and its isotope [23], though moderate nuclear dis-
+sipation was required to describe the data. It must be
+noted that the magnitude of dissipation invoked to ex-
+plain the experimental ER cross-sections varied within
+Rn isotopes [23, 24], which is again found to be different
+for the description of the νpre data [21]. Interestingly,
+in case of Fr nuclei, the finite-range liquid drop model
+fission barrier was scaled down, particularly for 213Fr to
+fit measured ER cross-section [15]. This reduction of the
+arXiv:2301.13461v1 [nucl-th] 31 Jan 2023
+
+2
+fission barrier is in disagreement with the predictions for
+the shell closed nuclei [25]. One notable observation is the
+reported interpretation of reduced survival probability of
+213Fr nucleus due to neutron shell which is in contrast
+to the isotopic trend reported for Rn isotopes. Further,
+the fission cross-section of 213Fr is reported to exhibit
+no extra stability from N=126 shell closure [26]. In the
+statistical model approach followed in these works, no at-
+tempts were made to extract a global prescription of the
+parameters, rather, a case specific adjustment of dissipa-
+tion strength was involved. The influence of neutron shell
+structures on the potential energy surface and hence fis-
+sion observables are still quite ambiguous. Apart from
+just shell influence, entrance channels effects are also
+probed in a couple of recent publications to understand
+the experimental νpre data for 213Fr nuclei [27, 28]. These
+studies reportedly observed a deviation in the measured
+data from the predictions of entrance channel model for
+16O- and 19F-induced reactions.
+These studies substantiate the view that no consistent
+picture has emerged from recent independent analysis
+of each fission observable for neutron shell closed nuclei
+210Po, 212Rn and 213Fr and their isotopes.
+Inadequa-
+cies of standard statistical model interpretations have
+been addressed by employing Kramers-modified fission
+width taking into account shape-dependent level den-
+sity, temperature-dependent fission transition points, ori-
+entation (K state) degree of freedom and temperature-
+independent reduced dissipation coefficient [16, 17]. At-
+tempts for restraining the statistical-model parameters
+have also been reported [29], but a consistent description
+of experimental data for all three observables, namely
+νpre, fission and ER excitation functions for shell closed
+nuclei still could not be achieved. Recent developments
+in multi-dimensional stochastic approach are fairly suc-
+cessful in describing the fission characteristics of excited
+nuclei [30–35]. However, a simultaneous description of
+the experimental data and a systematic study for shell
+closed nuclei has not been attempted yet and is further
+required.
+In this paper, we show that the dynamical model based
+on 1D Langevin equation coupled with a statistical ap-
+proach [36] can simultaneously reproduce νpre, fission
+and ER cross-section data of shell closed nuclei over a
+range of excitation energies (Eex ≈ 40−80 MeV) of the
+measurements. The present calculations are performed
+without adjusting any of the model parameters, thus pro-
+vides a unified framework for a simultaneous study of
+these fission observables for nuclei in A ≈ 200 region.
+We re-investigated the available experimental data for
+the neutron shell closed nuclei 210Po, 212Rn and 213Fr,
+and their non-shell closed isotopes 206Po, 214,216Rn and
+215,217Fr.
+It is observed that a universal deformation-
+dependent reduced friction parameter is able to describe
+the fission observables simultaneously at all measured en-
+ergies irrespective of the shell structure of the nuclei.
+II.
+THEORETICAL MODEL DESCRIPTION
+A combined dynamical and statistical model code [37]
+is utilized to compute the fission observables of nuclei
+under study.
+The detailed description of the theoreti-
+cal aspects of the model can be found in Refs. [36, 38].
+The dynamical part of the model is carried out with a
+1D Langevin equation of motion governed by a driving
+potential that is determined by free energy F(q, T), as
+employed in recent Refs.
+[39–44].
+The free energy as
+derived from the Fermi gas model is related to the defor-
+mation dependent level density parameter a(q, A) as F(q,
+T) = V(q) - a(q, A)T 2 where T is the nuclear temper-
+ature, q is the dimensionless deformation coordinate de-
+fined as the ratio of half the distance between the center
+of masses of future fission fragments to the radius of CN
+and V (q) is the nuclear potential energy obtained from
+the finite-range liquid drop model [45, 46]. Fr¨obrich [47]
+and Lestone et al. [17] have emphasized on using nuclear
+entropy given by, S(q, A, Etot) = 2
+�
+a(q, A)[Etot − V (q)]
+in determining the driving force and therefore, it is em-
+ployed as a crucial quantity in the model. The nuclear
+driving force K = - dV (q)
+dq
++ da(q)
+dq T 2, not only consists of
+a conservative force but also contain a thermodynami-
+cal correction that enters the dynamics via. level density
+parameter a(q, A). The deformation dependent level den-
+sity parameter used in constructing the entropy has the
+form [48]:
+a(q, A) = ˜a1A + ˜a2A2/3Bs(q)
+(1)
+where A is the mass number of the CN and ˜a1 = 0.073
+MeV−1 and ˜a2 = 0.095 MeV−1 are taken from Ref. [49].
+Bs(q) is the dimensionless functional of the surface en-
+ergy [34, 38, 43, 50], expressed as the ratio of surface
+energy of the composite system to that of a sphere.
+The over-damped Langevin equation which describes
+the fission process in the dynamical part of the model
+thus, has the form [36]:
+dq
+dt =
+T
+Mβ(q)
+�∂S(q)
+∂q
+�
+Etot
++
+�
+T
+Mβ(q)Γ(t)
+(2)
+where Etot is the total energy of the composite system
+that remains conserved and Γ(t) is a Markovian stochas-
+tic variable with a normal distribution. The reduced dis-
+sipation coefficient β(q) = γ/M (as employed in litera-
+ture, see e.g., Refs. [16, 29, 42, 44] (and Refs. therein))
+is the ratio of friction coefficient γ to the inertia param-
+eter M calculated with Werner-Wheeler approximation
+of an incompressible irrotational fluid [51]. The present
+model employs ”funny−hills” parameters {c,h,α} [52]
+for describing the shape of the fissioning nuclei.
+Tak-
+ing into account only symmetric fission, the mass asym-
+metry parameter of the shape evolution is set to α=0
+[36, 38, 50]. The dimensionless fission coordinate (q) is
+given by q(c,h)= ( 3c
+8 )(1+ 2
+15[2h+ (c−1)
+2
+]c3), where c and h
+
+3
+defines the elongation and neck degree of freedom of the
+fissioning nucleus, respectively [36, 43, 53, 54].
+Following the fission dynamics through full Langevin
+dynamical calculation is quite time consuming. Similar
+to previous Langevin studies [31, 36, 39–43], a compu-
+tationally less intensive approach is adopted in present
+study where the dynamical stage is coupled with a sta-
+tistical model. In the present calculations, the emission
+of light particles from ground state to scission config-
+uration along the Langevin trajectories is treated as a
+discrete process.
+The evaporation of pre-scission light
+particles from ground state of Langevin trajectories to
+the scission point is coupled to the fission mode by a
+Monte Carlo procedure. The decay width for light parti-
+cle evaporation at each Langevin time step is calculated
+with the formalism as suggested by Fr¨obrich et al. [36]
+and later incorporated in Refs. [34, 40–43]. The emission
+width of a particle of kind ν (n,p,α) is given by [55]:
+Γν = (2sν + 1)
+mν
+π2ℏ2ρc(Eex)
+×
+� (Eex−Bν)
+0
+dϵνρR(Eex − Bν − ϵν)ϵνσinv(ϵν)
+(3)
+where sν is the spin of emitted particle ν, and mν is
+its reduced mass with respect to the residual nucleus.
+The level densities of the compound and residual nuclei
+are denoted by ρc(Eex) and ρR(Eex − Bν − ϵν). Bν is
+the liquid-drop binding energy, ϵ is the kinetic energy
+of the emitted particle and σinv(ϵν) is the inverse cross
+sections [55]. The decay width for light particle emission
+is calculated at each Langevin time step τ [43, 53, 54].
+When a stationary flux over the barrier is reached af-
+ter a sufficiently long delay time, the decay of the CN
+is then modelled by an adequately modified statistical
+model [38, 56, 57]. To have continuity when switching
+from dynamical to statistical branch, an entropy depen-
+dent fission width is incorporated in the latter. While en-
+tering the statistical branch, the particle emission width
+Γν is re-calculated and the fission width Γf = ℏRf [36]
+is calculated with fission rate (Rf) given by,
+Rf =
+Tgs
+�
+|S
+′′
+gs|S
+′′
+sd
+2πMβgs
+exp[S(qgs) − S(qsd)]
+× 2(1+erf[(qsc − qsd)
+�
+S
+′′
+sd/2])−1
+(4)
+Here erf(x) = (2/√π)
+� x
+0 dt exp(−t2) is the error func-
+tion and βgs is ground state dissipation coefficient. The
+saddle-point (qsd) and the ground-state positions (qgs)
+are defined by the entropy and not, as in the conventional
+approach, by the potential energy. The standard Monte
+Carlo cascade procedure was used to select the kind of
+decay with weights Γi/Γtot (i=fission,n,p,d,α) and Γtot =
+�
+i Γi. Pre-scission particle multiplicities are calculated
+by counting the number of corresponding evaporated par-
+ticle events registered in the dynamical and statistical
+branch of the model.
+The Langevin equation is started from a ground state
+configuration with a temperature corresponding to the
+initial excitation energy.
+The fusion cross-section can
+be determined from the partial cross section dσ(l)
+dl
+which
+represent the contribution of angular momenta l to the
+total fusion cross-section.
+Each Langevin trajectory is
+started with an orbital angular momentum which is sam-
+pled from a fusion spin distribution that reads as [34, 36]:
+dσ(l)
+dl
+= 2π
+k2
+2l + 1
+1 + exp (l−lc)
+δl
+(5)
+The final results are weighted over all relevant waves, that
+is, the spin distribution is used as an angular momen-
+tum weight function with which the Langevin calcula-
+tions for fission are started. As shown in recent Langevin
+studies, [34, 39–44], the spin distribution is calculated
+with the surface friction model [58].
+This calculation
+also fixes the fusion cross-section thus guaranteeing the
+correct normalization of fission and evaporation residue
+cross-sections within the accuracy of the surface friction
+model. The parameters lc and δl are the critical angular
+momentum for fusion and diffuseness, respectively.
+The fission observables that will be discussed in sub-
+sequent sections are calculated in the model as follows.
+The pre-scission neutron multiplicity is the number of
+neutrons emitted by the CN till it reaches the scission
+configuration. The fission probability (Pf) is given by
+the ratio of fissioned trajectories to total trajectories.
+The CN survival probability (1-Pf) is given by number
+of trajectories leading to ER formation divided by total
+trajectories and the fission (ER) cross-section is given by
+the product of fission (survival) probability and fusion
+cross-section.
+III.
+RESULTS AND DISCUSSION
+In the present study, pre-scission neutron multiplic-
+ities, fission and ER excitation functions for 206,210Po,
+212,214,216Rn and 213,215,217Fr compound nuclei are com-
+puted and compared with available experimental data
+wherein 210Po, 212Rn and 213Fr are N=126 neutron shell
+closed nuclei. The table I shows important parameters
+for the reactions studied in this work. The dynamical cal-
+culations are performed with a universal frictional form
+of Refs. [36, 47, 57] without adjusting any of the model
+parameters with a consistent prescription of the dissipa-
+tion coefficient. To account for sufficient statistics, 107
+Langevin trajectories are considered in the model calcu-
+lations.
+Fig.
+1 shows the results of dynamical calculations
+compared with the experimental data of νpre, fission
+and ER cross-sections for 206Po formed via.
+12C+194Pt
+[18, 19, 59] reaction and 210Po formed through two dif-
+ferent entrance channel reactions, namely
+12C+198Pt
+[18, 19, 60] and 18O+192Os [5, 60, 61], spanning a wide
+range of excitation energy. The excitation energies shown
+
+4
+TABLE I. Important parameters of reactions studied
+CN
+fissility
+Sn
+Bf(l=0)
+Reaction
+Mass excess (MeV) α/αBG
+(MeV)
+(MeV)
+target(proj)
+CN
+206Po
+0.717
+7.99
+10.51
+12C+194Pt
+-34.79(0)
+-18.83
+1.043
+210Po
+0.711
+7.38
+11.22
+12C+198Pt
+-29.93(0)
+-16.33
+1.050
+18O+192Os -35.89(-0.78) -16.33
+0.982
+212Rn
+0.732
+7.83
+8.88
+18O+194Pt -34.79(-0.78)
+-9.26
+0.970
+214Rn
+0.729
+7.54
+9.19
+16O+198Pt -29.93(-4.74)
+-4.77
+0.996
+216Rn
+0.727
+7.25
+9.49
+18O+198Pt -29.93(-0.78)
+0.70
+0.977
+213Fr
+0.743
+8.06
+7.83
+16O+197Au -31.16(-4.74)
+-4.01
+0.987
+19F+194Pt
+-34.79(-1.49)
+-4.01
+0.954
+215Fr
+0.740
+7.76
+8.13
+19F+196Pt -32.67 (-1.49) -0.07
+0.958
+217Fr
+0.737
+7.47
+8.42
+19F+198Pt
+-29.93(-1.49)
+5.00
+0.961
+here are with respect to the liquid drop ground state CN
+mass and experimental mass of projectile and target [62].
+Our calculations are restricted to excitation energies at
+and above 40 MeV where the present macroscopic model
+is valid. We emphasize that the microscopic shell correc-
+tions are not accounted for in the present calculations,
+as we are dealing with hot nuclei where shell effects are
+expected to be negligible at high excitation energies that
+are populated in heavy-ion reactions. The results of cal-
+culations using only the statistical model (dashed line)
+are also shown in Fig. 1. These calculations are made
+with the same code with Langevin dynamics turned off.
+The statistical model calculations under-predict the mea-
+sured νpre data as shown in panels (a) to (c), even more
+so as excitation energy increases. The dynamical model
+calculations using universal reduced friction coefficient
+are in excellent agreement with the measured data of
+νpre (panels (a) to (c)), fission cross-sections σfiss (pan-
+els (d) to (f)) and ER cross-sections σER (panels (g) to
+(i)) for the neutron shell closed nuclei 210Po as well as
+its isotope 206Po. The measured data of 210Po formed
+through two different entrance channels agree well with
+the theory in a broad range of excitation energies up to
+80 MeV. The model calculations describe the available
+experimental data for 206,210Po simultaneously at these
+excitation energies without any microscopic corrections
+included in the model. These observations are at vari-
+ance with the statistical model analysis of 12C+194Pt and
+12C+198Pt reactions that reported a significant shell cor-
+rection at the saddle deformation to describe the angular
+anisotropy and νpre data [18, 19]. A recent 4D Langevin
+dynamical study [63] that was carried on 206Po and 210Po
+populated from reaction 12C+198Pt, reported a reason-
+able description of the measured data for these reactions
+without invoking any extra shell corrections at the saddle
+state; shown as open triangles in panels (a), (c), (d) and
+(f) of Fig. 1. A better agreement of the measured data is
+observed for 12C+198Pt reaction in comparison to its 4D
+Langevin calculations [63], particularly at low excitation
+energies as shown in panels (a) and (d) of Fig. 1. The
+overestimation of νpre and fission cross-section of 210Po
+in Ref.
+[63] was attributed to the remnant of ground
+state shells and hence, a consequence of not using a pure
+macroscopic potential energy surface as suggested in Ref.
+[64]. Nonetheless, the predictions of multi-dimensional
+Langevin model for νpre data of 206Po by Karpov et al.
+[30] are also found to be in reasonable agreement with the
+results of the present analysis. Moreover, the measured
+mass distribution of fragments in the fission of 206,210Po
+[65, 66] reaffirms the absence of any shell corrections on
+the potential energy surface at the saddle point.
+Figs.
+2 and 3 display the comparison between ex-
+perimental data and theoretical calculations of νpre, fis-
+sion, ER and fusion cross-sections for N=126 shell closed
+nuclei viz.
+212Rn [21, 23, 24, 67] formed through re-
+action 18O+194Pt and 213Fr formed through reactions
+19F+194Pt [15, 22, 26, 68] and 16O+197Au [5, 6], and
+their non-shell closed isotopes 214,216Rn populated via.
+reactions 16,18O+198Pt [21, 23, 24, 67] and 215,217Fr pop-
+ulated via. reactions 19F+196,198Pt [15, 22, 26, 68]. The
+model calculations describe the νpre and fission excita-
+tion functions for 212Rn and its isotopes 214,216Rn quite
+successfully. In reactions forming 213,215,217Fr nuclei, the
+same parameter set is able to account for the experi-
+mental fission excitation functions but not νpre. A re-
+cent work [26] using an extended version of statistical-
+model employing collective enhancement of level density
+also reported an under-estimation of νpre data for same
+reactions when fitted simultaneously with fission cross-
+section. In the present work, the disagreement between
+experimental νpre and theory is prominent above ≈50
+MeV excitation energy and it increases with rise in exci-
+tation energy. Considering that νpre of other studied nu-
+clei are well reproduced by the model, it is unclear why
+
+5
+0
+1
+2
+3
+4
+νpre
+(a)
+12C+198Pt −→ 210Po
+(b)
+18O+192Os −→ 210Po
+(c)
+12C+194Pt −→ 206Po
+100
+101
+102
+103
+σfiss(mb)
+(d)
+(e)
+(f)
+40
+60
+80
+100
+100
+101
+102
+103
+σER(mb)
+(g)
+40
+60
+80
+100
+(h)
+40
+60
+(i)
+Eex (MeV)
+FIG. 1. (Colour online) Measured and calculated pre-scission neutron multiplicities (νpre), fission cross-sections (σfiss) and evap-
+oration residue cross-sections (σER) as a function of excitation energy for the reactions 12C+198Pt, 18O+192Os and 12C+194Pt.
+The continuous line (red) denote calculated results with a universal frictional form factor and dashed line (black) represent
+statistical model calculations. The symbols in the legend represent different experimental data sets, for νpre: (filled squares)
+Ref. [19], (filled circles) Ref. [5] and (open square) Ref. [59]; for σfission and σER: (filled diamonds) Ref. [18], (filled hexagons)
+Ref. [61] and (open diamonds) Ref. [60]. The open triangles represent results of νpre and σfission from 4D Langevin calculations
+of Ref. [63].
+the same frictional form fails, particularly for reactions
+forming Fr nuclei. It is to be noted that, an energy de-
+pendent dissipation was used in Ref.[21, 22] to describe
+the νpre data for these reactions.
+We also attempted
+similar approach by employing a temperature-dependent
+friction (TDF) in the stochastic calculations [69] (with-
+out changing any other parameter). This frictional form
+factor is deformation dependent, unlike the ones used in
+Refs. [21, 22, 70]. The maximum of β(q) in TDF corre-
+sponds to the ground state, that tends to decrease with
+increasing deformation with its minimum near the sad-
+dle configuration and is followed by an increase in the
+dissipation strength when approaching the scission. The
+dissipation coefficient assumes a higher value with in-
+creasing temperature of the CN. It is observed that a
+better agreement of νpre data is achieved for reactions
+19F+194,196,198Pt and 16O+197Au after invoking temper-
+ature dependence of the dissipation. The same frictional
+form, however, is found to over-predict the measured νpre
+data of other studied nuclei and hence is not shown here.
+Deviation in ER excitation functions are also to be
+noted for 212Rn and 213,215,217Fr nuclei wherein the cal-
+culated ER cross-sections underpredict the experimental
+data for these nuclei at high excitation energies. The case
+
+6
+0
+2
+4
+νpre
+(a)
+18O+198Pt −→ 216Rn
+(b)
+16O+198Pt −→ 214Rn
+(c)
+18O+194Pt −→ 212Rn
+101
+102
+103
+σfiss(mb)
+(d)
+(e)
+(f)
+101
+102
+103
+σER(mb)
+(g)
+(h)
+(i)
+40
+60
+80
+101
+102
+103
+σfus(mb)
+(j)
+40
+60
+80
+(k)
+40
+60
+80
+(l)
+Eex (MeV)
+FIG. 2. (Colour online) Measured and calculated pre-scission neutron multiplicities (νpre), fission cross-sections (σfiss), evapora-
+tion residue cross-sections (σER) and fusion cross-sections (σfus) as a function of excitation energy for the reactions 18O+198Pt,
+16O+198Pt, 18O+194Pt. The continuous (red) and dashed (black) lines have the same meaning as in Fig. 1. The calculations
+of fusion cross-section are independent of the frictional form and are represented by dotted line (brown). The symbols in the
+legend represent different experimental data sets, for νpre: (filled squares) Ref. [21]; for σfiss: (filled diamonds) Ref. [67] and
+(open diamonds) Ref. [23]; for σER: (filled circles) Ref. [24] and (filled hexagons) Ref. [23] and for σfus: (filled triangles) Refs.
+[23, 24].
+of Rn isotopes is of particular interest as the ER cross-
+section data for 214,216Rn [24] agrees fairly well with the
+model calculations at all measured energies but differ for
+212Rn [23] except at the lowest energy. For 213,215,217Fr
+nuclei, the measured ER cross-sections of Ref. [15] differ
+above excitation energy ≈ 55 MeV and the deviation is
+prominent for 213,215Fr. It is quite interesting to note that
+the ER measurement by a different group [68] for same
+reactions forming 213,217Fr at Eex ≤ 55 MeV follows the
+trend of the model predictions quite successfully. Unfor-
+tunately, Ref. [68] has reported only three data points.
+Moreover, the ER cross-section data of 215Fr formed in
+reaction 18O+197Au [71] is reproduced reasonably well
+with results of 19F+196Pt particularly, above 50 MeV ex-
+citation energy (displayed as open pentagons in panel (j)
+of Fig. 3). The present dynamical calculations assume
+
+7
+0
+2
+4
+6
+νpre
+(a)
+19F+198Pt → 217Fr
+(b)
+19F+196Pt → 215Fr
+(c)
+19F+194Pt → 213Fr
+(d)
+16O+197Au → 213Fr
+101
+102
+103
+σfiss(mb)
+(e)
+(f)
+(g)
+(h)
+101
+102
+103
+σER(mb)
+(i)
+(j)
+(k)
+(l)
+50
+75
+100
+101
+102
+103
+σfus(mb)
+(m)
+50
+75
+100
+(n)
+50
+75
+100
+(o)
+50
+100
+(p)
+Eex (MeV)
+FIG. 3. (Colour online) Measured and calculated pre-scission neutron multiplicities (νpre), fission cross-sections (σfiss), evapora-
+tion residue cross-sections (σER) and fusion cross-sections (σfus) as a function of excitation energy for the reactions 19F+198Pt,
+19F+196Pt, 19F+194Pt and 16O+197Au. The continuous (red), dashed (black) and dotted (brown) lines have the same meaning
+as in Figs. 1 and 2. The dash-dotted line (magenta) represent calculated results with temperature-dependent friction. The
+symbols in the legend represent different experimental data sets, for νpre: (filled squares) Ref. [22] and (partially filled squares)
+Ref. [5] ; for σfiss: (filled diamonds) Ref. [26],(partially filled diamonds) Ref. [6] and (open diamonds) Ref. [68]; for σER:
+(filled circles) Ref. [15], (partially filled circles) Ref. [6] and (open circles) Ref. [68] and for σfus: (filled triangles) Refs.
+[15, 26, 68] and (open triangles) Refs. [6]. The open pentagons denote σER for 215Fr nuclei formed via 18O+197Au Ref. [71].
+decay from an equilibrated CN and any entrance channel
+effects are not included. It takes account of only the dif-
+ferent angular momenta that are populated in different
+entrance channels. Taking into consideration the insignif-
+icant difference in angular momenta between two en-
+trance channels forming 215Fr, the observed deviation in
+ER cross-section for 19F-induced reaction is quite unex-
+pected. These observations further necessitated the need
+to confront the deviations in describing ER cross-sections
+by comparing the measured fusion cross-sections for Rn
+and Fr nuclei with the model. It is revealed that the cal-
+culated fusion cross-sections are in good agreement with
+the measured fusion data, augmenting the validity of the
+present calculations. Furthermore, the under-prediction
+of ER cross-sections indicates the need for a strong dis-
+sipation in the pre-saddle region [72].
+However, 3D
+
+8
+Langevin dynamical calculations [31] reported a reduc-
+tion in the wall friction coefficient to reproduce the mass
+and kinetic energy distribution of fission fragments, and
+their influence on νpre for 215Fr nucleus. The strength
+of the reduction coefficient, ks = 0.25 − 0.5 indicates
+a weak dissipation in the initial stages of the fissioning
+nucleus. The experimental analysis of fission fragment
+nuclear-charge distributions and fission cross-sections of
+Fr, Rn isotopes and their neighbouring nuclei also re-
+ported a pre-saddle dissipation strength of magnitude
+(4.5 ± 0.5) × 1021 s−1 [73] and 2 × 1021 s−1 [74], respec-
+tively. The more recent microscopic study of energy de-
+pendent dissipation using time-dependent Hartree-Fock
++ BCS method [75] also observed a strength of deforma-
+tion dependent friction coefficient, ranging from 1 to 6
+× 1021 s−1 in heavy nuclei. The strength of these fric-
+tional parameterizations are quite in agreement with the
+dissipation form factor employed in the present calcula-
+tions.
+These observations affirm a weak dissipation in
+the pre-saddle region; so, the observed enhancement of
+ER cross-sections in Fr nuclei populated via. 19F-induced
+reactions is not well-understood from the perspective of
+dissipation strength alone.
+In fact, a satisfactory de-
+scription of the excitation functions including ER cross-
+sections for reactions 12C+194Pt, 12C+198Pt, 18O+192Os
+and 16,18O+198Pt and survival probabilities for a range
+of fissilities [36] is observed within the framework of this
+1D Langevin dynamics with a universal friction param-
+eter. However, it is also important to bear in mind the
+possible bias coming from experimental uncertainty. It is
+striking that the observed deviations are pronounced in
+ER cross-section data where measurements are reported
+to have large uncertainty in ER separator transmission
+efficiency [15, 23]. It would be highly desirable to have
+additional ER measurements to rule out any possible ex-
+perimental bias in the interpretation of ER data.
+It must be noted that, the entrance channel dynam-
+ics of the fusion stage might also play a role influenc-
+ing neutron emission at the formation stage [14]. It is
+known that interplay of CN excitation energy, angular
+momentum and fission barrier play crucial role in fission
+process [28]. Present study do not take into account any
+entrance channel dynamics influencing the fusion stage.
+The model only considers the entrance channel depen-
+dent ’l’ distribution calculated within the surface friction
+model [58]. In Fig. 4 we show the calculated fission bar-
+rier height Bf(l) for three compound systems and mean
+angular momentum < l > calculated from ’l’ distribu-
+tion for different entrance channels forming same CN.
+The variation of Bf is plotted as a function of ’l’ in Fig.
+4(a) and variation of < l > of the compound systems is
+plotted as a function of Eex in Fig 4(b). From Fig. 4,
+it is clear that, the difference in angular momenta be-
+tween two entrance channels forming same CN at similar
+Eex is not very significant to cause any ’l’ induced ef-
+fects on measured fission observable. This is evident in
+the νpre data for 210Po formed in reactions 12C+198Pt
+and 18O+192Os which are well described in the present
+work (see Fig. 1) without invoking any entrance channel
+effects in the model.
+Recent studies investigating en-
+trance channel dynamics [27, 28] reported disagreement
+between experimental νpre and predictions of entrance
+channel model for 213Fr nuclei formed via.
+16O+197Au
+and 19F+194Pt reactions. These studies were, however,
+not extended to other isotopes of Fr, namely 215,217Fr
+that also show similar discrepancy as reported in the
+present study.
+The current 1D Langevin analysis provides a simul-
+taneous description of the experimental data for neutron
+magic nuclei 210Po without invoking any saddle shell cor-
+rections or a nuclear dissipation strength dependent on
+system/observable under study. In order to understand
+qualitatively that consideration of saddle shell correc-
+tions are not required to explain νpre data, we consider
+the nature of neutron emission during the fission process.
+It is to be noted that these neutrons are emitted from dy-
+namical trajectories that originated from compact config-
+uration till scission point is reached. The prompt and
+beta-delayed neutron emissions from fission fragments
+are not taken into consideration.
+As recent publica-
+tions have advocated for the inclusion of shell correc-
+tions on the saddle configuration to describe the angular
+anisotropy and νpre data at moderate excitation energies
+[12, 18, 19], we have attempted to find the distribution
+of pre-scission neutrons as it evolves from ground state
+to scission point. The model calculated potential energy
+V(q) and distribution of percentage yield of pre-scission
+neutrons are plotted as a function of the deformation co-
+ordinate (q) for these nuclei at 50 MeV excitation energy
+and shown in Fig. 5. It is evident that more than 90% of
+the neutron emission occurs at an early stage of fission
+before the saddle deformation (q ≈ 0.8) [38] is reached.
+The mean of the distribution corresponds to νpre emis-
+sion close to the ground state configuration. In-fact, a
+multi-dimensional Langevin study of 215Fr by Nadtochy
+et al. [31] have also pointed out that an appreciable part
+of pre-scission neutrons are emitted at an early stage of
+fission before saddle is reached. As most of the neutrons
+are emitted close to the ground state configuration, it is
+unlikely to be influenced by any shell corrections applied
+at the saddle.
+Though the present code uses classical 1D approach to
+describe fission observables, the main objective of this
+work is to have a simultaneous description of experi-
+mental data without any parameter adjustment thus,
+removing some of the reported ambiguities.
+A com-
+parison between νpre calculated with 1D model and re-
+cent macroscopic multi-dimensional models is displayed
+in Fig. 6. It can be seen that the νpre values predicted
+by different models are very similar and also reproduce
+the measurements quite well for reactions spanning a
+wide range of fissility parameter Z2/A.
+Additionally,
+the multi-dimensional calculations [34, 50, 76] also use
+the formalisms adopted from Refs. [36, 69] such as the
+parameterization of surface friction model and weakest
+coordinate dependence of the level-density parameter as
+
+9
+0
+20
+40
+60
+80
+ℓ (¯h)
+0
+2
+4
+6
+8
+10
+12
+14
+Bf (MeV)
+(a)
+210Po
+212Rn
+213Fr
+30
+40
+50
+60
+70
+80
+90
+10
+15
+20
+25
+30
+35
+40
+45
+< ℓ > (¯h)
+(b)
+12C+198Pt
+18O+192Os
+19F+194Pt
+16O+197Au
+Eex (MeV)
+FIG. 4. (Colour online) (a) The angular momentum ’l’ de-
+pendent fission barrier height Bf(l) for three CN 210Po,212Rn
+and 213Fr and (b) Variation of mean angular momentum < l >
+with compound nucleus excitation energy for 210Po,and 213Fr
+populated by different entrance channels.
+employed in the present work.
+Hence, the qualitative
+nature of the observed features presented here is not ex-
+pected to be different with multi-dimensional approach.
+As the present framework is found to provide realistic
+values close to measured data, we believe that the 1D
+approach still can be a potential tool to study a wider
+systematics which can be accomplished within minimum
+0
+10
+20
+30
+40
+V (MeV)
+0.2
+0.4
+0.6
+0.8
+1.0
+1.2
+0
+1
+2
+3
+4
+5
+6
+7
+8
+d<νpre>/dq (%)
+qneck
+qsadd
+210Po
+212Rn
+213Fr
+deformation coordinate (q)
+FIG. 5.
+(Colour online) Potential energy distribution as a
+function of nuclear deformation coordinate (q) for three fis-
+sioning nuclei 210Po, 212Rn and 213Fr (top panel) and distri-
+bution of percentage yield of evaporated pre-scission neutrons
+as a function of (q) for three CN at 50 MeV excitation energy
+(bottom panel). The deformation coordinate (q) assumes a
+value of 0.6 (qneck) when the neck of the fissioning nucleus
+starts to develop and q=0.8 (qsadd) at the saddle state con-
+figuration.
+computational resources.
+It must be remarked here that, even though present
+analysis provides a reasonable reproduction of the exper-
+imental data without invoking any shell corrections at
+high excitation energies, it shall not be concluded from
+this work that shell effects are not relevant in the analy-
+sis. As present investigation consider only the first chance
+fission at Eex ∼ 40 MeV and above where shell effects are
+expected to be washed out, no indication for the need of
+including shell corrections was found. However, for the
+case when the CN is populated at low excitation energies
+or reaches low excitation energy due to neutron emission
+as a consequence of competition between neutron evapo-
+ration and fission (multi-chance fission), the microscopic
+effects are required to be taken into consideration. Re-
+cent microscopic study of dissipation within Hartree-Fock
++ BCS framework [75] have shown a strong dependence
+of dissipation on deformation and initial excitation ener-
+gies of the hot nuclei. Possible influence of microscopic
+temperature dependence of fission barrier height and its
+curvature were also emphasized in some recent studies
+of fully microscopic description of fission process [77, 78].
+
+10
+30
+35
+40
+Z2/A
+0
+1
+2
+3
+4
+5
+6
+νpre
+162Yb
+206Po
+210Po
+215Fr
+244Cm
+264Rf
+216Ra
+248Cf
+Expt. data
+Present
+Multi-dimensional model
+FIG. 6.
+(Colour online) Comparison of measured pre-
+scission neutron multiplicities (νpre) with the results of the
+1D model (present work) and multi-dimensional models. The
+filled triangles (blue) denote experimental data [6, 14, 59, 80–
+83], the present dynamical model calculations are represented
+by filled circles (orange) and the filled squares (green) de-
+note the results of multi-dimensional dynamical calculations
+[28, 30, 34, 84].
+A microscopic framework based on the finite-temperature
+Skyrme-HartreeFock+BCS approach [79] was adopted to
+demonstrate the essential role of energy dependent fission
+barriers by studying the experimental fission probability
+of 210Po. It would be quite interesting to extend the in-
+vestigation of Fr nuclei within such a microscopic frame-
+work.
+IV.
+SUMMARY AND CONCLUSION
+In the present work we report a systematic study on
+the fission dynamics of N=126 shell closed nuclei in mass
+region 200 with a simultaneous description of three fis-
+sion observables. The present work highlights the limited
+reliability of the conclusions drawn from the recent statis-
+tical model analysis of shell closed nuclei, namely 210Po,
+212Rn and 213Fr at excitation energies 40 MeV and above,
+that advocated for extra shell effects at saddle configu-
+ration even after their inclusion in the level density for-
+mulation. Earlier analyses of νpre and ER cross-sections
+were based on different assumptions and case dependent
+parameter adjustments, without reaching a definite con-
+clusion.
+On the basis of present analysis we conclude
+that, without many of those assumptions and parameter
+adjustments, a well established combined dynamical and
+statistical model can simultaneously reproduce the avail-
+able data of νpre, fission and evaporation residue excita-
+tion functions (also fusion cross-sections in certain cases)
+for neutron shell closed nuclei, viz.
+210Po, 212Rn and
+their non-shell closed isotopes 206Po and 214,216Rn with-
+out the need of including any extra shell effects. There
+appears to be no discernible influence of N=126 neutron
+shell structure on these measured fission observables in
+the medium excitation energy range. The present work
+also points to a relatively smaller role of entrance channel
+effects in the studied systems.
+However, we find a significant mismatch between mea-
+sured νpre data and its model predictions for Fr nuclei
+formed in reactions 19F+194,196,198Pt and 16O+197Au,
+despite a reasonable description of fission and fusion
+cross-sections.
+The νpre data in Fr nuclei could only
+be reproduced after invoking a temperature dependent
+frictional form.
+The difficulty in completely reproduc-
+ing some specific measurements of Fr nuclei still remains
+not well-understood and additional measurements are de-
+sired. Although the present work is limited to the study
+of three fission observables, it would also be interesting
+to extend the systematic study using recent microscopic
+theory within Hartree-Fock + BCS framework.
+V.
+ACKNOWLEDGMENTS
+We are thankful to K. S. Golda and N. Saneesh for
+fruitful discussions. One of the authors (D.A.) acknowl-
+edges the financial support in the form of research fel-
+lowship received from the University Grants Commission
+(UGC).
+REFERENCES
+[1] R. Vandenbosch, J. R. Huizenga, Nuclear Fission, Aca-
+demic Press, New York, 1973.
+[2] H. J. Krappe, K. Pomorski, Theory of Nuclear Fission,
+Lecture Notes in Physics, Springer, Heidelberg, 2012.
+[3] N. Bohr, J. A. Wheeler, Phys. Rev. 56 (1939) 426. doi:
+10.1103/PhysRev.56.426.
+[4] A. Gavron, J. R. Beene, B. Cheynis, R. L. Ferguson,
+F. E. Obenshain, F. Plasil, G. R. Young, G. A. Petitt,
+M. J¨a¨askel¨ainen, D. G. Sarantites, C. F. Maguire, Phys.
+Rev. Lett. 47 (1981) 1255. doi:10.1103/PhysRevLett.
+47.1255.
+[5] J. O. Newton,
+D. J. Hinde,
+R. J. Charity,
+J. R.
+Leigh, J. J. M. Bokhorst, A. Chatterjee, G. S. Foote,
+S. Ogaza, Nucl. Phys. A 483 (1988) 126. doi:10.1016/
+0375-9474(88)90068-1.
+
+11
+[6] D. J. Hinde, R. J. Charity, G. S. Foote, J. R. Leigh,
+J. O. Newton, S. Ogaza, A. Chatterjee, Nucl. Phys. A
+452 (1986) 550. doi:10.1016/0375-9474(86)90214-9.
+[7] J. P. Lestone, J. R. Leigh, J. O. Newton, D. J. Hinde,
+J. X. Wei, J. X. Chen, S. Elfstrom, D. G. Popescu, Phys.
+Rev. Lett. 67 (1991) 1078. doi:10.1103/PhysRevLett.
+67.1078.
+[8] H. Rossner, D. Hilscher, D. J. Hinde, B. Gebauer,
+M. Lehmann, M. Wilpert, E. Mordhorst, Phys. Rev. C
+40 (1989) 2629. doi:10.1103/PhysRevC.40.2629.
+[9] M. Thoennessen, D. R. Chakrabarty, M. G. Herman,
+R. Butsch, P. Paul, Phys. Rev. Lett. 59 (1987) 2860.
+doi:10.1103/PhysRevLett.59.2860.
+[10] D. J. Hofman, B. B. Back, P. Paul, Phys. Rev. C 51
+(1995) 2597. doi:10.1103/PhysRevC.51.2597.
+[11] H. A. Kramers, Physica 7 (1940) 284.
+doi:10.1016/
+S0031-8914(40)90098-2.
+[12] K. Mahata, S. Kailas, S. S. Kapoor, Phys. Rev. C 74
+(2006) 041301(R). doi:10.1103/PhysRevC.74.041301.
+[13] D. Mancusi, R. J. Charity, J. Cugnon, Phys. Rev. C 82
+(2010) 044610. doi:10.1103/PhysRevC.82.044610.
+[14] A. Saxena, A. Chatterjee, R. K. Choudhury, S. S.
+Kapoor, D. M. Nadkarni, Phys. Rev. C 49 (1994) 932.
+doi:10.1103/PhysRevC.49.932.
+[15] V. Singh, B. R. Behera, M. Kaur, A. Kumar, K. P. Singh,
+N. Madhavan, S. Nath, J. Gehlot, G. Mohanto, A. Jhin-
+gan, I. Mukul, T. Varughese, J. Sadhukhan, S. Pal,
+S. Goyal, A. Saxena, S. Santra, S. Kailas, Phys. Rev.
+C 89 (2014) 024609. doi:10.1103/PhysRevC.89.024609.
+[16] S. G. McCalla, J. P. Lestone, Phys. Rev. Lett. 101 (2008)
+032702. doi:10.1103/PhysRevLett.101.032702.
+[17] J. P. Lestone, S. G. McCalla, Phys. Rev. C 79 (2009)
+044611. doi:10.1103/PhysRevC.79.044611.
+[18] A. Shrivastava, S. Kailas, A. Chatterjee, A. M. Samant,
+A. Navin, P. Singh, B. S. Tomar, Phys. Rev. Lett. 82
+(1999) 699. doi:10.1103/PhysRevLett.82.699.
+[19] K. S. Golda, A. Saxena, V. K. Mittal, K. Mahata,
+P. Sugathan, A. Jhingan, V. Singh, R. Sandal, S. Goyal,
+J. Gehlot, A. Dhal, B. R. Behera, R. K. Bhowmik,
+S. Kailas, Nucl. Phys. A 913 (2013) 157. doi:10.1016/
+j.nuclphysa.2013.05.016.
+[20] K. Mahata, S. Kailas, S. S. Kapoor, Phys. Rev. C 92
+(2015) 034602. doi:10.1103/PhysRevC.92.034602.
+[21] R. Sandal, B. R. Behera, V. Singh, M. Kaur, A. Kumar,
+G. Singh, K. P. Singh, P. Sugathan, A. Jhingan, K. S.
+Golda, M. B. Chatterjee, R. K. Bhowmik, S. Kalkal,
+D. Siwal, S. Goyal, S. Mandal, E. Prasad, K. Mahata,
+A. Saxena, J. Sadhukhan, S. Pal, Phys. Rev. C 87 (2013)
+014604. doi:10.1103/PhysRevC.87.014604.
+[22] V. Singh, B. R. Behera, M. Kaur, P. Sugathan, K. S.
+Golda, A. Jhingan, J. Sadhukhan, D. Siwal, S. Goyal,
+S. Santra, A. Kumar, R. K. Bhowmik, M. B. Chatterjee,
+A. Saxena, S. Pal, S. Kailas, Phys. Rev. C 86 (2012)
+014609. doi:10.1103/PhysRevC.86.014609.
+[23] P. V. Laveen, E. Prasad, N. Madhavan, S. Pal, J. Sad-
+hukhan, S. Nath, J. Gehlot, A. Jhingan, K. M. Varier,
+R. G. Thomas,
+A. M. Vinodkumar,
+A. Shamlath,
+T. Varughese, P. Sugathan, B. R. S. Babu, S. Ap-
+pannababu, K. S. Golda, B. R. Behera, V. Singh, R. San-
+dal, A. Saxena, B. V. John, S. Kailas, J. Phys. G: Nucl.
+Part. Phys. 42 (2015) 095105. doi:10.1088/0954-3899/
+42/9/095105.
+[24] R. Sandal, B. R. Behera, V. Singh, M. Kaur, A. Kumar,
+G. Kaur, P. Sharma, N. Madhavan, S. Nath, J. Gehlot,
+A. Jhingan, K. S. Golda, H. Singh, S. Mandal, S. Verma,
+E. Prasad, K. M. Varier, A. M. Vinodkumar, A. Saxena,
+J. Sadhukhan, S. Pal, Phys. Rev. C 91 (2015) 044621.
+doi:10.1103/PhysRevC.91.044621.
+[25] P. M¨oller,
+A. J. Sierk,
+T. Ichikawa,
+A. Iwamoto,
+R. Bengtsson, H. Uhrenholt, S. ˚Aberg, Phys. Rev. C 79
+(2009) 064304. doi:10.1103/PhysRevC.79.064304.
+[26] V. Singh, B. R. Behera, M. Kaur, A. Jhingan, R. Kaur,
+P. Sugathan, D. Siwal, S. Goyal, K. P. Singh, S. Pal,
+A. Saxena, S. Kailas, J. Phys. G: Nucl. Part. Phys. 48
+(2021) 075104. doi:10.1088/1361-6471/abe8cd.
+[27] M. Shareef, A. Chatterjee, E. Prasad, Eur. Phys. J. A 52
+(2016) 342. doi:10.1140/epja/i2016-16342-4.
+[28] C. Schmitt, K. Mazurek, P. N. Nadtochy, Phys. Rev. C
+97 (2018) 014616. doi:10.1103/PhysRevC.97.014616.
+[29] T. Banerjee, S. Nath, S. Pal, Phys. Lett. B 776 (2018)
+163. doi:10.1016/j.physletb.2017.11.040.
+[30] A. V. Karpov, P. N. Nadtochy, D. V. Vanin, G. D. Adeev,
+Phys. Rev. C 63 (2001) 054610. doi:10.1103/PhysRevC.
+63.054610.
+[31] P. N. Nadtochy, G. D. Adeev, A. V. Karpov, Phys. Rev.
+C 65 (2002) 064615. doi:10.1103/PhysRevC.65.064615.
+[32] J. Randrup, P. M¨oller, A. J. Sierk, Phys. Rev. C 84 (2011)
+034613. doi:10.1103/PhysRevC.84.034613.
+[33] J. Randrup, P. M¨oller, Phys. Rev. Lett. 106 (2011)
+132503. doi:10.1103/PhysRevLett.106.132503.
+[34] P. N. Nadtochy, E. G. Ryabov, A. E. Gegechkori, Y. A.
+Anischenko, G. D. Adeev, Phys. Rev. C 89 (2014) 014616.
+doi:10.1103/PhysRevC.89.014616.
+[35] K. Mazurek, P. N. Nadtochy, E. G. Ryabov, G. D.
+Adeev, Eur. Phys. J. A 53 (2017) 79. doi:10.1140/epja/
+i2017-12262-1.
+[36] P. Fr¨obrich, I. I. Gontchar, Phys. Reports 292 (1998) 131.
+doi:10.1016/S0370-1573(97)00042-2.
+[37] I. I. Gontchar, L. A. Litnevsky, P. Fr¨obrich, Computer
+Physics Communications 107 (1997) 223. doi:10.1016/
+S0010-4655(97)00108-2.
+[38] I. I. Gontchar, P. Fr¨obrich, N. I. Pischasov, Phys. Rev.
+C 47 (1993) 2228. doi:10.1103/PhysRevC.47.2228.
+[39] G. Chaudhuri, S. Pal, Eur. Phys. J. A 18 (2003) 9. doi:
+10.1140/epja/i2003-10038-x.
+[40] W. Kun, M. Yu-Gang, W. Yi-Bin, C. Xiang-Zhou, C. Jin-
+Gen, F. De-Qing, G. Wei, M. Guo-Liang, S. Wen-Qing,
+T. Wen-Dong, Z. Chen, Z. Xing-Fei, Chinese Phys. Lett.
+22 (2005) 53. doi:10.1088/0256-307x/22/1/016.
+[41] Y. Wei, W. Feng, Y. Hong-Wei, Chinese Phys. C 32
+(2008) 816. doi:10.1088/1674-1137/32/10/010.
+[42] W. Feng, Y. Wei, Chinese Phys. C 34 (2010) 551. doi:
+10.1088/1674-1137/34/5/007.
+[43] N. Wang, W. Ye, Phys. Rev. C 97 (2018) 014603. doi:
+10.1103/PhysRevC.97.014603.
+[44] N. Wang, W. Ye, Phys. Rev. C 103 (2021) 024611. doi:
+10.1103/PhysRevC.103.024611.
+[45] W. D. Myers, W. J. Swiatecki, Nucl. Phys. 81 (1966) 1.
+doi:10.1016/0029-5582(66)90639-0.
+[46] H. J. Krappe, J. R. Nix, A. J. Sierk, Phys. Rev. C 20
+(1979) 992. doi:10.1103/physrevc.20.992.
+[47] P. Fr¨obrich, Nucl. Phys. A 787 (2007) 170. doi:10.1016/
+j.nuclphysa.2006.12.028.
+[48] R. Balian, C. Bloch, Annals of Phys. 60 (1970) 401. doi:
+10.1016/0003-4916(70)90497-5.
+[49] A. V. Ignatyuk, M. G. Itkis, V. N. Okolovich, G. N.
+Smirenkin, A. S. Tishin, Yad. Fiz. 21 (1975) 1185.
+
+12
+[50] G. Chaudhuri, S. Pal, Phys. Rev. C 65 (2002) 054612.
+doi:10.1103/physrevc.65.054612.
+[51] K. T. R. Davies, A. J. Sierk, J. R. Nix, Phys. Rev. C 13
+(1976) 2385. doi:10.1103/physrevc.13.2385.
+[52] M. Brack, J. E. N. S. Damgaard, A. S. Jensen, H. C.
+Pauli, V. M. Strutinsky, C. Y. Wong, Rev. Mod. Phys.
+44 (1972) 320. doi:10.1103/revmodphys.44.320.
+[53] W. Ye, J. Tian, Phys, Rev. C 93 (2016) 044603. doi:
+10.1103/physrevc.93.044603.
+[54] W. Ye, J. Tian, Phys. Rev. C 91 (2015) 064603. doi:
+10.1103/physrevc.91.064603.
+[55] M. Blann, Phys. Rev. C 21 (1980) 1770. doi:10.1103/
+PhysRevC.21.1770.
+[56] N. D. Mavlitov, P. Fr¨obrich, I. I. Gonchar, Z Physik A
+342 (1992) 195. doi:10.1007/bf01288469.
+[57] P. Fr¨obrich, I. I. Gontchar, N. D. Mavlitov, Nucl. Phys.
+A 556 (1993) 281. doi:10.1016/0375-9474(93)90352-X.
+[58] J. Marten, P. Fr¨obrich, Nucl. Phys. A 545 (1992) 854.
+doi:10.1016/0375-9474(92)90533-p.
+[59] G. Chubaryan, M. Itkis, S. Luk’yanov, V. Okolovich,
+Y.
+Penionzhkevich,
+A.
+Rusanov,
+V.
+Salamatin,
+G. Smirenkin, Phys. Atom. Nucl. 56 (1993) 286.
+[60] J. van der Plicht, H. C. Britt, M. M. Fowler, Z. Fraenkel,
+A. Gavron, J. B. Wilhelmy, F. Plasil, T. C. Awes, G. R.
+Young, Phys. Rev. C 28 (1983) 2022.
+doi:10.1103/
+PhysRevC.28.2022.
+[61] R. J. Charity, J. R. Leigh, J. J. M. Bokhorst, A. Chat-
+terjee, G. S. Foote, D. J. Hinde, J. O. Newton, S. Ogaza,
+D. Ward, Nucl. Phys. A 457 (1986) 441. doi:10.1016/
+0375-9474(86)90388-X.
+[62] P. Moller, J. R. Nix, W. D. Myers, W. J. Swiatecki, At.
+data and Nucl. data tables 59 (1995) 185–381. doi:10.
+1006/adnd.1995.1002.
+[63] C. Schmitt, K. Mazurek, P. N. Nadtochy, Phys. Lett. B
+737 (2014) 289. doi:10.1016/j.physletb.2014.08.072.
+[64] K. Mahata, S. Kailas, Phys. Rev. C 95 (2017) 054616.
+doi:10.1103/PhysRevC.95.054616.
+[65] A. Sen, T. K. Ghosh, S. Bhattacharya, K. Banerjee,
+C. Bhattacharya, S. Kundu, G. Mukherjee, A. Asgar,
+A. Dey, A. Dhal, M. M. Shaikh, J. K. Meena, S. Manna,
+R. Pandey, T. K. Rana, P. Roy, T. Roy, V. Srivas-
+tava, P. Bhattacharya, Phys. Rev. C 96 (2017) 064609.
+doi:10.1103/PhysRevC.96.064609.
+[66] A. Chaudhuri, T. K. Ghosh, K. Banerjee, S. Bhat-
+tacharya, J. Sadhukhan, S. Kundu, C. Bhattacharya,
+J. K. Meena, G. Mukherjee, A. K. Saha, M. A. As-
+gar, A. Dey, S. Manna, R. Pandey, T. K. Rana, P. Roy,
+T. Roy, V. Srivastava, P. Bhattacharya, D. C. Biswas,
+B. N. Joshi, K. Mahata, A. Shrivastava, R. P. Vind,
+S. Pal, B. R. Behera, V. Singh, Phys. Rev. C 92 (2015)
+041601(R). doi:10.1103/PhysRevC.92.041601.
+[67] R. Sandal, Ph.D. thesis, Panjab University (2013).
+[68] K. Mahata, S. Kailas, A. Shrivastava, A. Chatterjee,
+P. Singh, S. Santra, B. S. Tomar, Phys. Rev. C 65 (2002)
+034613. doi:10.1103/PhysRevC.65.034613.
+[69] I. Gontchar, L. A. Litnevsky, Z Physik A 359 (1997) 149.
+doi:10.1007/s002180050382.
+[70] M. Thoennessen, G. F. Bertsch, Phys. Rev. Lett. 71
+(1993) 4303. doi:10.1103/PhysRevLett.71.4303.
+[71] L. Corradi,
+B. R. Behera,
+E. Fioretto,
+A. Gadea,
+A. Latina, A. M. Stefanini, S. Szilner, M. Trotta,
+Y. Wu, S. Beghini, G. Montagnoli, F. Scarlassara, R. N.
+Sagaidak, S. N. Atutov, B. Mai, G. Stancari, L. Tomas-
+setti, E. Mariotti, A. Khanbekyan, S. Veronesi, Phys.
+Rev. C 71 (2005) 014609.
+doi:10.1103/PhysRevC.71.
+014609.
+[72] B. B. Back, D. J. Blumenthal, C. N. Davids, D. J. Hen-
+derson, R. Hermann, D. J. Hofman, C. L. Jiang, H. T.
+Penttil¨a, A. H. Wuosmaa, Phys. Rev. C 60 (1999) 044602.
+doi:10.1103/PhysRevC.60.044602.
+[73] C. Schmitt, P. N. Nadtochy, A. Heinz, B. Jurado,
+A. Keli´c, K.-H. Schmidt, Phys. Rev. Lett. 99 (2007)
+042701. doi:10.1103/PhysRevLett.99.042701.
+[74] B. Jurado, C. Schmitt, K.-H. Schmidt, J. Benlliure,
+T. Enqvist, A. R. Junghans, A. Keli´c, F. Rejmund, Phys.
+Rev. Lett. 93 (2004) 072501. doi:10.1103/PhysRevLett.
+93.072501.
+[75] Y. Qiang, J. C. Pei, Phys. Rev. C 104 (2021) 054604.
+doi:10.1103/physrevc.104.054604.
+[76] P. N. Nadtochy, E. G. Ryabov, A. E. Gegechkori, Y. A.
+Anischenko, G. D. Adeev, Phys. Rev. C 85 (2012) 064619.
+doi:10.1103/PhysRevC.85.064619.
+[77] Y. Zhu, J. C. Pei, Phys. Rev. C 94 (2016) 024329. doi:
+10.1103/physrevc.94.024329.
+[78] C. Y. Qiao, J. C. Pei, Phys. Rev. C 106 (2022) 014608.
+doi:10.1103/physrevc.106.014608.
+[79] A. L. Goodman, Nucl. Phys. A 352 (1981) 30. doi:10.
+1016/0375-9474(81)90557-1.
+[80] D. J. Hinde, D. Hilscher, H. Rossner, B. Gebauer,
+M. Lehmann, M. Wilpert, Physical Review C 45 (1992)
+1229. doi:10.1103/physrevc.45.1229.
+[81] H. Singh, K. S. Golda, S. Pal, Ranjeet, R. Sandal, B. R.
+Behera, G. Singh, A. Jhingan, R. P. Singh, P. Sugathan,
+M. B. Chatterjee, S. K. Datta, A. Kumar, G. Viesti,
+I. M. Govil, Physical Review C 78 (2008) 024609. doi:
+10.1103/physrevc.78.024609.
+[82] M. G. Itkis, S. M. Lukyanov, V. N. Okolovich, Y. E.
+Penionzkevich, A. Y. Rusanov, V. S. Salamatin, G. N.
+Smirenkin, G. G. Chubaryan, Phys. At. Nucl. 52 (1990)
+15.
+[83] R. L. Ferguson, F. Plasil, H. Freiesleben, C. E. Bemis,
+H. W. Schmitt, Phys. Rev. C 8 (1973) 1104. doi:doi:
+10.1103/physrevc.8.1104.
+[84] H. Eslamizadeh, Eur. Phys. J. A 47 (2011) 134. doi:
+10.1140/epja/i2011-11134-0.
+
diff --git a/DtFRT4oBgHgl3EQfATdu/content/tmp_files/load_file.txt b/DtFRT4oBgHgl3EQfATdu/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2b2d95770e1a989e676e0916f6c4ef48d37cb4fa
--- /dev/null
+++ b/DtFRT4oBgHgl3EQfATdu/content/tmp_files/load_file.txt
@@ -0,0 +1,1600 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf,len=1599
+page_content='Investigating fission dynamics of neutron shell closed nuclei 210Po, 212Rn and 213Fr within a stochastic dynamical approach Divya Arora, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan,∗ and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee Inter-University Accelerator Centre, Aruna Asaf Ali Marg, New Delhi 110067, India Dissipative dynamics of nuclear fission is a well confirmed phenomenon described either by a Kramers-modified statistical model or by a dynamical model employing the Langevin equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Though dynamical models as well as statistical models incorporating fission delay are found to explain the measured fission observables in many studies, it nonetheless shows conflicting results for shell closed nuclei in the mass region 200.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Analysis of recent data for neutron shell closed nuclei in excitation energy range 40−80 MeV failed to arrive at a satisfactory description of the data and attributed the mismatch to shell effects and/or entrance channel effects, without reaching a definite conclusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In the present work we show that a well established stochastic dynamical code simultaneously reproduces the available data of pre-scission neutron multiplicities, fission and evaporation residue excitation functions for neutron shell closed nuclei 210Po and 212Rn and their isotopes 206Po and 214,216Rn without the need for including any extra shell or entrance channel effects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The calculations are performed by using a phenomenological universal friction form factor with no ad-hoc adjustment of model parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, we note significant deviation, beyond experimental errors, in some cases of Fr isotopes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' INTRODUCTION Fission of atomic nuclei is considered to be one of the most complex physical phenomena in nuclear physics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It involves rapid re-arrangement of nuclear matter with a delicate interplay between the macroscopic bulk matter and the microscopic quantal properties [1, 2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Though properties of fission have been studied exhaustively, many aspects of the dynamics are still not well-understood.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' For instance, discrepancies are reported between the mea- sured fission observables and the predictions of the clas- sical theory based on the standard Bohr-Wheeler statis- tical model of fission [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fission hindrance, enhanced pre-scission particle and giant dipole resonance (GDR) γ-ray multiplicities observed in hot nuclei suggested the effects of nuclear dissipation slowing down the fission pro- cess [4–10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' To account for frictional effects, Kramers dif- fusion model formalism with modified fission width [11], referred to as Kramers-modified statistical model was in- cluded in the standard statistical theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Although nature and strength of the nuclear dissipa- tion have been studied quite extensively, a simultaneous description of the experimental observables, namely, pre- scission neutron multiplicities (νpre), fission excitation functions and evaporation residue (ER) cross-sections still remains challenging.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Additionally, the dissipation coefficient is treated as an adjustable free parameter in the statistical model analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The pre-fission lifetime (or the dissipation strength), the level density parameter at ground state and saddle point deformation and fission barrier are empirically fitted to explain the νpre and/or fission and ER cross-section data [6, 12–15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As a result, the conclusions reached are often system dependent and are inadequate to provide a consistent description of the ∗ sugathan@gmail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='com fission process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Inadequate modelling of fission in statistical model can drastically influence the understanding of the fission phe- nomenon [16, 17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' This is especially observed in mass (A) ≈ 200 region that is explored here, to understand the role of N=126 neutron shell closure in the fissioning com- pound nucleus (CN).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' An anomalous increase in the exper- imental fission fragment angular anisotropy was reported for 210Po (N=126) as compared to 206Po (non-shell closed nuclei) across an excitation energy range (Eex) ≈ 40−60 MeV and was conjectured to be a manifestation of shell effects at the unconditional saddle [18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Further, a con- siderable amount of saddle shell correction was invoked to describe the experimental νpre data for 210Po nuclei [19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, a re-investigation of the experimental ex- citation functions and νpre data of 210Po ruled out any significant shell influence on the saddle [20] after corre- lated tuning of statistical-model parameters and inclu- sion of fission delay.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Another interesting aspect is the contradictory inter- pretation for correlation between neutron shell structure and nuclear dissipation strength that was required to re- produce the measured ER and νpre excitation functions in N=126 shell closed nuclei, namely 212Rn and 213Fr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The theoretical analysis of νpre data of 212Rn [21] and 213Fr [22] reported a low dissipation strength at Eex ≈ 50 MeV which was attributed to the influence of neu- tron shell closure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' On the contrary, no discernible shell influence was reported from ER cross-section studies of 212Rn and its isotope [23], though moderate nuclear dis- sipation was required to describe the data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It must be noted that the magnitude of dissipation invoked to ex- plain the experimental ER cross-sections varied within Rn isotopes [23, 24], which is again found to be different for the description of the νpre data [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Interestingly, in case of Fr nuclei, the finite-range liquid drop model fission barrier was scaled down, particularly for 213Fr to fit measured ER cross-section [15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' This reduction of the arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='13461v1 [nucl-th] 31 Jan 2023 2 fission barrier is in disagreement with the predictions for the shell closed nuclei [25].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' One notable observation is the reported interpretation of reduced survival probability of 213Fr nucleus due to neutron shell which is in contrast to the isotopic trend reported for Rn isotopes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Further, the fission cross-section of 213Fr is reported to exhibit no extra stability from N=126 shell closure [26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In the statistical model approach followed in these works, no at- tempts were made to extract a global prescription of the parameters, rather, a case specific adjustment of dissipa- tion strength was involved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The influence of neutron shell structures on the potential energy surface and hence fis- sion observables are still quite ambiguous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Apart from just shell influence, entrance channels effects are also probed in a couple of recent publications to understand the experimental νpre data for 213Fr nuclei [27, 28].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These studies reportedly observed a deviation in the measured data from the predictions of entrance channel model for 16O- and 19F-induced reactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These studies substantiate the view that no consistent picture has emerged from recent independent analysis of each fission observable for neutron shell closed nuclei 210Po, 212Rn and 213Fr and their isotopes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Inadequa- cies of standard statistical model interpretations have been addressed by employing Kramers-modified fission width taking into account shape-dependent level den- sity, temperature-dependent fission transition points, ori- entation (K state) degree of freedom and temperature- independent reduced dissipation coefficient [16, 17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' At- tempts for restraining the statistical-model parameters have also been reported [29], but a consistent description of experimental data for all three observables, namely νpre, fission and ER excitation functions for shell closed nuclei still could not be achieved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Recent developments in multi-dimensional stochastic approach are fairly suc- cessful in describing the fission characteristics of excited nuclei [30–35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, a simultaneous description of the experimental data and a systematic study for shell closed nuclei has not been attempted yet and is further required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In this paper, we show that the dynamical model based on 1D Langevin equation coupled with a statistical ap- proach [36] can simultaneously reproduce νpre, fission and ER cross-section data of shell closed nuclei over a range of excitation energies (Eex ≈ 40−80 MeV) of the measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The present calculations are performed without adjusting any of the model parameters, thus pro- vides a unified framework for a simultaneous study of these fission observables for nuclei in A ≈ 200 region.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' We re-investigated the available experimental data for the neutron shell closed nuclei 210Po, 212Rn and 213Fr, and their non-shell closed isotopes 206Po, 214,216Rn and 215,217Fr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is observed that a universal deformation- dependent reduced friction parameter is able to describe the fission observables simultaneously at all measured en- ergies irrespective of the shell structure of the nuclei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' THEORETICAL MODEL DESCRIPTION A combined dynamical and statistical model code [37] is utilized to compute the fission observables of nuclei under study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The detailed description of the theoreti- cal aspects of the model can be found in Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [36, 38].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The dynamical part of the model is carried out with a 1D Langevin equation of motion governed by a driving potential that is determined by free energy F(q, T), as employed in recent Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [39–44].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The free energy as derived from the Fermi gas model is related to the defor- mation dependent level density parameter a(q, A) as F(q, T) = V(q) - a(q, A)T 2 where T is the nuclear temper- ature, q is the dimensionless deformation coordinate de- fined as the ratio of half the distance between the center of masses of future fission fragments to the radius of CN and V (q) is the nuclear potential energy obtained from the finite-range liquid drop model [45, 46].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich [47] and Lestone et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [17] have emphasized on using nuclear entropy given by, S(q, A, Etot) = 2 � a(q, A)[Etot − V (q)] in determining the driving force and therefore, it is em- ployed as a crucial quantity in the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The nuclear driving force K = - dV (q) dq + da(q) dq T 2, not only consists of a conservative force but also contain a thermodynami- cal correction that enters the dynamics via.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' level density parameter a(q, A).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The deformation dependent level den- sity parameter used in constructing the entropy has the form [48]: a(q, A) = ˜a1A + ˜a2A2/3Bs(q) (1) where A is the mass number of the CN and ˜a1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='073 MeV−1 and ˜a2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='095 MeV−1 are taken from Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [49].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bs(q) is the dimensionless functional of the surface en- ergy [34, 38, 43, 50], expressed as the ratio of surface energy of the composite system to that of a sphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The over-damped Langevin equation which describes the fission process in the dynamical part of the model thus, has the form [36]: dq dt = T Mβ(q) �∂S(q) ∂q � Etot + � T Mβ(q)Γ(t) (2) where Etot is the total energy of the composite system that remains conserved and Γ(t) is a Markovian stochas- tic variable with a normal distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The reduced dis- sipation coefficient β(q) = γ/M (as employed in litera- ture, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=', Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [16, 29, 42, 44] (and Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' therein)) is the ratio of friction coefficient γ to the inertia param- eter M calculated with Werner-Wheeler approximation of an incompressible irrotational fluid [51].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The present model employs ”funny−hills” parameters {c,h,α} [52] for describing the shape of the fissioning nuclei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tak- ing into account only symmetric fission, the mass asym- metry parameter of the shape evolution is set to α=0 [36, 38, 50].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The dimensionless fission coordinate (q) is given by q(c,h)= ( 3c 8 )(1+ 2 15[2h+ (c−1) 2 ]c3), where c and h 3 defines the elongation and neck degree of freedom of the fissioning nucleus, respectively [36, 43, 53, 54].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Following the fission dynamics through full Langevin dynamical calculation is quite time consuming.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Similar to previous Langevin studies [31, 36, 39–43], a compu- tationally less intensive approach is adopted in present study where the dynamical stage is coupled with a sta- tistical model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In the present calculations, the emission of light particles from ground state to scission config- uration along the Langevin trajectories is treated as a discrete process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The evaporation of pre-scission light particles from ground state of Langevin trajectories to the scission point is coupled to the fission mode by a Monte Carlo procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The decay width for light parti- cle evaporation at each Langevin time step is calculated with the formalism as suggested by Fr¨obrich et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [36] and later incorporated in Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [34, 40–43].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The emission width of a particle of kind ν (n,p,α) is given by [55]: Γν = (2sν + 1) mν π2ℏ2ρc(Eex) × � (Eex−Bν) 0 dϵνρR(Eex − Bν − ϵν)ϵνσinv(ϵν) (3) where sν is the spin of emitted particle ν, and mν is its reduced mass with respect to the residual nucleus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The level densities of the compound and residual nuclei are denoted by ρc(Eex) and ρR(Eex − Bν − ϵν).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bν is the liquid-drop binding energy, ϵ is the kinetic energy of the emitted particle and σinv(ϵν) is the inverse cross sections [55].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The decay width for light particle emission is calculated at each Langevin time step τ [43, 53, 54].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' When a stationary flux over the barrier is reached af- ter a sufficiently long delay time, the decay of the CN is then modelled by an adequately modified statistical model [38, 56, 57].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' To have continuity when switching from dynamical to statistical branch, an entropy depen- dent fission width is incorporated in the latter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' While en- tering the statistical branch, the particle emission width Γν is re-calculated and the fission width Γf = ℏRf [36] is calculated with fission rate (Rf) given by, Rf = Tgs � |S ′′ gs|S ′′ sd 2πMβgs exp[S(qgs) − S(qsd)] × 2(1+erf[(qsc − qsd) � S ′′ sd/2])−1 (4) Here erf(x) = (2/√π) � x 0 dt exp(−t2) is the error func- tion and βgs is ground state dissipation coefficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The saddle-point (qsd) and the ground-state positions (qgs) are defined by the entropy and not, as in the conventional approach, by the potential energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The standard Monte Carlo cascade procedure was used to select the kind of decay with weights Γi/Γtot (i=fission,n,p,d,α) and Γtot = � i Γi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pre-scission particle multiplicities are calculated by counting the number of corresponding evaporated par- ticle events registered in the dynamical and statistical branch of the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The Langevin equation is started from a ground state configuration with a temperature corresponding to the initial excitation energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The fusion cross-section can be determined from the partial cross section dσ(l) dl which represent the contribution of angular momenta l to the total fusion cross-section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Each Langevin trajectory is started with an orbital angular momentum which is sam- pled from a fusion spin distribution that reads as [34, 36]: dσ(l) dl = 2π k2 2l + 1 1 + exp (l−lc) δl (5) The final results are weighted over all relevant waves, that is, the spin distribution is used as an angular momen- tum weight function with which the Langevin calcula- tions for fission are started.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As shown in recent Langevin studies, [34, 39–44], the spin distribution is calculated with the surface friction model [58].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' This calculation also fixes the fusion cross-section thus guaranteeing the correct normalization of fission and evaporation residue cross-sections within the accuracy of the surface friction model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The parameters lc and δl are the critical angular momentum for fusion and diffuseness, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The fission observables that will be discussed in sub- sequent sections are calculated in the model as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The pre-scission neutron multiplicity is the number of neutrons emitted by the CN till it reaches the scission configuration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The fission probability (Pf) is given by the ratio of fissioned trajectories to total trajectories.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The CN survival probability (1-Pf) is given by number of trajectories leading to ER formation divided by total trajectories and the fission (ER) cross-section is given by the product of fission (survival) probability and fusion cross-section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' RESULTS AND DISCUSSION In the present study, pre-scission neutron multiplic- ities, fission and ER excitation functions for 206,210Po, 212,214,216Rn and 213,215,217Fr compound nuclei are com- puted and compared with available experimental data wherein 210Po, 212Rn and 213Fr are N=126 neutron shell closed nuclei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The table I shows important parameters for the reactions studied in this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The dynamical cal- culations are performed with a universal frictional form of Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [36, 47, 57] without adjusting any of the model parameters with a consistent prescription of the dissipa- tion coefficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' To account for sufficient statistics, 107 Langevin trajectories are considered in the model calcu- lations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1 shows the results of dynamical calculations compared with the experimental data of νpre, fission and ER cross-sections for 206Po formed via.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 12C+194Pt [18, 19, 59] reaction and 210Po formed through two dif- ferent entrance channel reactions, namely 12C+198Pt [18, 19, 60] and 18O+192Os [5, 60, 61], spanning a wide range of excitation energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The excitation energies shown 4 TABLE I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Important parameters of reactions studied CN fissility Sn Bf(l=0) Reaction Mass excess (MeV) α/αBG (MeV) (MeV) target(proj) CN 206Po 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='717 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='99 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='51 12C+194Pt 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='79(0) 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='83 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='043 210Po 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='711 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='38 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='22 12C+198Pt 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='93(0) 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='33 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='050 18O+192Os -35.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='89(-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='78) -16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='982 212Rn 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='732 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='83 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='88 18O+194Pt -34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='79(-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='78) 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='970 214Rn 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='729 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='54 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='19 16O+198Pt -29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='93(-4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='74) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='77 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='996 216Rn 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='727 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='25 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='49 18O+198Pt -29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='93(-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='78) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='977 213Fr 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='743 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='06 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='83 16O+197Au -31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='16(-4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='74) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='987 19F+194Pt 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='79(-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='49) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='954 215Fr 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='740 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='76 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='13 19F+196Pt -32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='67 (-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='49) -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='958 217Fr 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='737 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='47 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='42 19F+198Pt 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='93(-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='49) 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='961 here are with respect to the liquid drop ground state CN mass and experimental mass of projectile and target [62].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Our calculations are restricted to excitation energies at and above 40 MeV where the present macroscopic model is valid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' We emphasize that the microscopic shell correc- tions are not accounted for in the present calculations, as we are dealing with hot nuclei where shell effects are expected to be negligible at high excitation energies that are populated in heavy-ion reactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The results of cal- culations using only the statistical model (dashed line) are also shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These calculations are made with the same code with Langevin dynamics turned off.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The statistical model calculations under-predict the mea- sured νpre data as shown in panels (a) to (c), even more so as excitation energy increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The dynamical model calculations using universal reduced friction coefficient are in excellent agreement with the measured data of νpre (panels (a) to (c)), fission cross-sections σfiss (pan- els (d) to (f)) and ER cross-sections σER (panels (g) to (i)) for the neutron shell closed nuclei 210Po as well as its isotope 206Po.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The measured data of 210Po formed through two different entrance channels agree well with the theory in a broad range of excitation energies up to 80 MeV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The model calculations describe the available experimental data for 206,210Po simultaneously at these excitation energies without any microscopic corrections included in the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These observations are at vari- ance with the statistical model analysis of 12C+194Pt and 12C+198Pt reactions that reported a significant shell cor- rection at the saddle deformation to describe the angular anisotropy and νpre data [18, 19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A recent 4D Langevin dynamical study [63] that was carried on 206Po and 210Po populated from reaction 12C+198Pt, reported a reason- able description of the measured data for these reactions without invoking any extra shell corrections at the saddle state;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' shown as open triangles in panels (a), (c), (d) and (f) of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A better agreement of the measured data is observed for 12C+198Pt reaction in comparison to its 4D Langevin calculations [63], particularly at low excitation energies as shown in panels (a) and (d) of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The overestimation of νpre and fission cross-section of 210Po in Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [63] was attributed to the remnant of ground state shells and hence, a consequence of not using a pure macroscopic potential energy surface as suggested in Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [64].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nonetheless, the predictions of multi-dimensional Langevin model for νpre data of 206Po by Karpov et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [30] are also found to be in reasonable agreement with the results of the present analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Moreover, the measured mass distribution of fragments in the fission of 206,210Po [65, 66] reaffirms the absence of any shell corrections on the potential energy surface at the saddle point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 2 and 3 display the comparison between ex- perimental data and theoretical calculations of νpre, fis- sion, ER and fusion cross-sections for N=126 shell closed nuclei viz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 212Rn [21, 23, 24, 67] formed through re- action 18O+194Pt and 213Fr formed through reactions 19F+194Pt [15, 22, 26, 68] and 16O+197Au [5, 6], and their non-shell closed isotopes 214,216Rn populated via.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' reactions 16,18O+198Pt [21, 23, 24, 67] and 215,217Fr pop- ulated via.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' reactions 19F+196,198Pt [15, 22, 26, 68].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The model calculations describe the νpre and fission excita- tion functions for 212Rn and its isotopes 214,216Rn quite successfully.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In reactions forming 213,215,217Fr nuclei, the same parameter set is able to account for the experi- mental fission excitation functions but not νpre.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A re- cent work [26] using an extended version of statistical- model employing collective enhancement of level density also reported an under-estimation of νpre data for same reactions when fitted simultaneously with fission cross- section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In the present work, the disagreement between experimental νpre and theory is prominent above ≈50 MeV excitation energy and it increases with rise in exci- tation energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Considering that νpre of other studied nu- clei are well reproduced by the model, it is unclear why 5 0 1 2 3 4 νpre (a) 12C+198Pt −→ 210Po (b) 18O+192Os −→ 210Po (c) 12C+194Pt −→ 206Po 100 101 102 103 σfiss(mb) (d) (e) (f) 40 60 80 100 100 101 102 103 σER(mb) (g) 40 60 80 100 (h) 40 60 (i) Eex (MeV) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' (Colour online) Measured and calculated pre-scission neutron multiplicities (νpre), fission cross-sections (σfiss) and evap- oration residue cross-sections (σER) as a function of excitation energy for the reactions 12C+198Pt, 18O+192Os and 12C+194Pt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The continuous line (red) denote calculated results with a universal frictional form factor and dashed line (black) represent statistical model calculations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The symbols in the legend represent different experimental data sets, for νpre: (filled squares) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [19], (filled circles) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [5] and (open square) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [59];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' for σfission and σER: (filled diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [18], (filled hexagons) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [61] and (open diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [60].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The open triangles represent results of νpre and σfission from 4D Langevin calculations of Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [63].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' the same frictional form fails, particularly for reactions forming Fr nuclei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is to be noted that, an energy de- pendent dissipation was used in Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [21, 22] to describe the νpre data for these reactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' We also attempted similar approach by employing a temperature-dependent friction (TDF) in the stochastic calculations [69] (with- out changing any other parameter).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' This frictional form factor is deformation dependent, unlike the ones used in Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [21, 22, 70].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The maximum of β(q) in TDF corre- sponds to the ground state, that tends to decrease with increasing deformation with its minimum near the sad- dle configuration and is followed by an increase in the dissipation strength when approaching the scission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The dissipation coefficient assumes a higher value with in- creasing temperature of the CN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is observed that a better agreement of νpre data is achieved for reactions 19F+194,196,198Pt and 16O+197Au after invoking temper- ature dependence of the dissipation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The same frictional form, however, is found to over-predict the measured νpre data of other studied nuclei and hence is not shown here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Deviation in ER excitation functions are also to be noted for 212Rn and 213,215,217Fr nuclei wherein the cal- culated ER cross-sections underpredict the experimental data for these nuclei at high excitation energies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The case 6 0 2 4 νpre (a) 18O+198Pt −→ 216Rn (b) 16O+198Pt −→ 214Rn (c) 18O+194Pt −→ 212Rn 101 102 103 σfiss(mb) (d) (e) (f) 101 102 103 σER(mb) (g) (h) (i) 40 60 80 101 102 103 σfus(mb) (j) 40 60 80 (k) 40 60 80 (l) Eex (MeV) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' (Colour online) Measured and calculated pre-scission neutron multiplicities (νpre), fission cross-sections (σfiss), evapora- tion residue cross-sections (σER) and fusion cross-sections (σfus) as a function of excitation energy for the reactions 18O+198Pt, 16O+198Pt, 18O+194Pt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The continuous (red) and dashed (black) lines have the same meaning as in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The calculations of fusion cross-section are independent of the frictional form and are represented by dotted line (brown).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The symbols in the legend represent different experimental data sets, for νpre: (filled squares) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [21];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' for σfiss: (filled diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [67] and (open diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [23];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' for σER: (filled circles) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [24] and (filled hexagons) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [23] and for σfus: (filled triangles) Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [23, 24].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' of Rn isotopes is of particular interest as the ER cross- section data for 214,216Rn [24] agrees fairly well with the model calculations at all measured energies but differ for 212Rn [23] except at the lowest energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' For 213,215,217Fr nuclei, the measured ER cross-sections of Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [15] differ above excitation energy ≈ 55 MeV and the deviation is prominent for 213,215Fr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is quite interesting to note that the ER measurement by a different group [68] for same reactions forming 213,217Fr at Eex ≤ 55 MeV follows the trend of the model predictions quite successfully.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Unfor- tunately, Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [68] has reported only three data points.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Moreover, the ER cross-section data of 215Fr formed in reaction 18O+197Au [71] is reproduced reasonably well with results of 19F+196Pt particularly, above 50 MeV ex- citation energy (displayed as open pentagons in panel (j) of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The present dynamical calculations assume 7 0 2 4 6 νpre (a) 19F+198Pt → 217Fr (b) 19F+196Pt → 215Fr (c) 19F+194Pt → 213Fr (d) 16O+197Au → 213Fr 101 102 103 σfiss(mb) (e) (f) (g) (h) 101 102 103 σER(mb) (i) (j) (k) (l) 50 75 100 101 102 103 σfus(mb) (m) 50 75 100 (n) 50 75 100 (o) 50 100 (p) Eex (MeV) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' (Colour online) Measured and calculated pre-scission neutron multiplicities (νpre), fission cross-sections (σfiss), evapora- tion residue cross-sections (σER) and fusion cross-sections (σfus) as a function of excitation energy for the reactions 19F+198Pt, 19F+196Pt, 19F+194Pt and 16O+197Au.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The continuous (red), dashed (black) and dotted (brown) lines have the same meaning as in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1 and 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The dash-dotted line (magenta) represent calculated results with temperature-dependent friction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The symbols in the legend represent different experimental data sets, for νpre: (filled squares) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [22] and (partially filled squares) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [5] ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' for σfiss: (filled diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [26],(partially filled diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [6] and (open diamonds) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [68];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' for σER: (filled circles) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [15], (partially filled circles) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [6] and (open circles) Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [68] and for σfus: (filled triangles) Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [15, 26, 68] and (open triangles) Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The open pentagons denote σER for 215Fr nuclei formed via 18O+197Au Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [71].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' decay from an equilibrated CN and any entrance channel effects are not included.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It takes account of only the dif- ferent angular momenta that are populated in different entrance channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Taking into consideration the insignif- icant difference in angular momenta between two en- trance channels forming 215Fr, the observed deviation in ER cross-section for 19F-induced reaction is quite unex- pected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These observations further necessitated the need to confront the deviations in describing ER cross-sections by comparing the measured fusion cross-sections for Rn and Fr nuclei with the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is revealed that the cal- culated fusion cross-sections are in good agreement with the measured fusion data, augmenting the validity of the present calculations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Furthermore, the under-prediction of ER cross-sections indicates the need for a strong dis- sipation in the pre-saddle region [72].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, 3D 8 Langevin dynamical calculations [31] reported a reduc- tion in the wall friction coefficient to reproduce the mass and kinetic energy distribution of fission fragments, and their influence on νpre for 215Fr nucleus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The strength of the reduction coefficient, ks = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='25 − 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='5 indicates a weak dissipation in the initial stages of the fissioning nucleus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The experimental analysis of fission fragment nuclear-charge distributions and fission cross-sections of Fr, Rn isotopes and their neighbouring nuclei also re- ported a pre-saddle dissipation strength of magnitude (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='5) × 1021 s−1 [73] and 2 × 1021 s−1 [74], respec- tively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The more recent microscopic study of energy de- pendent dissipation using time-dependent Hartree-Fock + BCS method [75] also observed a strength of deforma- tion dependent friction coefficient, ranging from 1 to 6 × 1021 s−1 in heavy nuclei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The strength of these fric- tional parameterizations are quite in agreement with the dissipation form factor employed in the present calcula- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These observations affirm a weak dissipation in the pre-saddle region;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' so, the observed enhancement of ER cross-sections in Fr nuclei populated via.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 19F-induced reactions is not well-understood from the perspective of dissipation strength alone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In fact, a satisfactory de- scription of the excitation functions including ER cross- sections for reactions 12C+194Pt, 12C+198Pt, 18O+192Os and 16,18O+198Pt and survival probabilities for a range of fissilities [36] is observed within the framework of this 1D Langevin dynamics with a universal friction param- eter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, it is also important to bear in mind the possible bias coming from experimental uncertainty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is striking that the observed deviations are pronounced in ER cross-section data where measurements are reported to have large uncertainty in ER separator transmission efficiency [15, 23].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It would be highly desirable to have additional ER measurements to rule out any possible ex- perimental bias in the interpretation of ER data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It must be noted that, the entrance channel dynam- ics of the fusion stage might also play a role influenc- ing neutron emission at the formation stage [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is known that interplay of CN excitation energy, angular momentum and fission barrier play crucial role in fission process [28].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Present study do not take into account any entrance channel dynamics influencing the fusion stage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The model only considers the entrance channel depen- dent ’l’ distribution calculated within the surface friction model [58].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 4 we show the calculated fission bar- rier height Bf(l) for three compound systems and mean angular momentum < l > calculated from ’l’ distribu- tion for different entrance channels forming same CN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The variation of Bf is plotted as a function of ’l’ in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 4(a) and variation of < l > of the compound systems is plotted as a function of Eex in Fig 4(b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' From Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 4, it is clear that, the difference in angular momenta be- tween two entrance channels forming same CN at similar Eex is not very significant to cause any ’l’ induced ef- fects on measured fission observable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' This is evident in the νpre data for 210Po formed in reactions 12C+198Pt and 18O+192Os which are well described in the present work (see Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1) without invoking any entrance channel effects in the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Recent studies investigating en- trance channel dynamics [27, 28] reported disagreement between experimental νpre and predictions of entrance channel model for 213Fr nuclei formed via.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 16O+197Au and 19F+194Pt reactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' These studies were, however, not extended to other isotopes of Fr, namely 215,217Fr that also show similar discrepancy as reported in the present study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The current 1D Langevin analysis provides a simul- taneous description of the experimental data for neutron magic nuclei 210Po without invoking any saddle shell cor- rections or a nuclear dissipation strength dependent on system/observable under study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In order to understand qualitatively that consideration of saddle shell correc- tions are not required to explain νpre data, we consider the nature of neutron emission during the fission process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is to be noted that these neutrons are emitted from dy- namical trajectories that originated from compact config- uration till scission point is reached.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The prompt and beta-delayed neutron emissions from fission fragments are not taken into consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As recent publica- tions have advocated for the inclusion of shell correc- tions on the saddle configuration to describe the angular anisotropy and νpre data at moderate excitation energies [12, 18, 19], we have attempted to find the distribution of pre-scission neutrons as it evolves from ground state to scission point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The model calculated potential energy V(q) and distribution of percentage yield of pre-scission neutrons are plotted as a function of the deformation co- ordinate (q) for these nuclei at 50 MeV excitation energy and shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It is evident that more than 90% of the neutron emission occurs at an early stage of fission before the saddle deformation (q ≈ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='8) [38] is reached.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The mean of the distribution corresponds to νpre emis- sion close to the ground state configuration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' In-fact, a multi-dimensional Langevin study of 215Fr by Nadtochy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [31] have also pointed out that an appreciable part of pre-scission neutrons are emitted at an early stage of fission before saddle is reached.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As most of the neutrons are emitted close to the ground state configuration, it is unlikely to be influenced by any shell corrections applied at the saddle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Though the present code uses classical 1D approach to describe fission observables, the main objective of this work is to have a simultaneous description of experi- mental data without any parameter adjustment thus, removing some of the reported ambiguities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A com- parison between νpre calculated with 1D model and re- cent macroscopic multi-dimensional models is displayed in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It can be seen that the νpre values predicted by different models are very similar and also reproduce the measurements quite well for reactions spanning a wide range of fissility parameter Z2/A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Additionally, the multi-dimensional calculations [34, 50, 76] also use the formalisms adopted from Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [36, 69] such as the parameterization of surface friction model and weakest coordinate dependence of the level-density parameter as 9 0 20 40 60 80 ℓ (¯h) 0 2 4 6 8 10 12 14 Bf (MeV) (a) 210Po 212Rn 213Fr 30 40 50 60 70 80 90 10 15 20 25 30 35 40 45 < ℓ > (¯h) (b) 12C+198Pt 18O+192Os 19F+194Pt 16O+197Au Eex (MeV) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' (Colour online) (a) The angular momentum ’l’ de- pendent fission barrier height Bf(l) for three CN 210Po,212Rn and 213Fr and (b) Variation of mean angular momentum < l > with compound nucleus excitation energy for 210Po,and 213Fr populated by different entrance channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' employed in the present work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hence, the qualitative nature of the observed features presented here is not ex- pected to be different with multi-dimensional approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As the present framework is found to provide realistic values close to measured data, we believe that the 1D approach still can be a potential tool to study a wider systematics which can be accomplished within minimum 0 10 20 30 40 V (MeV) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2 0 1 2 3 4 5 6 7 8 d<νpre>/dq (%) qneck qsadd 210Po 212Rn 213Fr deformation coordinate (q) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' (Colour online) Potential energy distribution as a function of nuclear deformation coordinate (q) for three fis- sioning nuclei 210Po, 212Rn and 213Fr (top panel) and distri- bution of percentage yield of evaporated pre-scission neutrons as a function of (q) for three CN at 50 MeV excitation energy (bottom panel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The deformation coordinate (q) assumes a value of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='6 (qneck) when the neck of the fissioning nucleus starts to develop and q=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='8 (qsadd) at the saddle state con- figuration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' computational resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It must be remarked here that, even though present analysis provides a reasonable reproduction of the exper- imental data without invoking any shell corrections at high excitation energies, it shall not be concluded from this work that shell effects are not relevant in the analy- sis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As present investigation consider only the first chance fission at Eex ∼ 40 MeV and above where shell effects are expected to be washed out, no indication for the need of including shell corrections was found.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, for the case when the CN is populated at low excitation energies or reaches low excitation energy due to neutron emission as a consequence of competition between neutron evapo- ration and fission (multi-chance fission), the microscopic effects are required to be taken into consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Re- cent microscopic study of dissipation within Hartree-Fock + BCS framework [75] have shown a strong dependence of dissipation on deformation and initial excitation ener- gies of the hot nuclei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Possible influence of microscopic temperature dependence of fission barrier height and its curvature were also emphasized in some recent studies of fully microscopic description of fission process [77, 78].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 10 30 35 40 Z2/A 0 1 2 3 4 5 6 νpre 162Yb 206Po 210Po 215Fr 244Cm 264Rf 216Ra 248Cf Expt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' data Present Multi-dimensional model FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' (Colour online) Comparison of measured pre- scission neutron multiplicities (νpre) with the results of the 1D model (present work) and multi-dimensional models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The filled triangles (blue) denote experimental data [6, 14, 59, 80– 83], the present dynamical model calculations are represented by filled circles (orange) and the filled squares (green) de- note the results of multi-dimensional dynamical calculations [28, 30, 34, 84].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A microscopic framework based on the finite-temperature Skyrme-HartreeFock+BCS approach [79] was adopted to demonstrate the essential role of energy dependent fission barriers by studying the experimental fission probability of 210Po.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' It would be quite interesting to extend the in- vestigation of Fr nuclei within such a microscopic frame- work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' SUMMARY AND CONCLUSION In the present work we report a systematic study on the fission dynamics of N=126 shell closed nuclei in mass region 200 with a simultaneous description of three fis- sion observables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The present work highlights the limited reliability of the conclusions drawn from the recent statis- tical model analysis of shell closed nuclei, namely 210Po, 212Rn and 213Fr at excitation energies 40 MeV and above, that advocated for extra shell effects at saddle configu- ration even after their inclusion in the level density for- mulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Earlier analyses of νpre and ER cross-sections were based on different assumptions and case dependent parameter adjustments, without reaching a definite con- clusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' On the basis of present analysis we conclude that, without many of those assumptions and parameter adjustments, a well established combined dynamical and statistical model can simultaneously reproduce the avail- able data of νpre, fission and evaporation residue excita- tion functions (also fusion cross-sections in certain cases) for neutron shell closed nuclei, viz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 210Po, 212Rn and their non-shell closed isotopes 206Po and 214,216Rn with- out the need of including any extra shell effects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' There appears to be no discernible influence of N=126 neutron shell structure on these measured fission observables in the medium excitation energy range.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The present work also points to a relatively smaller role of entrance channel effects in the studied systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' However, we find a significant mismatch between mea- sured νpre data and its model predictions for Fr nuclei formed in reactions 19F+194,196,198Pt and 16O+197Au, despite a reasonable description of fission and fusion cross-sections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The νpre data in Fr nuclei could only be reproduced after invoking a temperature dependent frictional form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' The difficulty in completely reproduc- ing some specific measurements of Fr nuclei still remains not well-understood and additional measurements are de- sired.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Although the present work is limited to the study of three fission observables, it would also be interesting to extend the systematic study using recent microscopic theory within Hartree-Fock + BCS framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' ACKNOWLEDGMENTS We are thankful to K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saneesh for fruitful discussions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' One of the authors (D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=') acknowl- edges the financial support in the form of research fel- lowship received from the University Grants Commission (UGC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' REFERENCES [1] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Vandenbosch, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Huizenga, Nuclear Fission, Aca- demic Press, New York, 1973.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [2] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Krappe, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pomorski, Theory of Nuclear Fission, Lecture Notes in Physics, Springer, Heidelberg, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [3] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bohr, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wheeler, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 56 (1939) 426.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='426.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [4] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gavron, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Beene, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Cheynis, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ferguson, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Obenshain, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Plasil, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Young, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Petitt, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J¨a¨askel¨ainen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sarantites, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Maguire, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 47 (1981) 1255.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 47.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1255.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [5] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Newton, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hinde, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Charity, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Leigh, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bokhorst, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Foote, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ogaza, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 483 (1988) 126.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/ 0375-9474(88)90068-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 11 [6] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hinde, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Charity, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Foote, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Leigh, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Newton, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ogaza, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 452 (1986) 550.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/0375-9474(86)90214-9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [7] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lestone, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Leigh, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Newton, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hinde, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wei, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Elfstrom, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Popescu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 67 (1991) 1078.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1078.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [8] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rossner, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hilscher, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hinde, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gebauer, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lehmann, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wilpert, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mordhorst, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 40 (1989) 2629.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2629.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [9] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Thoennessen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chakrabarty, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Herman, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Butsch, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Paul, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 59 (1987) 2860.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2860.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [10] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hofman, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Back, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Paul, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 51 (1995) 2597.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2597.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [11] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kramers, Physica 7 (1940) 284.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/ S0031-8914(40)90098-2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [12] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kapoor, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 74 (2006) 041301(R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='041301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [13] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mancusi, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Charity, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Cugnon, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 82 (2010) 044610.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='044610.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [14] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Choudhury, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kapoor, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadkarni, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 49 (1994) 932.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='932.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [15] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kumar, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Madhavan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nath, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gehlot, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mohanto, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhin- gan, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mukul, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Varughese, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sadhukhan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Goyal, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Santra, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 89 (2014) 024609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='024609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [16] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' McCalla, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lestone, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 101 (2008) 032702.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='032702.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [17] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lestone, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' McCalla, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 79 (2009) 044611.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='044611.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [18] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Shrivastava, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Samant, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Navin, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tomar, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 82 (1999) 699.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='699.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [19] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mittal, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sandal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Goyal, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gehlot, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Dhal, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhowmik, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 913 (2013) 157.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/ j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='nuclphysa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='05.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [20] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kapoor, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 92 (2015) 034602.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='034602.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [21] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sandal, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kumar, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhowmik, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kalkal, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Siwal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Goyal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mandal, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Prasad, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sadhukhan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 87 (2013) 014604.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='014604.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [22] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sadhukhan, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Siwal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Goyal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Santra, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kumar, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhowmik, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 86 (2012) 014609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='014609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [23] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Laveen, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Prasad, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Madhavan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sad- hukhan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nath, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gehlot, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Varier, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Thomas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Vinodkumar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Shamlath, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Varughese, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Babu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ap- pannababu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' San- dal, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' John, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G: Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 42 (2015) 095105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1088/0954-3899/ 42/9/095105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [24] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sandal, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kumar, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sharma, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Madhavan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nath, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gehlot, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mandal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Verma, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Prasad, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Varier, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Vinodkumar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sadhukhan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 91 (2015) 044621.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='044621.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [25] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M¨oller, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sierk, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ichikawa, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Iwamoto, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bengtsson, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Uhrenholt, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' ˚Aberg, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 79 (2009) 064304.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='064304.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [26] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kaur, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Siwal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Goyal, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saxena, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G: Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 48 (2021) 075104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1088/1361-6471/abe8cd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [27] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Shareef, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Prasad, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 52 (2016) 342.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1140/epja/i2016-16342-4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [28] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmitt, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mazurek, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 97 (2018) 014616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='014616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [29] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Banerjee, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nath, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B 776 (2018) 163.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='physletb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='040.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [30] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Karpov, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Vanin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Adeev, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 63 (2001) 054610.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='054610.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [31] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Adeev, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Karpov, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 65 (2002) 064615.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='064615.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [32] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Randrup, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M¨oller, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sierk, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 84 (2011) 034613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='034613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [33] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Randrup, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M¨oller, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 106 (2011) 132503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='132503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [34] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ryabov, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gegechkori, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Anischenko, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Adeev, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 89 (2014) 014616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='014616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [35] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mazurek, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ryabov, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Adeev, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 53 (2017) 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1140/epja/ i2017-12262-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [36] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gontchar, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Reports 292 (1998) 131.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/S0370-1573(97)00042-2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [37] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gontchar, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Litnevsky, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, Computer Physics Communications 107 (1997) 223.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/ S0010-4655(97)00108-2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [38] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gontchar, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pischasov, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 47 (1993) 2228.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='47.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2228.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [39] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chaudhuri, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 18 (2003) 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1140/epja/i2003-10038-x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [40] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kun, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Yu-Gang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Yi-Bin, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Xiang-Zhou, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jin- Gen, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' De-Qing, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wei, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Guo-Liang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wen-Qing, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wen-Dong, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Xing-Fei, Chinese Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 22 (2005) 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1088/0256-307x/22/1/016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [41] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wei, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Feng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hong-Wei, Chinese Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 32 (2008) 816.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1088/1674-1137/32/10/010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [42] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Feng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wei, Chinese Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 34 (2010) 551.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1088/1674-1137/34/5/007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [43] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ye, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 97 (2018) 014603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='014603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [44] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ye, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 103 (2021) 024611.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='024611.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [45] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Myers, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Swiatecki, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 81 (1966) 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/0029-5582(66)90639-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [46] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Krappe, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nix, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sierk, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 20 (1979) 992.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='992.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [47] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 787 (2007) 170.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/ j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='nuclphysa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='028.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [48] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Balian, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bloch, Annals of Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 60 (1970) 401.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/0003-4916(70)90497-5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [49] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ignatyuk, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Itkis, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Okolovich, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Smirenkin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tishin, Yad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fiz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 21 (1975) 1185.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 12 [50] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chaudhuri, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 65 (2002) 054612.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='054612.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [51] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Davies, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sierk, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nix, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 13 (1976) 2385.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2385.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [52] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Brack, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Damgaard, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jensen, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pauli, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Strutinsky, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wong, Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mod.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 44 (1972) 320.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/revmodphys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='320.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [53] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ye, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tian, Phys, Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 93 (2016) 044603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='044603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [54] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ye, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tian, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 91 (2015) 064603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='064603.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [55] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Blann, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 21 (1980) 1770.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/ PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1770.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [56] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mavlitov, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gonchar, Z Physik A 342 (1992) 195.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1007/bf01288469.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [57] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gontchar, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mavlitov, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 556 (1993) 281.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/0375-9474(93)90352-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [58] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Marten, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fr¨obrich, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 545 (1992) 854.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/0375-9474(92)90533-p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [59] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chubaryan, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Itkis, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Luk’yanov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Okolovich, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Penionzhkevich, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rusanov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Salamatin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Smirenkin, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Atom.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 56 (1993) 286.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [60] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' van der Plicht, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Britt, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fowler, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fraenkel, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gavron, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wilhelmy, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Plasil, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Awes, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Young, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 28 (1983) 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/ PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [61] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Charity, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Leigh, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bokhorst, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chat- terjee, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Foote, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hinde, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Newton, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ogaza, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ward, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 457 (1986) 441.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/ 0375-9474(86)90388-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [62] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Moller, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nix, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Myers, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Swiatecki, At.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' data and Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' data tables 59 (1995) 185–381.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1006/adnd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1995.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [63] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmitt, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mazurek, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B 737 (2014) 289.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='physletb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='08.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='072.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [64] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 95 (2017) 054616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='054616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [65] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ghosh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhattacharya, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Banerjee, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhattacharya, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kundu, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mukherjee, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Asgar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Dey, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Dhal, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Shaikh, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Meena, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Manna, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pandey, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rana, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Roy, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Roy, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Srivas- tava, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhattacharya, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 96 (2017) 064609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='064609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [66] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chaudhuri, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ghosh, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Banerjee, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhat- tacharya, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sadhukhan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kundu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhattacharya, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Meena, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mukherjee, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Saha, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' As- gar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Dey, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Manna, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pandey, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rana, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Roy, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Roy, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Srivastava, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bhattacharya, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Biswas, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Joshi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Shrivastava, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Vind, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 92 (2015) 041601(R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='041601.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [67] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sandal, Ph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' thesis, Panjab University (2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [68] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mahata, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kailas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Shrivastava, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Santra, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tomar, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 65 (2002) 034613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='034613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [69] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gontchar, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Litnevsky, Z Physik A 359 (1997) 149.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1007/s002180050382.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [70] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Thoennessen, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bertsch, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 71 (1993) 4303.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='4303.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [71] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Corradi, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Fioretto, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gadea, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Latina, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Stefanini, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Szilner, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Trotta, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Beghini, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Montagnoli, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Scarlassara, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sagaidak, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Atutov, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mai, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Stancari, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Tomas- setti, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Mariotti, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Khanbekyan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Veronesi, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 71 (2005) 014609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 014609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [72] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Back, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Blumenthal, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Davids, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hen- derson, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hermann, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hofman, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jiang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Penttil¨a, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wuosmaa, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 60 (1999) 044602.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='044602.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [73] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmitt, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Heinz, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jurado, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Keli´c, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmidt, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 99 (2007) 042701.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='042701.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [74] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jurado, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmitt, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmidt, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Benlliure, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Enqvist, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Junghans, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Keli´c, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rejmund, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 93 (2004) 072501.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='072501.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [75] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Qiang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pei, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 104 (2021) 054604.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='054604.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [76] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nadtochy, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ryabov, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gegechkori, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Anischenko, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Adeev, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 85 (2012) 064619.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='064619.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [77] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Zhu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pei, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 94 (2016) 024329.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='024329.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [78] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Qiao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pei, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 106 (2022) 014608.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='014608.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [79] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Goodman, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 352 (1981) 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 1016/0375-9474(81)90557-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [80] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hinde, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Hilscher, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rossner, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Gebauer, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lehmann, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Wilpert, Physical Review C 45 (1992) 1229.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1229.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [81] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Golda, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Pal, Ranjeet, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sandal, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Behera, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Jhingan, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Singh, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Sugathan, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chatterjee, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Datta, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Kumar, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Viesti, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Govil, Physical Review C 78 (2008) 024609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='024609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [82] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Itkis, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Lukyanov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Okolovich, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Penionzkevich, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rusanov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Salamatin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Smirenkin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Chubaryan, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' At.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' 52 (1990) 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [83] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Ferguson, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Plasil, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Freiesleben, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Bemis, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Schmitt, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' C 8 (1973) 1104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi:doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1103/physrevc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' [84] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Eslamizadeh, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' A 47 (2011) 134.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
+page_content='1140/epja/i2011-11134-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/DtFRT4oBgHgl3EQfATdu/content/2301.13461v1.pdf'}
diff --git a/E9E5T4oBgHgl3EQfVA-E/vector_store/index.pkl b/E9E5T4oBgHgl3EQfVA-E/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c661a31ce819ab255adaab90ce6083cc32015a34
--- /dev/null
+++ b/E9E5T4oBgHgl3EQfVA-E/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:090a747c540e22257860bf791a23cf50c605c5ae527adf06e50d2463514883b0
+size 140592
diff --git a/F9AyT4oBgHgl3EQfSveK/content/2301.00092v1.pdf b/F9AyT4oBgHgl3EQfSveK/content/2301.00092v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..c9122e869ff59577cd40794a883ab62973ef0cc2
--- /dev/null
+++ b/F9AyT4oBgHgl3EQfSveK/content/2301.00092v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1854a136dbc8347d1ce8628cb8117bce3cd0829710e3f90895bad14373cc492
+size 640750
diff --git a/F9AyT4oBgHgl3EQfSveK/vector_store/index.pkl b/F9AyT4oBgHgl3EQfSveK/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..ce910c75e379b08d2050a9da8b361bcb787c0427
--- /dev/null
+++ b/F9AyT4oBgHgl3EQfSveK/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0e11080dd82a6238d3fd8a84a2d8e5919afb2abf334e09f91c100641ad2e51c
+size 326724
diff --git a/F9E1T4oBgHgl3EQf-wbm/content/2301.03574v1.pdf b/F9E1T4oBgHgl3EQf-wbm/content/2301.03574v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..447ed18ca1f8727791cf528db081c4ee0834733a
--- /dev/null
+++ b/F9E1T4oBgHgl3EQf-wbm/content/2301.03574v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a50db6c95094dcef8d1ff843d0bbaf33f2d86669776d31bc5a9a20dda283a65a
+size 400397
diff --git a/F9E1T4oBgHgl3EQf-wbm/vector_store/index.faiss b/F9E1T4oBgHgl3EQf-wbm/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..159931e268da05283efe833283d3e0f5efb0d94f
--- /dev/null
+++ b/F9E1T4oBgHgl3EQf-wbm/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2edf41f660d8ec22479683209718a6a6d7a8e68be9c18776392b4e74f69282e7
+size 5505069
diff --git a/F9E1T4oBgHgl3EQf-wbm/vector_store/index.pkl b/F9E1T4oBgHgl3EQf-wbm/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..2d37a5e1f7f3be50ed78198ecb0cf72bf83a4ea8
--- /dev/null
+++ b/F9E1T4oBgHgl3EQf-wbm/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b73447a4c685a45e43a0b8ae2b967f7e9e79ee5ca78645f9d9fd17f3567bffb4
+size 190443
diff --git a/FtE0T4oBgHgl3EQfRAB3/content/2301.02201v1.pdf b/FtE0T4oBgHgl3EQfRAB3/content/2301.02201v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..dfd2204ae91aa67e00351454571d9f82ebaafc68
--- /dev/null
+++ b/FtE0T4oBgHgl3EQfRAB3/content/2301.02201v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2df3c9d83574f249f367d043f80683214f1d19132bbfa92453309b8041149539
+size 1149472
diff --git a/FtE0T4oBgHgl3EQfRAB3/vector_store/index.pkl b/FtE0T4oBgHgl3EQfRAB3/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..b59d75ccf1487105be42debf1e70054444962e3e
--- /dev/null
+++ b/FtE0T4oBgHgl3EQfRAB3/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9613d4a5b52d11a5e650d87a2e18388b6e3215333a8b039bdb114e1ae430179
+size 105036
diff --git a/G9E2T4oBgHgl3EQfTge-/content/2301.03804v1.pdf b/G9E2T4oBgHgl3EQfTge-/content/2301.03804v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..196d8702a62c414fa33e947b05e4d20daa6369fb
--- /dev/null
+++ b/G9E2T4oBgHgl3EQfTge-/content/2301.03804v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a68c3fc4a7de3a434471b81fe59cfbe78f72fec2430d675a9efbf38d809b767
+size 849875
diff --git a/G9E2T4oBgHgl3EQfTge-/vector_store/index.faiss b/G9E2T4oBgHgl3EQfTge-/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..f85d0a42e65cbbadd4e47656ced2c7088be936ab
--- /dev/null
+++ b/G9E2T4oBgHgl3EQfTge-/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1886ff16d9f6c571b4dc737d0e7543d26d8c28e557d5126685848d979f7ca3f
+size 15728685
diff --git a/G9E2T4oBgHgl3EQfTge-/vector_store/index.pkl b/G9E2T4oBgHgl3EQfTge-/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..377a4cc990d26eff7c371538bf2d521285eaeb4b
--- /dev/null
+++ b/G9E2T4oBgHgl3EQfTge-/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da765036a541969f0b0507941ab3aef426448b686bdc475ec1b74d41658bbfe6
+size 868337
diff --git a/I9E5T4oBgHgl3EQfXA_-/content/tmp_files/2301.05564v1.pdf.txt b/I9E5T4oBgHgl3EQfXA_-/content/tmp_files/2301.05564v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..47456b606ffc51feddaee7049adb562b8ceafded
--- /dev/null
+++ b/I9E5T4oBgHgl3EQfXA_-/content/tmp_files/2301.05564v1.pdf.txt
@@ -0,0 +1,529 @@
+arXiv:2301.05564v1 [math.NT] 13 Jan 2023
+ABELIAN SURFACES WITH SUPERSINGULAR GOOD REDUCTION
+AND NON SEMISIMPLE TATE MODULE
+MAJA VOLKOV
+Abstract. We show the existence of abelian surfaces A over Qp having good reduction
+with supersingular special fibre whose associated p-adic Galois module Vp(A) is not
+semisimple.
+2000 Mathematics Subject Classification: 11G10, 14K15, 14G20.
+Keywords: Abelian varieties, local fields, Galois representations.
+Contents
+Introduction
+1
+1.
+The general method
+2
+2.
+A lift of the twofold product of a supersingular elliptic curve
+4
+3.
+A lift of a simple supersingular abelian surface
+6
+References
+7
+Introduction
+Fix a prime number p and an algebraic closure Qp of Qp. Write G = Gal(Qp/Qp) for
+the absolute Galois group of Qp. For a d-dimensional abelian variety A over Qp let A[pn]
+be the group of pn-torsion points with values in Qp and
+Vp(A) =
+def Qp ⊗Zp lim
+←−
+n≥1
+A[pn].
+This is a 2d-dimensional Qp-vector space on which G acts linearly and continuously. We
+want to consider the following problem: find abelian varieties A over Qp having good
+reduction with supersingular special fibre and such that the Galois module Vp(A) is not
+semisimple. In this paper we show the existence of two such varieties with nonisogenous
+special fibres for the least dimension possible, namely for d = 2. In fact our procedure
+easily generalises to any d ≥ 2, however we stick to surfaces as they furnish low-dimensional
+hence simple to describe representations.
+The existence of such surfaces follows from the characterisation of p-adic representations
+of G arising from abelian varieties with (tame) potential good reduction obtained in [Vo],
+and indeed provides an example of application of this result. In order to explicitely describe
+our objects we use Fontaine’s contravariant functor establishing an equivalence between
+crystalline p-adic representations of G and admissible filtered ϕ-modules. In section 1 we
+briefly review this theory as well as the characterisation in [Vo] (Theorem 1.2), and outline
+the general strategy.
+In sections 2 and 3 we construct two filtered ϕ-modules arising
+1
+
+2
+MAJA VOLKOV
+from abelian surfaces over Qp with good reduction that enjoy the required properties
+(Propositions 2.1 and 3.1).
+1. The general method
+Recall from [Fo2] that the objects D in the category MFQp(ϕ) of filtered ϕ-modules
+are finite dimensional Qp-vector spaces together with a Frobenius map ϕ ∈ AutQp(D) and
+a decreasing filtration Fil = (Fili D)i∈Z on D by subspaces such that Fili D = D for i ≪ 0
+and Fili D = 0 for i ≫ 0, and the morphisms are Qp-linear maps commuting with ϕ and
+preserving the filtration. The dual of (D, Fil) is the Qp-linear dual D∗ with ϕD∗ = ϕ∗ −1
+and Fili D∗ consists of linear forms on D vanishing on Filj D for all j > −i. The Tate twist
+D{−1} of (D, Fil) is D as a Qp-vector space with ϕD{−1} = pϕ and Fili D{−1} = Fili−1 D.
+The filtration Fil has Hodge-Tate type (0, 1) if Fili D = D for i ≤ 0, Fili D = 0 for i ≥ 2,
+and Fil1 D is a nontrivial subspace. The full subcategory MFad
+Qp(ϕ) of MFQp(ϕ) consists
+of objects (D, Fil) satisfying a property relating the Frobenius with the filtration, called
+admissibility and defined as follows. For a ϕ-stable sub-Qp-vector space D′ of D consider
+the Hodge and Newton invariants
+tH(D′) =
+def
+�
+i∈Z
+i dimQp
+�
+D′∩ FiliD
+�
+D′∩ Fili+1D
+�
+and
+tN(D′) =
+def vp(det ϕ|D′)
+where vp is the normalised p-adic valuation on Qp. Then (D, Fil) is admissible if
+(i) tH(D) = tN(D)
+(ii) tH(D′) ≤ tN(D′) for any sub-Qp[ϕ]-module D′ of D.
+A sub-Qp[ϕ]-module D′ endowed with the induced filtration Fili D′ = D′ ∩ Fili D is a
+subobject of (D, Fil) in MFad
+Qp(ϕ) if and only if tH(D′) = tN(D′).
+Let Bcris be the ring of p-adic periods constructed in [Fo1] and for a p-adic representation
+V of G put
+D*
+cris(V ) =
+def HomQp[G](V, Bcris).
+We always have dimQp D*
+cris(V ) ≤ dimQp V and V is said to be crystalline when equality
+holds. The functor V �→ D*
+cris(V ) establishes an anti-equivalence between the category of
+crystalline p-adic representations of G and MFad
+Qp(ϕ), a quasi-inverse being V*
+cris(D, Fil) =
+Homϕ,Fil(D, Bcris) ([Co-Fo]). These categories are well-suited to our problem since for an
+abelian variety A over Qp the G-module Vp(A) is crystalline if and only if A has good
+reduction ([Co-Io] Thm.4.7 or [Br] Cor.5.3.4.).
+A p-Weil number is an algebraic integer such that all its conjugates have absolute value
+√p in C. Call a monic polynomial in Z[X] a p-Weil polynomial if all its roots in Q are
+p-Weil numbers and its valuation at X2 − p is even. Consider the following conditions on
+a filtered ϕ-module (D, Fil) over Qp:
+(1) ϕ acts semisimply and Pchar(ϕ) is a p-Weil polynomial
+(2) the filtration has Hodge-Tate type (0, 1)
+(3) there exists a nondegenerate skew form on D under which ϕ is a p-similitude and
+Fil1 D is totally isotropic.
+
+SUPERSINGULAR ABELIAN SURFACES WITH NON SEMISIMPLE TATE MODULE
+3
+Recall that ϕ is a p-similitude under a bilinear form β if β(ϕx, ϕy) = pβ(x, y) for all
+x, y ∈ D and Fil1 D is totally isotropic if β(x, y) = 0 for all x, y ∈ Fil1 D. The map sending
+δ ∈ IsomQp(D∗, D) to β : (x, y) �→ δ−1(x)(y) identifies the antisymmetric isomorphisms
+of filtered ϕ-modules from D∗{−1} to D with the forms satisfying (3). A Qp-linear map
+δ : D∗ → D is an antisymmetric morphism in MFQp(ϕ) if δ∗ = −δ (under the canonical
+isomorphism D∗∗ ≃ D), ϕδ = pδϕ∗ −1, and δ(Fil1 D)⊥ ⊆ Fil1 D.
+Remark 1.1. Let Homa
+ϕ(D∗{−1}, D) be the Qp-vector space of antisymmetric ϕ-module
+morphisms from D∗{−1} to D and pick any δ ∈ Isoma
+ϕ(D∗{−1}, D). Then α† = δα∗δ−1
+defines an involution † on Endϕ(D) and the map α �→ αδ establishes an isomorphim
+Endϕ(D)† ∼
+−→ Homa
+ϕ(D∗{−1}, D) where Endϕ(D)† is the subspace of elements fixed by †.
+Theorem 1.2 ([Vo] Corollary 5.9). Let V be a crystalline p-adic representation of G. The
+following are equivalent:
+(i) there is an abelian variety A over Qp such that V ≃ Vp(A)
+(ii) D∗
+cris(V ) satisfies conditions (1), (2) and (3).
+Note that the restriction p ̸= 2 in [Vo] Theorem 5.7 and its Corollary 5.9 is unnecessary
+as Kisin shows that a crystalline representation with Hodge-Tate weights in {0, 1} arises
+from a p-divisible group unrestrictidly on the prime p ([Ki] Thm.0.3).
+Let A be an abelian variety over Qp having good reduction and (D, Fil) = D*
+cris(Vp(A)).
+The ϕ-module D satisfies (1) by the Weil conjectures for abelian varieties over Fp. Tate’s
+theorem on endomorphisms of the latter (see [Wa-Mi] II) shows that the isomorphism class
+of the ϕ-module D, given by semisimplicity by Pchar(ϕ), determines the isogeny class of
+the special fibre of A over Fp. Any polarisation on A induces a form on D satisfying (3)
+and the filtration satisfies (2) by the Hodge decomposition for p-divisible groups and (3).
+Conversely let V be a crystalline p-adic representation of G such that D*
+cris(V ) satisfies
+(1), (2), (3). From (1) the Honda-Tate theory ([Ho-Ta]) furnishes an abelian variety A over
+Fp with the right Frobenius. From (2) Kisin’s result [Ki] furnishes a p-divisible group over
+Zp lifting A(p). The Serre-Tate theory of liftings then produces a formal abelian scheme A
+over Zp with special fibre isogenous to A. Finally (3) furnishes a polarisation on A which
+ensures by Grothendieck’s theorem on algebraisation of formal schemes ([Gr] 5.4.5) that
+A is a true abelian scheme. The proof of Theorem 5.7 in [Vo] details this construction.
+Thus we want to construct an admissible filtered ϕ-module (D, Fil) over Qp satisfying
+conditions (1), (2), (3) of theorem 1.2 and such that
+(a) Pchar(ϕ) is a supersingular p-Weil polynomial
+(b) (D, Fil) is not semisimple.
+Recall that a p-Weil polynomial is supersingular if its roots are of the form ζ√p with
+ζ ∈ Q a root of unity, and that an abelian variety A over Fp is supersingular if and only if
+the characteristic polynomial of its Frobenius is supersingular. Regarding (a) in section 2
+we take Pchar(ϕ)(X) = (X2 + p)2 which is the characteristic polynomial of the Frobenius
+of the product of a supersingular elliptic curve E over Fp with itself. In section 3 we take
+Pchar(ϕ)(X) = X4 + pX2 + p2 which is the characteristic polynomial of the Frobenius of
+a simple supersingular abelian surface over Fp.
+
+4
+MAJA VOLKOV
+Regarding (b) we assume p ≡ 1 mod 3Z in section 3. In each (a)-case we find a subobject
+D1 of (D, Fil) in MFad
+Qp(ϕ) and a quotient object D2 (endowed with the quotient filtration
+Fili D2 = Fili D mod D1) such that the sequence
+(s)
+1
+� D1
+incl � D
+proj � D2
+� 1
+is exact and D2 is not a subobject. Thus (s) does not split and therefore (D, Fil) is not
+semisimple. Of course when (D, Fil) ≃ D*
+cris(Vp(A)) this means that there is a nonsplit
+short exact sequence of G-modules
+1
+� V2
+� Vp(A)
+� V1
+� 1
+with Vi ≃ V*
+cris(Di) for i = 1, 2, and it follows that Vp(A) is not a semisimple G-module.
+2. A lift of the twofold product of a supersingular elliptic curve
+Consider the filtered ϕ-module (D, Fil) over Qp defined as follows. There is a Qp-basis
+B = (x1, y1, x2, y2) for D so that
+D = Qpx1 ⊕ Qpy1 ⊕ Qpx2 ⊕ Qpy2
+is a 4-dimensional Qp-vector space. The matrix of ϕ over B is
+MatB(ϕ) =
+
+
+
+
+0
+−p
+0
+0
+1
+0
+0
+0
+0
+0
+0
+−p
+0
+0
+1
+0
+
+
+
+ ∈ GL4(Qp)
+and the filtration is given by
+Fil0 D = D,
+Fil1 D = Qpx1 ⊕ Qp(y1 + x2),
+Fil2 D = 0.
+Proposition 2.1. There is an abelian surface A over Qp such that (D, Fil) ≃ D∗
+cris(Vp(A)).
+Further
+(a) A has good reduction with special fibre isogenous to the product of two supersingular
+elliptic curves over Fp
+(b) the G-module Vp(A) is not semisimple.
+Proof. The filtration has Hodge-Tate type (0, 1) with dim Fil1 D = 2 and det ϕ = p2 hence
+tH(D) = 2 = tN(D). Since Pchar(ϕ)(X) = (X2 + p)2 the nontrivial ϕ-stable subspaces
+of D are the Di = Qpxi ⊕ Qpyi for i = 1, 2 both having Newton invariant tN(Di) = 1.
+However D1 ∩ Fil1 D = Qpx1 whereas D2 ∩ Fil1 D = 0, so tH(D1) = 1 and tH(D2) = 0.
+Therefore (D, Fil) is admissible, D1 is a subobject, D2 is a quotient that is not a subobject,
+the short exact sequence (s) does not split and (D, Fil) is not semisimple.
+The action of ϕ is semisimple and Pchar(ϕ) = Pchar(FrE)2 where E is a supersingular
+elliptic curve over Fp with Pchar(FrE)(X) = X2 +p. Thus (D, Fil) satisfies condition (1) of
+theorem 1.2 as well as condition (a) of section 1 and it obviously satisfies (2). It remains to
+check condition (3) that is to find a δ ∈ IsomQp(D∗, D) satisfying δ∗ = −δ, ϕδ = pδϕ∗−1,
+and δ(Fil1 D)⊥ = Fil1 D. Let B∗ = (x∗
+1, y∗
+1, x∗
+2, y∗
+2) be the dual basis of B for D∗ where z∗
+
+SUPERSINGULAR ABELIAN SURFACES WITH NON SEMISIMPLE TATE MODULE
+5
+is the linear form on D sending z ∈ D to 1 and vanishing on all vectors noncolinear to z.
+The matrix of pϕ∗ −1 over B∗ is
+p MatB(ϕ−1)t =
+
+
+
+
+0
+−1
+0
+0
+p
+0
+0
+0
+0
+0
+0
+−1
+0
+0
+p
+0
+
+
+
+
+where Mt is the transpose of M and
+(Fil1 D)⊥ = Qpy∗
+2 ⊕ Qp(y∗
+1 − x∗
+2).
+Let δ : D∗ → D be the Qp-linear morphism with matrix over the bases B∗ and B
+MatB∗,B(δ) =
+
+
+
+
+0
+0
+0
+−1
+0
+0
+1
+0
+0
+−1
+0
+0
+1
+0
+0
+0
+
+
+
+ .
+Then δ is invertible and satisfies the relations δ∗ = −δ and ϕδ = pδϕ∗ −1.
+Further
+δ(Fil1 D)⊥ = δ
+�
+Qpy∗
+2 ⊕ Qp(y∗
+1 − x∗
+2)
+�
+= Qpx1 ⊕ Qp(y1 + x2) = Fil1 D.
+□
+Remark 2.2. Any 2-dimensional object satisfying conditions (1) and (2) of theorem 1.2
+also satisfies condition (3). Hence theorem 1.2 applied to the admissible filtered ϕ-modules
+(D1, Fili D ∩ D1) and (D2, Fili D mod D1) shows the existence of elliptic schemes Ei over
+Zp such that Di ≃ D*
+cris(Vp(Ei)) for i = 1, 2. The special fibres of the Ei are Fp-isogenous
+to E. Thus we obtain a nonsplit exact sequence of G-modules
+1
+� Vp(E2)
+� Vp(A)
+� Vp(E1)
+� 1 .
+By Tate’s full faithfulness theorem [Ta] the G-module Vp(A) determines the p-divisible
+group A(p) over Zp up to isogeny, therefore A(p) is not Zp-isogenous to E1(p) × E2(p).
+Remark 2.3. The same construction works starting with the square of any supersingular
+p-Weil polynomial of degree two (when p ≥ 5 there is only X2 + p but when p = 2
+or 3 there are also the X2 ± pX + p). However it fails when dealing with the product
+of two distinct such. Indeed let α1 ̸= α2 ∈ pZp and D be a semisimple 4-dimensional
+ϕ-module with Pchar(ϕ)(X) = (X2 + α1X + p)(X2 + α2X + p). Then D = D1 ⊕ D2
+with Di = Ker(ϕ2 + αiϕ + p), which are the nontrivial ϕ-stable subspaces of D, and
+tN(Di) = 1. Since α1 ̸= α2 one checks that any Qp-linear δ : D∗ → D satisfying δ∗ = −δ
+and ϕδ = pδϕ∗ −1 sends D⊥
+2 into D1 and D⊥
+1 into D2. Endowing D with an admissible
+Hodge-Tate (0, 1) filtration such that (s) does not split amounts to picking a 2-dimensional
+subspace Fil1 D such that dim D1 ∩ Fil1 D = 1 and dim D2 ∩ Fil1 D = 0 (or vice versa) ;
+then dim D1 ∩ δ(Fil1 D)⊥ = 0 and dim D2 ∩ δ(Fil1 D)⊥ = 1, therefore δ(Fil1 D)⊥ ̸= Fil1 D.
+This shows that the p-adic Tate modules of abelian schemes over Zp with special fibre Fp-
+isogenous to the product of two nonisogenous supersingular elliptic curves are semisimple.
+Remark 2.4. One constructs in a similar fashion for each integer n ≥ 2 a lift of the n-fold
+product of a supersingular elliptic curve over Fp with nonsemisimple p-adic Tate module.
+
+6
+MAJA VOLKOV
+3. A lift of a simple supersingular abelian surface
+In this section we assume p ≡ 1 mod 3Z which is equivalent to ζ3 ∈ Qp where ζ3 is
+a primitive 3rd root of unity. Consider the filtered ϕ-module (D, Fil) defined as follows.
+There is a Qp-basis B = (x1, y1, x2, y2) for D so that
+D = Qpx1 ⊕ Qpy1 ⊕ Qpx2 ⊕ Qpy2
+is a 4-dimensional Qp-vector space. The matrix of ϕ over B is
+MatB(ϕ) =
+
+
+
+
+0
+ζ3p
+0
+0
+1
+0
+0
+0
+0
+0
+0
+ζ−1
+3 p
+0
+0
+1
+0
+
+
+
+ ∈ GL4(Qp)
+and the filtration is given by
+Fil0 D = D,
+Fil1 D = Qpx1 ⊕ Qp(y1 + x2),
+Fil2 D = 0.
+Proposition 3.1. There is an abelian surface A over Qp such that (D, Fil) ≃ D∗
+cris(Vp(A)).
+Further
+(a) A has good reduction with special fibre isogenous to a supersingular simple abelian
+surface over Fp
+(b) the G-module Vp(A) is not semisimple.
+Proof. Just as in the proof of proposition 2.1 we have tH(D) = 2 = tN(D). Since
+Pchar(ϕ)(X) = X4 + pX2 + p2 = (X2 − ζ3p)(X2 − ζ−1
+3 p)
+the nontrivial sub-Qp[ϕ]-modules of D are the Di = Qpxi ⊕ Qpyi for i = 1, 2 both having
+Newton invariant tN(Di) = 1, and Hodge invariants tH(D1) = 1, tH(D2) = 0. Again we
+obtain a nonsplit exact sequence (s) in MFad
+Qp(ϕ) and (D, Fil) is not semisimple.
+The action of ϕ is semisimple and Pchar(ϕ) = Pchar(FrA) where A is a supersingular
+simple abelian surface over Fp with Pchar(FrA)(X) = X4 +pX2 +p2. Thus (D, Fil) satisfies
+condition (1) of theorem 1.2 as well as condition (a) of section 1. It obviously satisfies (2)
+and it remains to check (3). Let B∗ = (x∗
+1, y∗
+1, x∗
+2, y∗
+2) be the dual basis of B for D∗. Again
+(Fil1 D)⊥ = Qpy∗
+2 ⊕ Qp(y∗
+1 − x∗
+2) and the matrix of pϕ∗ −1 over B∗ is
+p MatB(ϕ−1)t =
+
+
+
+
+0
+ζ−1
+3
+0
+0
+p
+0
+0
+0
+0
+0
+0
+ζ3
+0
+0
+p
+0
+
+
+
+ .
+Let δ : D∗ → D be the Qp-linear morphism with matrix over the bases B∗ and B
+MatB∗,B(δ) =
+
+
+
+
+0
+0
+0
+ζ3
+0
+0
+1
+0
+0
+−1
+0
+0
+−ζ3
+0
+0
+0
+
+
+
+ .
+As in the proof of proposition 2.1 one checks that δ is invertible, satisfies δ∗ = −δ,
+ϕδ = pδϕ∗ −1, and that δ(Fil1 D)⊥ = Fil1 D.
+□
+
+SUPERSINGULAR ABELIAN SURFACES WITH NON SEMISIMPLE TATE MODULE
+7
+Remark 3.2. The objects (D1, Fili D ∩ D1) and (D2, Fili D mod D1) in MFad
+Qp(ϕ) do not
+arise from elliptic schemes over Zp, however [Ki] Thm.0.3 shows the existence of p-divisible
+groups Gi over Zp such that Di ≃ D*
+cris(Vp(Gi)). The special fibre of A(p) is Fp-isogenous
+to the product of the special fibres of the Gi, themselves being nonisogenous. Thus we
+obtain a nonsplit exact sequence of G-modules
+1
+� Vp(G2)
+� Vp(A)
+� Vp(G1)
+� 1
+and Tate’s full faithfulness theorem shows that A(p) is not Zp-isogenous to G1 × G2.
+Remark 3.3. Starting with X4 − pX2 + p2 when p ≡ 1 mod 3Z and X4 + p2 when
+p ≡ 1 mod 4Z one obtains alike nonsemisimple 4-dimensional supersingular representations
+(just replace ζ3 by ζ6 or ζ4). More generally the
+pdΦn
+�X2
+p
+�
+=
+�
+i∈(Z/nZ)×
+(X2 − ζi
+np)
+with d = #
+�
+Z/nZ
+�× ≥ 2
+where Φn is the nth cyclotomic polynomial are supersingular p-Weil polynomials leading
+when p ≡ 1 mod nZ to similar higher-dimensional constructions.
+References
+[Br]
+C. Breuil, Groupes p-divisibles, groupes finis et modules filtr´es, Annals of Math. 152 (2000),
+489-549.
+[Co-Io]
+R. Coleman and A. Iovita, The Frobenius and Monodromy operators for Curves and Abelian
+Varieties, Duke Math. J. 97 (1999), 171-215.
+[Co-Fo]
+P. Colmez et J.-M. Fontaine, Construction des repr´esentations p-adiques semi-stables, Invent.
+math. 140, 1 (2000), 1-43.
+[Fo1]
+J.-M. Fontaine, Le corps des p´eriodes p-adiques, in P´eriodes p-adiques, Ast´erisque 223, Soc.
+Math. de France (1994).
+[Fo2]
+J.-M. Fontaine, Repr´esentations p-adiques semi-stables, in P´eriodes p-adiques, Ast´erisque 223,
+Soc. Math. de France (1994).
+[Gr]
+A. Grothendieck, EGA III, Inst. Hautes ´Etudes Sci. Publ. Math. 11 (1961).
+[Ho-Ta] J. Tate, Classes d’isog´enie des vari´et´es ab´eliennes sur un corps fini (d’apr`es T. Honda), S´eminaire
+Bourbaki 352 (1968), 15p.
+[Ki]
+M. Kisin, Crystalline representations and F-crystals, in Algebraic Geometry and Number Theory
+In Honor of Vladimir Drinfeld’s 50th Birthday, Progress in Mathematics 253 (2006), 459-496.
+[Ta]
+J. Tate, p-Divisible groups over local fields, in Proceedings of a Conference on Local Fields,
+Driebergen 1966, Springer-Verlag (1967), 158-183.
+[Vo]
+M. Volkov, A class of p-adic Galois representations arising from abelian varieties over Qp, Math.
+Ann. 331 (2005), no. 4, 889–923.
+[Wa-Mi] W.C. Waterhouse and J.S. Milne, Abelian Varieties over Finite Fields, in AMS Proceedings of
+Symposia in Pure Mathematics XX (1971), 53-64.
+Universit´e de Mons, D´epartement de Math´ematique, Place du Parc 20, B-7000 Mons, Bel-
+gium.
+Email address: maja.volkov@umons.ac.be
+
diff --git a/I9E5T4oBgHgl3EQfXA_-/content/tmp_files/load_file.txt b/I9E5T4oBgHgl3EQfXA_-/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..213740518a341a97e31bd344a8b9bf9c33c20c0c
--- /dev/null
+++ b/I9E5T4oBgHgl3EQfXA_-/content/tmp_files/load_file.txt
@@ -0,0 +1,215 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf,len=214
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='05564v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='NT] 13 Jan 2023 ABELIAN SURFACES WITH SUPERSINGULAR GOOD REDUCTION AND NON SEMISIMPLE TATE MODULE MAJA VOLKOV Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' We show the existence of abelian surfaces A over Qp having good reduction with supersingular special fibre whose associated p-adic Galois module Vp(A) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 2000 Mathematics Subject Classification: 11G10, 14K15, 14G20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Keywords: Abelian varieties, local fields, Galois representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Contents Introduction 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The general method 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A lift of the twofold product of a supersingular elliptic curve 4 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A lift of a simple supersingular abelian surface 6 References 7 Introduction Fix a prime number p and an algebraic closure Qp of Qp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Write G = Gal(Qp/Qp) for the absolute Galois group of Qp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' For a d-dimensional abelian variety A over Qp let A[pn] be the group of pn-torsion points with values in Qp and Vp(A) = def Qp ⊗Zp lim ←− n≥1 A[pn].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' This is a 2d-dimensional Qp-vector space on which G acts linearly and continuously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' We want to consider the following problem: find abelian varieties A over Qp having good reduction with supersingular special fibre and such that the Galois module Vp(A) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In this paper we show the existence of two such varieties with nonisogenous special fibres for the least dimension possible, namely for d = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In fact our procedure easily generalises to any d ≥ 2, however we stick to surfaces as they furnish low-dimensional hence simple to describe representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The existence of such surfaces follows from the characterisation of p-adic representations of G arising from abelian varieties with (tame) potential good reduction obtained in [Vo], and indeed provides an example of application of this result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In order to explicitely describe our objects we use Fontaine’s contravariant functor establishing an equivalence between crystalline p-adic representations of G and admissible filtered ϕ-modules.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In section 1 we briefly review this theory as well as the characterisation in [Vo] (Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2), and outline the general strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In sections 2 and 3 we construct two filtered ϕ-modules arising 1 2 MAJA VOLKOV from abelian surfaces over Qp with good reduction that enjoy the required properties (Propositions 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The general method Recall from [Fo2] that the objects D in the category MFQp(ϕ) of filtered ϕ-modules are finite dimensional Qp-vector spaces together with a Frobenius map ϕ ∈ AutQp(D) and a decreasing filtration Fil = (Fili D)i∈Z on D by subspaces such that Fili D = D for i ≪ 0 and Fili D = 0 for i ≫ 0, and the morphisms are Qp-linear maps commuting with ϕ and preserving the filtration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The dual of (D, Fil) is the Qp-linear dual D∗ with ϕD∗ = ϕ∗ −1 and Fili D∗ consists of linear forms on D vanishing on Filj D for all j > −i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The Tate twist D{−1} of (D, Fil) is D as a Qp-vector space with ϕD{−1} = pϕ and Fili D{−1} = Fili−1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The filtration Fil has Hodge-Tate type (0, 1) if Fili D = D for i ≤ 0, Fili D = 0 for i ≥ 2, and Fil1 D is a nontrivial subspace.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The full subcategory MFad Qp(ϕ) of MFQp(ϕ) consists of objects (D, Fil) satisfying a property relating the Frobenius with the filtration, called admissibility and defined as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' For a ϕ-stable sub-Qp-vector space D′ of D consider the Hodge and Newton invariants tH(D′) = def � i∈Z i dimQp � D′∩ FiliD � D′∩ Fili+1D � and tN(D′) = def vp(det ϕ|D′) where vp is the normalised p-adic valuation on Qp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Then (D, Fil) is admissible if (i) tH(D) = tN(D) (ii) tH(D′) ≤ tN(D′) for any sub-Qp[ϕ]-module D′ of D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A sub-Qp[ϕ]-module D′ endowed with the induced filtration Fili D′ = D′ ∩ Fili D is a subobject of (D, Fil) in MFad Qp(ϕ) if and only if tH(D′) = tN(D′).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let Bcris be the ring of p-adic periods constructed in [Fo1] and for a p-adic representation V of G put D* cris(V ) = def HomQp[G](V, Bcris).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' We always have dimQp D* cris(V ) ≤ dimQp V and V is said to be crystalline when equality holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The functor V �→ D* cris(V ) establishes an anti-equivalence between the category of crystalline p-adic representations of G and MFad Qp(ϕ), a quasi-inverse being V* cris(D, Fil) = Homϕ,Fil(D, Bcris) ([Co-Fo]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' These categories are well-suited to our problem since for an abelian variety A over Qp the G-module Vp(A) is crystalline if and only if A has good reduction ([Co-Io] Thm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='7 or [Br] Cor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A p-Weil number is an algebraic integer such that all its conjugates have absolute value √p in C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Call a monic polynomial in Z[X] a p-Weil polynomial if all its roots in Q are p-Weil numbers and its valuation at X2 − p is even.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Consider the following conditions on a filtered ϕ-module (D, Fil) over Qp: (1) ϕ acts semisimply and Pchar(ϕ) is a p-Weil polynomial (2) the filtration has Hodge-Tate type (0, 1) (3) there exists a nondegenerate skew form on D under which ϕ is a p-similitude and Fil1 D is totally isotropic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' SUPERSINGULAR ABELIAN SURFACES WITH NON SEMISIMPLE TATE MODULE 3 Recall that ϕ is a p-similitude under a bilinear form β if β(ϕx, ϕy) = pβ(x, y) for all x, y ∈ D and Fil1 D is totally isotropic if β(x, y) = 0 for all x, y ∈ Fil1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The map sending δ ∈ IsomQp(D∗, D) to β : (x, y) �→ δ−1(x)(y) identifies the antisymmetric isomorphisms of filtered ϕ-modules from D∗{−1} to D with the forms satisfying (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A Qp-linear map δ : D∗ → D is an antisymmetric morphism in MFQp(ϕ) if δ∗ = −δ (under the canonical isomorphism D∗∗ ≃ D), ϕδ = pδϕ∗ −1, and δ(Fil1 D)⊥ ⊆ Fil1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Remark 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let Homa ϕ(D∗{−1}, D) be the Qp-vector space of antisymmetric ϕ-module morphisms from D∗{−1} to D and pick any δ ∈ Isoma ϕ(D∗{−1}, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Then α† = δα∗δ−1 defines an involution † on Endϕ(D) and the map α �→ αδ establishes an isomorphim Endϕ(D)† ∼ −→ Homa ϕ(D∗{−1}, D) where Endϕ(D)† is the subspace of elements fixed by †.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2 ([Vo] Corollary 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let V be a crystalline p-adic representation of G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The following are equivalent: (i) there is an abelian variety A over Qp such that V ≃ Vp(A) (ii) D∗ cris(V ) satisfies conditions (1), (2) and (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Note that the restriction p ̸= 2 in [Vo] Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='7 and its Corollary 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='9 is unnecessary as Kisin shows that a crystalline representation with Hodge-Tate weights in {0, 1} arises from a p-divisible group unrestrictidly on the prime p ([Ki] Thm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let A be an abelian variety over Qp having good reduction and (D, Fil) = D* cris(Vp(A)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The ϕ-module D satisfies (1) by the Weil conjectures for abelian varieties over Fp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Tate’s theorem on endomorphisms of the latter (see [Wa-Mi] II) shows that the isomorphism class of the ϕ-module D, given by semisimplicity by Pchar(ϕ), determines the isogeny class of the special fibre of A over Fp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Any polarisation on A induces a form on D satisfying (3) and the filtration satisfies (2) by the Hodge decomposition for p-divisible groups and (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Conversely let V be a crystalline p-adic representation of G such that D* cris(V ) satisfies (1), (2), (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' From (1) the Honda-Tate theory ([Ho-Ta]) furnishes an abelian variety A over Fp with the right Frobenius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' From (2) Kisin’s result [Ki] furnishes a p-divisible group over Zp lifting A(p).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The Serre-Tate theory of liftings then produces a formal abelian scheme A over Zp with special fibre isogenous to A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Finally (3) furnishes a polarisation on A which ensures by Grothendieck’s theorem on algebraisation of formal schemes ([Gr] 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='5) that A is a true abelian scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The proof of Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='7 in [Vo] details this construction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Thus we want to construct an admissible filtered ϕ-module (D, Fil) over Qp satisfying conditions (1), (2), (3) of theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2 and such that (a) Pchar(ϕ) is a supersingular p-Weil polynomial (b) (D, Fil) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Recall that a p-Weil polynomial is supersingular if its roots are of the form ζ√p with ζ ∈ Q a root of unity, and that an abelian variety A over Fp is supersingular if and only if the characteristic polynomial of its Frobenius is supersingular.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Regarding (a) in section 2 we take Pchar(ϕ)(X) = (X2 + p)2 which is the characteristic polynomial of the Frobenius of the product of a supersingular elliptic curve E over Fp with itself.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In section 3 we take Pchar(ϕ)(X) = X4 + pX2 + p2 which is the characteristic polynomial of the Frobenius of a simple supersingular abelian surface over Fp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 4 MAJA VOLKOV Regarding (b) we assume p ≡ 1 mod 3Z in section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' In each (a)-case we find a subobject D1 of (D, Fil) in MFad Qp(ϕ) and a quotient object D2 (endowed with the quotient filtration Fili D2 = Fili D mod D1) such that the sequence (s) 1 � D1 incl � D proj � D2 � 1 is exact and D2 is not a subobject.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Thus (s) does not split and therefore (D, Fil) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Of course when (D, Fil) ≃ D* cris(Vp(A)) this means that there is a nonsplit short exact sequence of G-modules 1 � V2 � Vp(A) � V1 � 1 with Vi ≃ V* cris(Di) for i = 1, 2, and it follows that Vp(A) is not a semisimple G-module.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A lift of the twofold product of a supersingular elliptic curve Consider the filtered ϕ-module (D, Fil) over Qp defined as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' There is a Qp-basis B = (x1, y1, x2, y2) for D so that D = Qpx1 ⊕ Qpy1 ⊕ Qpx2 ⊕ Qpy2 is a 4-dimensional Qp-vector space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The matrix of ϕ over B is MatB(ϕ) = \uf8eb \uf8ec \uf8ec \uf8ed 0 −p 0 0 1 0 0 0 0 0 0 −p 0 0 1 0 \uf8f6 \uf8f7 \uf8f7 \uf8f8 ∈ GL4(Qp) and the filtration is given by Fil0 D = D, Fil1 D = Qpx1 ⊕ Qp(y1 + x2), Fil2 D = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' There is an abelian surface A over Qp such that (D, Fil) ≃ D∗ cris(Vp(A)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Further (a) A has good reduction with special fibre isogenous to the product of two supersingular elliptic curves over Fp (b) the G-module Vp(A) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The filtration has Hodge-Tate type (0, 1) with dim Fil1 D = 2 and det ϕ = p2 hence tH(D) = 2 = tN(D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Since Pchar(ϕ)(X) = (X2 + p)2 the nontrivial ϕ-stable subspaces of D are the Di = Qpxi ⊕ Qpyi for i = 1, 2 both having Newton invariant tN(Di) = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' However D1 ∩ Fil1 D = Qpx1 whereas D2 ∩ Fil1 D = 0, so tH(D1) = 1 and tH(D2) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Therefore (D, Fil) is admissible, D1 is a subobject, D2 is a quotient that is not a subobject, the short exact sequence (s) does not split and (D, Fil) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The action of ϕ is semisimple and Pchar(ϕ) = Pchar(FrE)2 where E is a supersingular elliptic curve over Fp with Pchar(FrE)(X) = X2 +p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Thus (D, Fil) satisfies condition (1) of theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2 as well as condition (a) of section 1 and it obviously satisfies (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' It remains to check condition (3) that is to find a δ ∈ IsomQp(D∗, D) satisfying δ∗ = −δ, ϕδ = pδϕ∗−1, and δ(Fil1 D)⊥ = Fil1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let B∗ = (x∗ 1, y∗ 1, x∗ 2, y∗ 2) be the dual basis of B for D∗ where z∗ SUPERSINGULAR ABELIAN SURFACES WITH NON SEMISIMPLE TATE MODULE 5 is the linear form on D sending z ∈ D to 1 and vanishing on all vectors noncolinear to z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The matrix of pϕ∗ −1 over B∗ is p MatB(ϕ−1)t = \uf8eb \uf8ec \uf8ec \uf8ed 0 −1 0 0 p 0 0 0 0 0 0 −1 0 0 p 0 \uf8f6 \uf8f7 \uf8f7 \uf8f8 where Mt is the transpose of M and (Fil1 D)⊥ = Qpy∗ 2 ⊕ Qp(y∗ 1 − x∗ 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let δ : D∗ → D be the Qp-linear morphism with matrix over the bases B∗ and B MatB∗,B(δ) = \uf8eb \uf8ec \uf8ec \uf8ed 0 0 0 −1 0 0 1 0 0 −1 0 0 1 0 0 0 \uf8f6 \uf8f7 \uf8f7 \uf8f8 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Then δ is invertible and satisfies the relations δ∗ = −δ and ϕδ = pδϕ∗ −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Further δ(Fil1 D)⊥ = δ � Qpy∗ 2 ⊕ Qp(y∗ 1 − x∗ 2) � = Qpx1 ⊕ Qp(y1 + x2) = Fil1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' □ Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Any 2-dimensional object satisfying conditions (1) and (2) of theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2 also satisfies condition (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Hence theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2 applied to the admissible filtered ϕ-modules (D1, Fili D ∩ D1) and (D2, Fili D mod D1) shows the existence of elliptic schemes Ei over Zp such that Di ≃ D* cris(Vp(Ei)) for i = 1, 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The special fibres of the Ei are Fp-isogenous to E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Thus we obtain a nonsplit exact sequence of G-modules 1 � Vp(E2) � Vp(A) � Vp(E1) � 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' By Tate’s full faithfulness theorem [Ta] the G-module Vp(A) determines the p-divisible group A(p) over Zp up to isogeny, therefore A(p) is not Zp-isogenous to E1(p) × E2(p).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The same construction works starting with the square of any supersingular p-Weil polynomial of degree two (when p ≥ 5 there is only X2 + p but when p = 2 or 3 there are also the X2 ± pX + p).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' However it fails when dealing with the product of two distinct such.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Indeed let α1 ̸= α2 ∈ pZp and D be a semisimple 4-dimensional ϕ-module with Pchar(ϕ)(X) = (X2 + α1X + p)(X2 + α2X + p).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Then D = D1 ⊕ D2 with Di = Ker(ϕ2 + αiϕ + p), which are the nontrivial ϕ-stable subspaces of D, and tN(Di) = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Since α1 ̸= α2 one checks that any Qp-linear δ : D∗ → D satisfying δ∗ = −δ and ϕδ = pδϕ∗ −1 sends D⊥ 2 into D1 and D⊥ 1 into D2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Endowing D with an admissible Hodge-Tate (0, 1) filtration such that (s) does not split amounts to picking a 2-dimensional subspace Fil1 D such that dim D1 ∩ Fil1 D = 1 and dim D2 ∩ Fil1 D = 0 (or vice versa) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' then dim D1 ∩ δ(Fil1 D)⊥ = 0 and dim D2 ∩ δ(Fil1 D)⊥ = 1, therefore δ(Fil1 D)⊥ ̸= Fil1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' This shows that the p-adic Tate modules of abelian schemes over Zp with special fibre Fp- isogenous to the product of two nonisogenous supersingular elliptic curves are semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' One constructs in a similar fashion for each integer n ≥ 2 a lift of the n-fold product of a supersingular elliptic curve over Fp with nonsemisimple p-adic Tate module.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 6 MAJA VOLKOV 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' A lift of a simple supersingular abelian surface In this section we assume p ≡ 1 mod 3Z which is equivalent to ζ3 ∈ Qp where ζ3 is a primitive 3rd root of unity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Consider the filtered ϕ-module (D, Fil) defined as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' There is a Qp-basis B = (x1, y1, x2, y2) for D so that D = Qpx1 ⊕ Qpy1 ⊕ Qpx2 ⊕ Qpy2 is a 4-dimensional Qp-vector space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The matrix of ϕ over B is MatB(ϕ) = \uf8eb \uf8ec \uf8ec \uf8ed 0 ζ3p 0 0 1 0 0 0 0 0 0 ζ−1 3 p 0 0 1 0 \uf8f6 \uf8f7 \uf8f7 \uf8f8 ∈ GL4(Qp) and the filtration is given by Fil0 D = D, Fil1 D = Qpx1 ⊕ Qp(y1 + x2), Fil2 D = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Proposition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' There is an abelian surface A over Qp such that (D, Fil) ≃ D∗ cris(Vp(A)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Further (a) A has good reduction with special fibre isogenous to a supersingular simple abelian surface over Fp (b) the G-module Vp(A) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Just as in the proof of proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1 we have tH(D) = 2 = tN(D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Since Pchar(ϕ)(X) = X4 + pX2 + p2 = (X2 − ζ3p)(X2 − ζ−1 3 p) the nontrivial sub-Qp[ϕ]-modules of D are the Di = Qpxi ⊕ Qpyi for i = 1, 2 both having Newton invariant tN(Di) = 1, and Hodge invariants tH(D1) = 1, tH(D2) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Again we obtain a nonsplit exact sequence (s) in MFad Qp(ϕ) and (D, Fil) is not semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The action of ϕ is semisimple and Pchar(ϕ) = Pchar(FrA) where A is a supersingular simple abelian surface over Fp with Pchar(FrA)(X) = X4 +pX2 +p2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Thus (D, Fil) satisfies condition (1) of theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2 as well as condition (a) of section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' It obviously satisfies (2) and it remains to check (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let B∗ = (x∗ 1, y∗ 1, x∗ 2, y∗ 2) be the dual basis of B for D∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Again (Fil1 D)⊥ = Qpy∗ 2 ⊕ Qp(y∗ 1 − x∗ 2) and the matrix of pϕ∗ −1 over B∗ is p MatB(ϕ−1)t = \uf8eb \uf8ec \uf8ec \uf8ed 0 ζ−1 3 0 0 p 0 0 0 0 0 0 ζ3 0 0 p 0 \uf8f6 \uf8f7 \uf8f7 \uf8f8 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Let δ : D∗ → D be the Qp-linear morphism with matrix over the bases B∗ and B MatB∗,B(δ) = \uf8eb \uf8ec \uf8ec \uf8ed 0 0 0 ζ3 0 0 1 0 0 −1 0 0 −ζ3 0 0 0 \uf8f6 \uf8f7 \uf8f7 \uf8f8 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' As in the proof of proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='1 one checks that δ is invertible, satisfies δ∗ = −δ, ϕδ = pδϕ∗ −1, and that δ(Fil1 D)⊥ = Fil1 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' □ SUPERSINGULAR ABELIAN SURFACES WITH NON SEMISIMPLE TATE MODULE 7 Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The objects (D1, Fili D ∩ D1) and (D2, Fili D mod D1) in MFad Qp(ϕ) do not arise from elliptic schemes over Zp, however [Ki] Thm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='3 shows the existence of p-divisible groups Gi over Zp such that Di ≃ D* cris(Vp(Gi)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' The special fibre of A(p) is Fp-isogenous to the product of the special fibres of the Gi, themselves being nonisogenous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Thus we obtain a nonsplit exact sequence of G-modules 1 � Vp(G2) � Vp(A) � Vp(G1) � 1 and Tate’s full faithfulness theorem shows that A(p) is not Zp-isogenous to G1 × G2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Starting with X4 − pX2 + p2 when p ≡ 1 mod 3Z and X4 + p2 when p ≡ 1 mod 4Z one obtains alike nonsemisimple 4-dimensional supersingular representations (just replace ζ3 by ζ6 or ζ4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' More generally the pdΦn �X2 p � = � i∈(Z/nZ)× (X2 − ζi np) with d = # � Z/nZ �× ≥ 2 where Φn is the nth cyclotomic polynomial are supersingular p-Weil polynomials leading when p ≡ 1 mod nZ to similar higher-dimensional constructions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' References [Br] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Breuil, Groupes p-divisibles, groupes finis et modules filtr´es, Annals of Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 152 (2000), 489-549.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Co-Io] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Coleman and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Iovita, The Frobenius and Monodromy operators for Curves and Abelian Varieties, Duke Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 97 (1999), 171-215.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Co-Fo] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Colmez et J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Fontaine, Construction des repr´esentations p-adiques semi-stables, Invent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 140, 1 (2000), 1-43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Fo1] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Fontaine, Le corps des p´eriodes p-adiques, in P´eriodes p-adiques, Ast´erisque 223, Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' de France (1994).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Fo2] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Fontaine, Repr´esentations p-adiques semi-stables, in P´eriodes p-adiques, Ast´erisque 223, Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' de France (1994).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Gr] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Grothendieck, EGA III, Inst.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Hautes ´Etudes Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Publ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 11 (1961).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Ho-Ta] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Tate, Classes d’isog´enie des vari´et´es ab´eliennes sur un corps fini (d’apr`es T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Honda), S´eminaire Bourbaki 352 (1968), 15p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Ki] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Kisin, Crystalline representations and F-crystals, in Algebraic Geometry and Number Theory In Honor of Vladimir Drinfeld’s 50th Birthday, Progress in Mathematics 253 (2006), 459-496.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Ta] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Tate, p-Divisible groups over local fields, in Proceedings of a Conference on Local Fields, Driebergen 1966, Springer-Verlag (1967), 158-183.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Vo] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Volkov, A class of p-adic Galois representations arising from abelian varieties over Qp, Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Ann.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 331 (2005), no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' 4, 889–923.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' [Wa-Mi] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Waterhouse and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Milne, Abelian Varieties over Finite Fields, in AMS Proceedings of Symposia in Pure Mathematics XX (1971), 53-64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Universit´e de Mons, D´epartement de Math´ematique, Place du Parc 20, B-7000 Mons, Bel- gium.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content=' Email address: maja.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='volkov@umons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
+page_content='be' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/I9E5T4oBgHgl3EQfXA_-/content/2301.05564v1.pdf'}
diff --git a/J9AzT4oBgHgl3EQfVfw4/content/tmp_files/2301.01284v1.pdf.txt b/J9AzT4oBgHgl3EQfVfw4/content/tmp_files/2301.01284v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8a46d423d5d76c1b01d72330f9e101611825e617
--- /dev/null
+++ b/J9AzT4oBgHgl3EQfVfw4/content/tmp_files/2301.01284v1.pdf.txt
@@ -0,0 +1,1197 @@
+Draft version January 4, 2023
+Typeset using LATEX twocolumn style in AASTeX631
+Polar circumtriple planets and disks can only form close to a triple star
+Stephen Lepp,1, 2 Rebecca G. Martin,1, 2 and Stephen H. Lubow3
+1Nevada Center for Astrophysics, University of Nevada, Las Vegas, 4505 S. Maryland Pkwy., Las Vegas, NV 89154, USA
+2Department of Physics and Astronomy,University of Nevada, Las Vegas, 4505 S. Maryland Pkwy., Las Vegas, NV 89154, USA
+3Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA
+ABSTRACT
+Observations of protoplanetary disks around binary and triple star systems suggest that misalign-
+ments between the orbital plane of the stars and the disks are common. Motivated by recent observa-
+tions of polar circumbinary disks, we explore the possibility for polar circumtriple disks and therefore
+polar circumtriple planets that could form in such a disk. With n-body simulations and analytic meth-
+ods we find that the inclusion of the third star, and the associated apsidal precession, significantly
+reduces the radial range of polar orbits so that circumtriple polar disks and planets can only be found
+close to the stellar system. Outside of a critical radius, that is typically in the range of 3−10 times the
+outer binary separation depending upon the binary parameters, the orbits behave the same as they do
+around a circular orbit binary.
+For some observed systems that have shorter period inner binaries,
+the critical radius is considerably larger. If polar circumtriple planets can form, we suggest that it is
+likely that they form in a disk that was subject to breaking.
+Keywords: Binary stars (154) — Celestial mechanics (211) — Planet formation (1241)
+1. INTRODUCTION
+Multiple stellar systems are common in star forming
+regions (Duchˆene & Kraus 2013). Disks around triple
+star systems are also expected to be common (Tobin
+et al. 2016; Bate 2018) and there are several well known
+examples including GG Tauri A (Di Folco et al. 2014;
+Keppler et al. 2020; Phuong et al. 2020a) and GW Ori
+(Bi et al. 2020; Kraus et al. 2020; Smallwood et al. 2021).
+A common feature of these disks is that they are tilted
+with respect to the orbital plane of the stars. Disk mis-
+alignment may initially occur, for example, because of
+turbulence in the molecular gas cloud (Offner et al. 2010;
+Tokuda et al. 2014; Bate 2012) or later accretion of ma-
+terial by the young binary (Bate et al. 2010; Bate 2018).
+Misalignment may be increased later by stellar flybys
+(Nealon et al. 2020) or bound stellar companions (e.g.
+Martin & Lubow 2017; Martin et al. 2022).
+Around an eccentric binary star system, test particle
+orbits have two stable stationary states: coplanar align-
+ment to the binary orbit, and polar alignment in which
+the angular momentum of the particle orbit is aligned
+to the binary eccentricity vector and 90◦ to the binary
+orbital plane (Verrier & Evans 2009; Farago & Laskar
+2010; Doolin & Blundell 2011; Chen et al. 2019). A par-
+ticle that is misaligned from one of these two stationary
+states undergoes nodal precession. Low initial inclina-
+tion orbits precess about the binary angular momentum
+vector while high initial inclination orbits precess about
+the binary eccentricity vector.
+Since the test particle
+does not affect the dynamics of the binary, the qualita-
+tive behaviour does not depend on orbital radius of the
+particle around the binary unless general relativity or
+tides become important (Lepp et al. 2022).
+A circumbinary disk with a low-mass can undergo sim-
+ilar dynamical behaviour to a test particle (e.g Aly et al.
+2015; Martin & Lubow 2018). If the disk is in good ra-
+dial communication, it can undergo solid body preces-
+sion at a
+angular momentum weighted average
+rate
+(Papaloizou & Terquem 1995; Larwood et al. 1996).
+For protoplanetary disks, the radial communication is
+wave-like (Papaloizou & Pringle 1983; Lubow & Ogilvie
+2001). Dissipation in the disk leads to alignment either
+towards coplanar (Nixon et al. 2013; Facchini et al. 2013)
+or polar depending on the initial tilt (Martin & Lubow
+2017; Lubow & Martin 2018; Zanazzi & Lai 2018; Cuello
+& Giuppone 2019).
+Several polar circumbinary disks
+around eccentric binaries have been observed (Kennedy
+et al. 2012, 2019; Kenworthy et al. 2022), although none
+have yet been observed around a triple star. While polar
+circumbinary planets have not yet been observed, their
+formation may be as efficient as in a coplanar configu-
+ration (Childs & Martin 2021a,b).
+arXiv:2301.01284v1 [astro-ph.EP] 3 Jan 2023
+
+2
+While the evolution of circumbinary particles and
+disks is now fairly well understood, the inclusion of an in-
+ner hierarchical triple star system has not been explored
+in detail. In this work, for the first time, we examine the
+effect of an inner triple star system on the existence of
+polar orbits. In Section 2 we use n-body simulations and
+in Section 3 we compare to analytic models. The inner
+and outer binaries that compose the triple star undergo
+apsidal precession. We show that this can remove the
+possibility of polar orbits outside of a critical radius from
+the triple star. This is similar to the effects of general
+relativity that also causes apsidal precession of the bi-
+nary (Lepp et al. 2022) but with much higher precession
+rates. In Section 4 we draw our conclusions and discuss
+implications both for observations of circumtriple disks
+and for the properties of planets that may form in such
+disks.
+2. CIRCUMTRIPLE PARTICLE ORBITS
+In this Section we first consider the dynamics of a par-
+ticle orbiting a triple star with our standard parameters
+and then we consider the effect of varying different triple
+star parameters. We use the REBOUND N-body code
+(Rein & Liu 2012). The simulations were integrated us-
+ing a combination of IAS15, a 15th order Gauss-Radau
+integrator (Rein & Spiegel 2015) and the WHFast, a
+symplectic Wisdom-Holman integrator (Rein & Tamayo
+2015; Wisdom & Holman 1991).
+2.1. Triple star parameters
+Triple star systems are found to occur with a large
+range of properties. Figure 3 of Tokovinin (2021) plots
+the outer binary period as a function of inner binary pe-
+riod for a sample of 1820 systems that lie within a dis-
+tance of 200pc. The sample is subject to strong selection
+effects that favor the detection of close spectroscopic bi-
+naries and resolved wide binaries. Nonetheless, the plot
+suggests that for longer period inner binaries (> 1y),
+the outer to inner semi-major axis ratios typically range
+from 3 to 50. For shorter period inner binaries (< 1y),
+the ratio range is typically from 20 to 100, and ratios of
+greater than 1000 also occur.
+Triple star systems may be unstable for a wide range
+of parameter space (Mardling & Aarseth 2001; Valtonen
+& Karttunen 2006; Vynatheya et al. 2022).
+We con-
+sider hierarchical triple systems composed of an inner
+binary with an outer binary companion. The inclina-
+tion of the inner binary to the inclination of the binary
+companion must be small enough to avoid von Zeipel-
+Kozai-Lidov (ZKL) oscillations (von Zeipel 1910; Kozai
+1962; Lidov 1962; Naoz 2016; Hamers 2021). Figure 3
+of Tokovinin (2021) shows evidence of the stability limit
+at a period ratio of about 4.7 predicted by Mardling &
+Aarseth (2001).
+For the range of parameters studied,
+the eccentricity of all the particle orbits are relatively
+constant.
+The orbits are scale free in mass and length and we
+adopt as our mass unit, the total mass of the triple
+star system, mAB, and for our length unit, the semi-
+major axis of the outer companion, aAB. For our stan-
+dard parameters, the inner binary has a total mass
+mA and is composed of an equal mass binary with
+mAa = mAb = 0.25 mAB, semi-major axis aA = aAB/20
+and eccentricity eA = 0 and an inclination of iA = 0
+(coplanar) relative to outer companions orbit. The outer
+companion to the binary has mass mB = mA = 0.5 mAB
+and is in an orbit with an eccentricity of eAB = 0.5.
+More generally, we define the relative mass of inner bi-
+nary as fA =
+mAb
+mAa+mAb , where mAb is the smaller of the
+two masses and so this parameter ranges from 0 to 0.5.
+The relative mass of companion is fB =
+mB
+mAB . Since
+the companion may be smaller or larger in mass than
+the inner binary this parameter ranges from 0 to 1. Our
+standard parameters have fA = 0.5 and fB = 0.5. These
+parameters are in the stable region for circumtriple sys-
+tems. The system becomes unstable with larger aA and
+eAB. Adopting the Multilayer Perceptron (MLP) model
+from Vynatheya et al. (2022) and varying eAB we find
+it is stable for eAB ≲ 0.8 and aA/aAB ≲ 0.13.
+We
+check our ranges of parameters with the MLP model
+(Vynatheya et al. 2022) to avoid unstable regions. How-
+ever, the transition between stable and unstable is grad-
+ual rather than abrupt (Hayashi et al. 2022) and so we
+have chosen our standard parameters to be well clear of
+unstable regions.
+2.2. Test particle orbits around our standard triple star
+We run test particle orbits at radius r around the
+triple star. The test particles can have unstable orbits if
+they are too close to the AB binary (Holman & Wiegert
+1999; Quarles et al. 2020; Chen et al. 2020). We only
+consider orbits at radii large enough to be stable. We
+analyse the test particle orbits in the frame of the AB
+binary made up of the companion star orbiting the in-
+ner binary. We characterise the test particle orbit by its
+inclination and nodal phase angle relative to this binary.
+The inclination of the orbit is given by
+i = cos−1(ˆlAB · ˆlp) ,
+(1)
+where ˆlAB is a unit vector in the direction of the AB
+binary angular momentum and ˆlp is a unit vector in
+the direction of the particles angular momentum. The
+nodal phase angle is the angle measured relative to the
+
+3
+150
+100
+50
+0
+50
+100
+150
+i sin
+r=3.5 aAB
+r=4.5 aAB
+100
+0
+100
+i cos
+150
+100
+50
+0
+50
+100
+150
+i sin
+r=5.5 aAB
+100
+0
+100
+i cos
+r=6.0 aAB
+r=3.5 aAB
+r=4.5 aAB
+r=5.5 aAB
+r=6.0 aAB
+Figure 1.
+Test particle orbits around our standard triple star at orbital radii of r = 3.5, 4.5, 5.5 and 6 aAB.
+Left: The
+(i cos φ, i sin φ) phase plane. Right: Precession paths for the angular momentum vector of the test particles plotted on surface
+of sphere. The angular momentum and eccentricity unit vectors of binary shown in green and red, respectively. The circulating
+orbits are shown in green, librating orbits are red and retrograde circulating orbits are blue.
+eccentricity vector of the outer binary and is given by
+φ = tan−1
+�ˆlp · (ˆlAB × ˆeAB)
+ˆlp · ˆeAB
+�
++ 90◦,
+(2)
+(Chen et al. 2019, 2020) where φ is the phase angle and
+ˆeAB is the eccentricity vector of the outer binary.
+We run test particle orbits around our standard triple
+star that begin in circular orbits at radii of r = 3.5, 4.5,
+5.5 and 6 aAB. We start with initial inclinations in 10◦
+increments from 10◦ to 170◦ and with an initial longi-
+tude of the ascending node of 90◦. The resulting orbits
+are plotted in the (i cos φ, i sin φ) phase plane in the left
+panel of Fig. 1. The right panel shows the same infor-
+mation but displays the paths of the particles orbital
+angular momentum vector on the unit sphere.
+For low initial inclinations, there is a circulating region
+shown in green, in which the particle angular momen-
+tum vector precesses around the binary angular momen-
+tum vector. The retrograde circulation region is shown
+in blue where the particles angular momentum vector
+is orbiting about the negative of the binaries angular
+momentum vector. There is a librating region, shown
+in red, where the particle angular momentum vector
+precesses around a stationary inclination. This station-
+ary inclination for close in particles is at i = 90◦ and
+aligned with the binary eccentricity vector. As the par-
+ticle moves to larger orbital radii, the stationary inclina-
+tion moves to higher inclinations. Once the stationary
+inclination is > 180◦ there are no more librating or-
+bits and the particle has similar dynamics to one around
+a circular orbit binary since it nodally precesses about
+the binary angular momentum vector for all inclinations.
+This is very similar to the behavior seen in Lepp et al.
+(2022) where we considered test particle orbits about a
+binary which was precessing due to the effects of general
+relativity. Here the behavior of the triple star system is
+causing a similar precession but at a timescale over an
+order of magnitude higher.
+All the simulations in this paper were run with zero
+mass test particles but to see the effects of a massive
+particle we ran select simulations with various mass par-
+ticles. The simulations are essentially unchanged by in-
+troducing a particle up to mAB/1000 (about a Jupiter
+mass if mAB ≈ 1 M⊙). A Jupiter mass particle follows
+the test particle evolution. Masses significantly above
+this mass can change the evolution. In particular large
+masses in polar orbits induce a precession in the outer
+binary in the opposite direction of that caused by the
+inner binary and cause the total precession of the outer
+binary to be slower.
+2.3. Critical radius for librating orbits
+In Fig. 2 we show the smallest initial inclination for
+a librating orbit, imin, the largest initial inclination for
+a librating orbit, imax, and the stationary inclination,
+is, where the orbit stays at a fixed inclination with no
+nodal precession. The upper left panel in Figure 2 rep-
+
+4
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+imin
+is
+imax
+analytic
+standard model
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+eA=0.4
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+eA=0.6
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+eAB=0.2
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+eAB=0.6
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+aA=1/16 aAB
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+aA=1/30 aAB
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+iA=10
+ 0
+ 20
+ 40
+ 60
+ 80
+ 100
+ 120
+ 140
+ 160
+ 180
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+i (°)
+r (aAB)
+iA=20
+Figure 2. The minimum initial inclination (imin, magenta) and maximum initial inclination (imax, blue) for librating orbits
+and the stationary polar inclination (green) for varying eA, eAB, aAB or iA (in degrees) from our standard model. The analytic
+curve for the stationary state is from equation (8) with α = 2. The key in upper left panel applies to all nine panels.
+resents our standard triple star parameters. There are
+circulating orbits at low inclinations i < imin, and ret-
+rograde circulating orbits for i > imax. If imax > 180◦
+then there is no retrograde circulating region and since
+the librating orbits occur around is, there are no librat-
+ing orbits when is > 180◦. The precession of the triple
+star system causes the stationary inclination to move to
+higher inclinations with increasing test particle radius,
+until it becomes greater than 180◦ and then there are no
+librating orbits. We call this radius the critical radius,
+rc, and it represents the maximum radius at which test
+particles can orbit the outer binary in a polar orbit. For
+our standard triple star parameters, rc = 5.7 aAB.
+We now consider the effect of varying the triple star
+orbital parameters on the test particle orbits. The other
+panels in Figure 2 take our standard model and vary one
+of the parameters. In the next two panels across the top,
+we vary the eccentricity of the inner binary from eA = 0
+to eA = 0.4 and eA = 0.6. The change in eA increases
+the apsidal precession rate of the AB binary by about
+30% for eA = 0.4 and about 70% for 0.6. The radius rc
+then occurs at smaller orbital radii of 5.28 and 4.87 aAB,
+for eA = 0.4 and eA = 0.6, respectively.
+Next, we vary from our standard case the eccentricity
+of the companion from its value of eAB = 0.5 to eAB =
+0.2 and eAB = 0.6. In both cases, rc is reduced, though
+the effect is much weaker than that seen for varying eA.
+This is because eAB affects both the precession rate of
+the ascending node of the test particle and the apsidal
+precession rate of the AB binary (see Section 3).
+We then vary the ratio of the semi-major axis of the
+inner binary to the companion from its standard value of
+aA/aAB = 1/20. For aA/aAB = 1/16 we find rc ≈ 5 aAB
+and for aA/aAB = 1/30, rc = 7.25 aAB which is off the
+range of the plot. This again reflects the change in ap-
+sidal precession rate of the binary with changing geom-
+etry. Finally, we consider the effect of the inclination
+of the inner binary relative to the triple star compan-
+ion. We have restricted our simulations to small angle
+inclinations to avoid ZKL oscillations that would intro-
+
+5
+duce additional time variations. We change our stan-
+dard model to have the inner binary’s orbit inclined to
+the orbital plane of the companion and this increases rc.
+The critical radius gets larger as the apsidal precession
+rate gets smaller. However, we note that the inclination
+has a weak effect on the critical radius.
+Figure 3 shows the critical radius rc as a function of
+some of the triple star parameters. The crosses show
+the numerical determination of the radius. We vary eA,
+eAB, aA, iA, fA and fB. The critical radius depends
+on the rate of apsidal precession of the AB binary as
+well as on the nodal precession rate of the test parti-
+cle orbit (see the next section). The faster the apsidal
+precession the smaller the critical radius.
+For typical
+triple star parameters (Tokovinin 2008, 2021), the criti-
+cal radius is in the approximate range 3−10 aAB, unless
+one of the stars has a much smaller mass than the oth-
+ers. The innermost stable orbit for a polar circumbinary
+test particle is typically around 2 − 2.5 aAB (Chen et al.
+2020) and so the radial range of stable polar circum-
+triple orbits may be quite small. However, as discussed
+in Section 2.1, some observed triples found with short
+period inner binaries have more extreme outer to inner
+semi-major axis ratios that allow the critical radius to
+extend to more than 80 aAB.
+The libration time scale increases with the semi-major
+axis of the test particle. For our standard model, the
+critical radius outside of which there are no polar or-
+bits is rc = 5.73 aAB. In this case, a test particle at
+r = 5.5 aAB orbits near the stationary inclination li-
+brate about it with a period of about 4000PAB, where
+PAB is the orbital period of the outer binary.
+At an
+orbital radius of r = 4.5 aAB, orbits near the stationary
+inclination librate with a period of about 800 PAB.
+3. ANALYTICAL ESTIMATION
+The stationary inclination occurs where the apsidal
+precession rate of the binary is equal to the nodal pre-
+cession rate of the test particle.
+We follow Zanardi
+et al. (2018) to analytically find the stationary incli-
+nation based on the quadrupole order expansion of the
+Hamiltonian. They derived it for the case where general
+relativity drives the apsidal precession. The precession
+of the ascending node of the test particle is given by
+equation (4) in Zanardi et al. (2018).
+For a circular
+(e = 0) polar stationary orbit (Ω = 90◦), the nodal pre-
+cession rate is
+˙Ωs = − mAmBk
+m3/2
+ABr3/2
+�aAB
+r
+�2 3 cos i(1 + 4e2
+AB)
+4
+,
+(3)
+where k2 is the gravitational constant. We equate this
+to the rate of change of the longitude of the periapsis for
+the binary, ˙ϖAB = ˙ωAB + ˙ΩAB, to find the stationary
+inclination for the test particle
+is = cos−1
+�
+− ˙ϖAB
+4
+3k
+(mAB)3/2
+mAmB
+r7/2
+a2
+AB
+1
+(1 + 4e2
+AB)
+�
+.
+(4)
+This formula is general and the apsidal precession rate
+for the binary could come from general relativity (e.g.
+Zanardi et al. 2018), tidal interactions (e.g. Sterne 1939)
+or interactions with a companion star (e.g. Morais &
+Correia 2012).
+The precession rate of the longitude of the periapsis
+of the companion in a triple in the quadrupole approxi-
+mation is given by
+˙ϖAB =
+�3k
+4
+� �mAamAb(mAB)1/2
+(mA)2
+�
+×
+�
+a2
+A
+a7/2
+AB
+� �
+1
+(1 − e2
+AB)2
+�
+F(eA, iA),
+(5)
+where
+F(eA, iA) = (1 + αe2
+A)
+�3 cos(iA)2 − 1
+2
+�
++ 15
+4 e2
+A(1 − cos(iA)2) cos(2ωA)
+(6)
+(see equations 25 and 26 in Morais & Correia 2012),
+where iA is the inclination of the inner binary relative to
+the outer binary, and ωA is the argument of the periapsis
+of the inner binary measured relative to the outer binary.
+The same rate may also be found for the co-planar case
+by adding all the quadrupole terms for ˙ω and ˙Ω for the
+outer binary (equations 74 and 76 in Naoz (2016)), the
+inclination dependence is slightly different as Morais &
+Correia (2012) approximate the outer binary as the fixed
+plane. In all our configurations the outer binary carries
+most of the angular momentum and so this is a good
+approximation, as seen in Figs. 2 to 4. The first term in
+equation (6) sets the average rate of apsidal precession
+and the second term causes an oscillation about this
+average precession rate. If ωA is an odd multiple of 45◦
+then the second term is zero. In practice, one can ignore
+the second term if one wants the average precession over
+long times or in a time which is centered around an
+odd multiple of 45◦. The expression is valid so long as
+aA ≪ aAB and mB is not much less than mA meaning
+that the AB binary has most of the angular momentum
+and thus the AB binary plane is very nearly a fixed plane
+in the system.
+For an inclination of zero, the function in equation (6)
+simplifies to
+F(eAB, 0) = (1 + αe2
+A).
+(7)
+
+6
+ 4
+ 4.2
+ 4.4
+ 4.6
+ 4.8
+ 5
+ 5.2
+ 5.4
+ 5.6
+ 5.8
+ 0
+ 0.2
+ 0.4
+ 0.6
+ 0.8
+ 1
+rc (aAB)
+eA
+rc
+analytic
+rc vs eA
+ 4.4
+ 4.6
+ 4.8
+ 5
+ 5.2
+ 5.4
+ 5.6
+ 5.8
+ 6
+ 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
+rc (aAB)
+eAB
+rc
+analytic
+rc vs eAB
+ 4.5
+ 5
+ 5.5
+ 6
+ 6.5
+ 7
+ 7.5
+ 8
+ 8.5
+ 9
+ 9.5
+ 10
+ 0.02
+ 0.03
+ 0.04
+ 0.05
+ 0.06
+ 0.07
+rc (aAB)
+aA (aAB)
+rc
+analytic
+rc vs aA
+ 5.7
+ 5.8
+ 5.9
+ 6
+ 6.1
+ 6.2
+ 6.3
+ 6.4
+ 6.5
+ 6.6
+ 6.7
+ 0
+ 5
+ 10
+ 15
+ 20
+ 25
+ 30
+rc (aAB)
+iA (°)
+rc
+analytic
+rc vs iA
+ 4
+ 6
+ 8
+ 10
+ 12
+ 14
+ 16
+ 18
+ 0
+ 0.1
+ 0.2
+ 0.3
+ 0.4
+ 0.5
+rc (aAB)
+fA
+rc
+analytic
+rc vs fA
+ 2
+ 2.5
+ 3
+ 3.5
+ 4
+ 4.5
+ 5
+ 5.5
+ 6
+ 0
+ 0.2
+ 0.4
+ 0.6
+ 0.8
+ 1
+rc (aAB)
+fB
+rc
+analytic
+rc vs fB
+Figure 3. The critical radius, rc, inside of which there are polar orbits for varying eA, eAB, aA, iA, fA and fB from our
+standard model parameters. The purple crosses show numerical simulations and the green lines show the analytical estimate.
+The analytic lines only depends on α in the top left plot (since eA = 0 everywhere else) and there we take α = 2.
+ 0.002
+ 0.004
+ 0.006
+ 0.008
+ 0.01
+ 0.012
+ 0.014
+ 0
+ 0.1
+ 0.2
+ 0.3
+ 0.4
+ 0.5
+ 0.6
+ 0.7
+ 0.8
+ 0.9
+precession rate (radians/PAB)
+eA
+standard
+eAB=0.4
+i=30°
+analytic α=1.5
+analytic α=2.0
+ 1.3
+ 1.4
+ 1.5
+ 1.6
+ 1.7
+ 1.8
+ 1.9
+ 2
+ 2.1
+ 2.2
+ 2.3
+ 0.001
+ 0.01
+α
+aA/aAB
+eAB = 0.1
+eAB = 0.3
+eAB = 0.5
+iA = 10°
+iA = 20°
+iA = 30°
+Figure 4.
+Left: Apsidal precession rate ( ˙ϖ) variation with eA around our standard triple star, one with eAB = 0.4 and one
+with an inclination iA = 30◦. The analytical fits are shown with coefficients of both α = 1.5 and α = 2.0. Right: The coefficient
+α in front of e2
+A in the precession formula averaged over the values of eA from 0.1 to 0.9.
+
+7
+Here, the α = 3
+2 factor given in Morais & Correia (2012)
+works well in the limit of aAB ≫ aA and we find it works
+well for ratios of aA/aAB ≲ 0.005. However, our stan-
+dard triple star has aA/aAB = 0.05 and so higher order
+terms have changed this parameter. In Naoz (2016) it
+is clear that the octupole terms vanish for our standard
+case of an equal mass inner binary and so it must be due
+to even higher order terms (e.g. Yokoyama et al. 2003;
+Vinson & Chiang 2018; de El´ıa et al. 2019). We now
+consider numerical fits for this parameter.
+The left panel of Figure 4 shows the numerically de-
+termined particle precession rates around the triple star
+as a function of eA. We consider the standard model,
+the standard model with eAB = 0.4 and the standard
+model with iA = 30◦.
+To get an accurate precession
+rate, we average the precession rate over 40 periods of
+the AB binary and over a time with the relative angle of
+the precession of 45◦, this assures that the second term
+in equation (6) averages to zero.
+In the right hand panel, we show a numerical determi-
+nation of α as a function of aA/aAB. We take eAB = 0.1,
+0.3 and 0.5 with a coplanar binary and inclinations of
+10, 20 and 30◦ with eAB = 0.5. For each point, we vary
+eA between 0.1 to 0.9 in steps of 0.1 and find for each an
+exact α that would give that rate relative to the eA = 0
+rate. We then average all of these. At small ratios of
+aA/aAB the best fit for α is 3/2 but close to our standard
+model aA/aAB = 0.05, a much better fit is α = 2.
+We now take ˙ϖAB from equation (5) and find the sta-
+tionary inclination with equation (4) to be
+is = cos−1
+�
+−mAamAbm2
+AB
+m3
+AmB
+� r
+aAB
+�7/2
+×
+� aA
+aAB
+�2
+1
+(1 + 4e2
+AB)(1 − e2
+AB)2 Fe(eA, iA)
+�
+.
+(8)
+The orange lines in Fig. 2 show the analytic station-
+ary inclination with α = 2. There is good agreement
+between this and the numerical solutions.
+We find the critical particle orbital radius outside of
+which there are no polar orbits by setting is = 180◦ and
+solving for r to find
+rc
+aAB
+=
+�
+m3
+AmB
+mAamAbm2
+AB
+�aAB
+aA
+�2
+×
+(1 − e2
+AB)2(1 + 4e2
+AB)
+Fe(eA, i)
+�2/7
+.
+(9)
+Our standard parameters have rc/aAB = 5.73, in agree-
+ment with the top left panel of Fig. 2. More generally
+we find
+rc
+aAB
+= 5.73 M A E
+F 2/7 ,
+(10)
+where M, A and E are scaling functions for the radius
+in terms of mass, semi-major axis, and eccentricity of
+the companion which have been normalized to one for
+our standard parameters. The scaling with masses is
+M(mAa, mAb, mB) =
+�
+m3
+AmB
+mAamAbm2
+AB
+�2/7
+(11)
+=
+�(1 − fB)fB
+(1 − fA)fA
+�2/7
+,
+(12)
+with semi-major axis is
+A(aA, aAB) =
+� aAB
+20 aA
+�4/7
+,
+(13)
+with the companion eccentricity is
+E(eAB) =
+�8(1 − e2
+AB)2(1 + 4e2
+AB)
+9
+�2/7
+.
+(14)
+The green lines in Fig. 3 show this analytic solution for
+the critical radius. We see that there is good agreement
+between the numerical and analytic solutions.
+4. DISCUSSION AND CONCLUSIONS
+Misaligned circumbinary test particle orbits around an
+eccentric binary undergo nodal precession either about
+the binary angular momentum vector (i = 0◦) or about
+the stationary polar inclination that is aligned to the
+binary eccentricity vector (i = 90◦). The orbit type de-
+pends on the initial particle inclination and the binary
+eccentricity but it does not depend upon the particle
+semi-major axis. With n-body simulations and analytic
+methods we have investigated the dynamics of circum-
+triple particle orbits. For close in particles, the polar
+inclination is 90◦ and the orbits around the triple star
+are similar to those around the outer binary with the
+inner binary replaced by a single star. However, with
+a hierarchical triple star, the inner and outer binaries
+undergo apsidal precession and this leads to an increas-
+ing polar stationary inclination with increasing particle
+semi-major axis.
+There is a critical radius rc outside
+of which there are no polar orbits, only circulating or-
+bits that precess about the binary angular momentum
+vector. We find for typical parameters that the criti-
+cal radius is in the approximate range 3 − 10 times the
+outer binary semi-major axis.
+Therefore, polar cir-
+cumtriple orbits typically exist only relatively close to a
+triple star. But for some observed shorter period inner
+binaries (< 1y), the ratio of the outer to inner semi-
+major axis is quite large (Tokovinin 2021). In such cases,
+the circumtriple orbits can occur at relatively large dis-
+tances from the outer binary.
+
+8
+A low-mass circumtriple disk can undergo similar be-
+haviour to the particles, but the radii of the disc com-
+municate with each other allowing solid body preces-
+sion. Therefore, a disk with an outer radius larger than
+rc could reach a polar state. However, because rc can
+be only a few times the outer binary separation, even
+if a disk began with an outer radius smaller than rc,
+it may quickly spread out beyond this, depending upon
+the disk viscosity. This suggests that a polar circum-
+triple disk could form, although it may be the inner part
+of a broken disk. If rc is small, communication through
+the disk may instead lead the outer parts to dominate
+the behaviour and the disc to move towards coplanar
+alignment. These effects should be investigated in fu-
+ture work.
+There are two triple star systems that may have plan-
+ets orbiting them, GG Tauri A (Phuong et al. 2020b)
+and GW Ori (Bi et al. 2020; Smallwood et al. 2021).
+The GG Tauri A system consists of three stars (Di Folco
+et al. 2014) with mB = 0.6 M⊙, mAa = 0.38 M⊙ and
+mAb = 0.3 M⊙. The outer binary semimajor axis is es-
+timated to be aAB = 36 au and inner binary semi-major
+axis is about aA = 5.1 au. The other orbital parameters
+are uncertain, but we can estimate M ≈ 1 and A ≈ 0.55
+and rc = 3.2 aAB = 113 au. The disk around the triple
+extends from r = 180 au ≈ 5 aAB to 800 au and the
+proposed planet is at about 230 au. The second system,
+GW Ori, has the triple star parameters mAa = 2.47 M⊙,
+mAb = 1.43 M⊙ and mAB = 1.36 M⊙, aA = 1.2 au,
+aAB = 8.89 au, eA = 0.069 and eAB = 0.379 (Kraus
+et al. 2020). These give A = 0.56, M = 0.95, F = 1,
+E = 1.05 and we find rc = 3.0 aAB = 28.4 au. The ob-
+served disk is at r > 36 au ≈ 4 aAB and the proposed
+planet is at r = 100 au. Again the disk and the planet
+are well outside the critical radius.
+For both of these observed circumtriple systems, the
+inner edge of the disk and the orbits of the potential
+planets are larger than rc. This suggests that the dy-
+namics of the planet and disk are similar to that around
+a circular orbit binary. A coplanar circumbinary disc is
+truncated at 2 − 3 times the binary separation (Arty-
+mowicz & Lubow 1994) and the cavity size decreases
+with increasing tilt of the disc (Miranda & Lai 2015;
+Lubow & Martin 2018; Franchini et al. 2019). There-
+fore the inner truncation radius of both of these disks
+are larger than would be predicted for a circumbinary
+disk. This could be a result of the triple star effects de-
+scribed in this work that limit the disc to be in r > rc.
+The tidal truncation of a misaligned circumtriple disk
+should be investigated in future work.
+We thank an anonymous referee for useful comments
+that improved the manuscript.
+We acknowledge sup-
+port from NASA through grants 80NSSC21K0395 and
+80NSSC19K0443.
+SHL thanks the Institute for Ad-
+vanced Study for visitor support.
+REFERENCES
+Aly, H., Dehnen, W., Nixon, C., & King, A. 2015, MNRAS,
+449, 65, doi: 10.1093/mnras/stv128
+Artymowicz, P., & Lubow, S. H. 1994, ApJ, 421, 651,
+doi: 10.1086/173679
+Bate, M. R. 2012, Monthly Notices of the Royal
+Astronomical Society, 419, 3115
+Bate, M. R. 2018, MNRAS, 475, 5618,
+doi: 10.1093/mnras/sty169
+Bate, M. R., Lodato, G., & Pringle, J. E. 2010, MNRAS,
+401, 1505, doi: 10.1111/j.1365-2966.2009.15773.x
+Bi, J., van der Marel, N., Dong, R., et al. 2020, ApJL, 895,
+L18, doi: 10.3847/2041-8213/ab8eb4
+Chen, C., Franchini, A., Lubow, S. H., & Martin, R. G.
+2019, MNRAS, 490, 5634, doi: 10.1093/mnras/stz2948
+—. 2020, MNRAS, 495, 141, doi: 10.1093/mnras/staa1143
+Chen, C., Lubow, S. H., & Martin, R. G. 2020, Monthly
+Notices of the Royal Astronomical Society, 494, 4645,
+doi: 10.1093/mnras/staa1037
+Childs, A. C., & Martin, R. G. 2021a, MNRAS, 507, 3461,
+doi: 10.1093/mnras/stab2419
+—. 2021b, ApJL, 920, L8, doi: 10.3847/2041-8213/ac2957
+Cuello, N., & Giuppone, C. A. 2019, A&A, 628, A119,
+doi: 10.1051/0004-6361/201833976
+de El´ıa, G. C., Zanardi, M., Dugaro, A., & Naoz, S. 2019,
+A&A, 627, A17, doi: 10.1051/0004-6361/201935220
+Di Folco, E., Dutrey, A., Le Bouquin, J. B., et al. 2014,
+A&A, 565, L2, doi: 10.1051/0004-6361/201423675
+Doolin, S., & Blundell, K. M. 2011, MNRAS, 418, 2656,
+doi: 10.1111/j.1365-2966.2011.19657.x
+Duchˆene, G., & Kraus, A. 2013, ARA&A, 51, 269,
+doi: 10.1146/annurev-astro-081710-102602
+Facchini, S., Lodato, G., & Price, D. J. 2013, MNRAS, 433,
+2142, doi: 10.1093/mnras/stt877
+Farago, F., & Laskar, J. 2010, MNRAS, 401, 1189,
+doi: 10.1111/j.1365-2966.2009.15711.x
+Franchini, A., Lubow, S. H., & Martin, R. G. 2019, ApJL,
+880, L18, doi: 10.3847/2041-8213/ab2fd8
+Hamers, A. S. 2021, MNRAS, 500, 3481,
+doi: 10.1093/mnras/staa3498
+
+9
+Hayashi, T., Trani, A. A., & Suto, Y. 2022, arXiv e-prints,
+arXiv:2209.08487. https://arxiv.org/abs/2209.08487
+Holman, M. J., & Wiegert, P. A. 1999, AJ, 117, 621,
+doi: 10.1086/300695
+Kennedy, G. M., Wyatt, M. C., Sibthorpe, B., et al. 2012,
+MNRAS, 421, 2264,
+doi: 10.1111/j.1365-2966.2012.20448.x
+Kennedy, G. M., Matr`a, L., Facchini, S., et al. 2019, Nature
+Astronomy, 3, 230, doi: 10.1038/s41550-018-0667-x
+Kenworthy, M. A., Gonz´alez Picos, D., Elizondo, E., et al.
+2022, A&A, 666, A61, doi: 10.1051/0004-6361/202243441
+Keppler, M., Penzlin, A., Benisty, M., et al. 2020, A&A,
+639, A62, doi: 10.1051/0004-6361/202038032
+Kozai, Y. 1962, AJ, 67, 591, doi: 10.1086/108790
+Kraus, S., Kreplin, A., Young, A. K., et al. 2020, Science,
+369, 1233, doi: 10.1126/science.aba4633
+Larwood, J. D., Nelson, R. P., Papaloizou, J. C. B., &
+Terquem, C. 1996, MNRAS, 282, 597
+Lepp, S., Martin, R. G., & Childs, A. C. 2022, ApJL, 929,
+L5, doi: 10.3847/2041-8213/ac61e1
+Lidov, M. L. 1962, Planet. Space Sci., 9, 719,
+doi: 10.1016/0032-0633(62)90129-0
+Lubow, S. H., & Martin, R. G. 2018, MNRAS, 473, 3733,
+doi: 10.1093/mnras/stx2643
+Lubow, S. H., & Ogilvie, G. I. 2001, ApJ, 560, 997,
+doi: 10.1086/322493
+Mardling, R. A., & Aarseth, S. J. 2001, MNRAS, 321, 398,
+doi: 10.1046/j.1365-8711.2001.03974.x
+Martin, R. G., Lepp, S., Lubow, S. H., et al. 2022, ApJL,
+927, L26, doi: 10.3847/2041-8213/ac54b4
+Martin, R. G., & Lubow, S. H. 2017, ApJL, 835, L28,
+doi: 10.3847/2041-8213/835/2/L28
+—. 2018, MNRAS, 479, 1297, doi: 10.1093/mnras/sty1648
+Miranda, R., & Lai, D. 2015, MNRAS, 452, 2396,
+doi: 10.1093/mnras/stv1450
+Morais, M. H. M., & Correia, A. C. M. 2012, MNRAS, 419,
+3447, doi: 10.1111/j.1365-2966.2011.19986.x
+Naoz, S. 2016, ARA&A, 54, 441,
+doi: 10.1146/annurev-astro-081915-023315
+Nealon, R., Cuello, N., & Alexander, R. 2020, MNRAS,
+491, 4108, doi: 10.1093/mnras/stz3186
+Nixon, C., King, A., & Price, D. 2013, MNRAS, 434, 1946,
+doi: 10.1093/mnras/stt1136
+Offner, S. S. R., Kratter, K. M., Matzner, C. D., Krumholz,
+M. R., & Klein, R. I. 2010, ApJ, 725, 1485,
+doi: 10.1088/0004-637X/725/2/1485
+Papaloizou, J. C. B., & Pringle, J. E. 1983, MNRAS, 202,
+1181, doi: 10.1093/mnras/202.4.1181
+Papaloizou, J. C. B., & Terquem, C. 1995, MNRAS, 274,
+987
+Phuong, N. T., Dutrey, A., Diep, P. N., et al. 2020a, A&A,
+635, A12, doi: 10.1051/0004-6361/201936173
+Phuong, N. T., Dutrey, A., Di Folco, E., et al. 2020b, A&A,
+635, L9, doi: 10.1051/0004-6361/202037682
+Quarles, B., Li, G., Kostov, V., & Haghighipour, N. 2020,
+AJ, 159, 80, doi: 10.3847/1538-3881/ab64fa
+Rein, H., & Liu, S. F. 2012, A&A, 537, A128,
+doi: 10.1051/0004-6361/201118085
+Rein, H., & Spiegel, D. S. 2015, MNRAS, 446, 1424,
+doi: 10.1093/mnras/stu2164
+Rein, H., & Tamayo, D. 2015, MNRAS, 452, 376,
+doi: 10.1093/mnras/stv1257
+Smallwood, J. L., Nealon, R., Chen, C., et al. 2021,
+MNRAS, 508, 392, doi: 10.1093/mnras/stab2624
+Sterne, T. E. 1939, MNRAS, 99, 451,
+doi: 10.1093/mnras/99.5.451
+Tobin, J. J., Looney, L. W., Li, Z.-Y., et al. 2016, The
+Astrophysical Journal, 818, 73,
+doi: 10.3847/0004-637x/818/1/73
+Tokovinin, A. 2008, MNRAS, 389, 925,
+doi: 10.1111/j.1365-2966.2008.13613.x
+—. 2021, Universe, 7, 352, doi: 10.3390/universe7090352
+Tokuda, K., Onishi, T., Saigo, K., et al. 2014, ApJL, 789,
+L4, doi: 10.1088/2041-8205/789/1/L4
+Valtonen, M., & Karttunen, H. 2006, The Three-Body
+Problem (Cambridge University Press)
+Verrier, P. E., & Evans, N. W. 2009, MNRAS, 394, 1721,
+doi: 10.1111/j.1365-2966.2009.14446.x
+Vinson, B. R., & Chiang, E. 2018, MNRAS, 474, 4855,
+doi: 10.1093/mnras/stx3091
+von Zeipel, H. 1910, Astronomische Nachrichten, 183, 345,
+doi: 10.1002/asna.19091832202
+Vynatheya, P., Hamers, A. S., Mardling, R. A., &
+Bellinger, E. P. 2022, arXiv e-prints, arXiv:2207.03151.
+https://arxiv.org/abs/2207.03151
+Wisdom, J., & Holman, M. 1991, j, 102, 1528,
+doi: 10.1086/115978
+Yokoyama, T., Santos, M. T., Cardin, G., & Winter, O. C.
+2003, A&A, 401, 763, doi: 10.1051/0004-6361:20030174
+Zanardi, M., de El´ıa, G. C., Di Sisto, R. P., & Naoz, S.
+2018, A&A, 615, A21, doi: 10.1051/0004-6361/201732127
+Zanazzi, J. J., & Lai, D. 2018, MNRAS, 473, 603,
+doi: 10.1093/mnras/stx2375
+
diff --git a/J9AzT4oBgHgl3EQfVfw4/content/tmp_files/load_file.txt b/J9AzT4oBgHgl3EQfVfw4/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..75b685737e1511d2ae023c9ca001354e089c6703
--- /dev/null
+++ b/J9AzT4oBgHgl3EQfVfw4/content/tmp_files/load_file.txt
@@ -0,0 +1,890 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf,len=889
+page_content='Draft version January 4, 2023 Typeset using LATEX twocolumn style in AASTeX631 Polar circumtriple planets and disks can only form close to a triple star Stephen Lepp,1, 2 Rebecca G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Martin,1, 2 and Stephen H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Lubow3 1Nevada Center for Astrophysics, University of Nevada, Las Vegas, 4505 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Maryland Pkwy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Las Vegas, NV 89154, USA 2Department of Physics and Astronomy,University of Nevada, Las Vegas, 4505 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Maryland Pkwy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Las Vegas, NV 89154, USA 3Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA ABSTRACT Observations of protoplanetary disks around binary and triple star systems suggest that misalign- ments between the orbital plane of the stars and the disks are common.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Motivated by recent observa- tions of polar circumbinary disks, we explore the possibility for polar circumtriple disks and therefore polar circumtriple planets that could form in such a disk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' With n-body simulations and analytic meth- ods we find that the inclusion of the third star, and the associated apsidal precession, significantly reduces the radial range of polar orbits so that circumtriple polar disks and planets can only be found close to the stellar system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Outside of a critical radius, that is typically in the range of 3−10 times the outer binary separation depending upon the binary parameters, the orbits behave the same as they do around a circular orbit binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For some observed systems that have shorter period inner binaries, the critical radius is considerably larger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' If polar circumtriple planets can form, we suggest that it is likely that they form in a disk that was subject to breaking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Keywords: Binary stars (154) — Celestial mechanics (211) — Planet formation (1241) 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' INTRODUCTION Multiple stellar systems are common in star forming regions (Duchˆene & Kraus 2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Disks around triple star systems are also expected to be common (Tobin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Bate 2018) and there are several well known examples including GG Tauri A (Di Folco et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2014;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Keppler et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Phuong et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020a) and GW Ori (Bi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Kraus et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Smallwood et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A common feature of these disks is that they are tilted with respect to the orbital plane of the stars.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Disk mis- alignment may initially occur, for example, because of turbulence in the molecular gas cloud (Offner et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2010;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Tokuda et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2014;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Bate 2012) or later accretion of ma- terial by the young binary (Bate et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2010;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Bate 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Misalignment may be increased later by stellar flybys (Nealon et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020) or bound stellar companions (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Martin & Lubow 2017;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Martin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Around an eccentric binary star system, test particle orbits have two stable stationary states: coplanar align- ment to the binary orbit, and polar alignment in which the angular momentum of the particle orbit is aligned to the binary eccentricity vector and 90◦ to the binary orbital plane (Verrier & Evans 2009;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Farago & Laskar 2010;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Doolin & Blundell 2011;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A par- ticle that is misaligned from one of these two stationary states undergoes nodal precession.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Low initial inclina- tion orbits precess about the binary angular momentum vector while high initial inclination orbits precess about the binary eccentricity vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Since the test particle does not affect the dynamics of the binary, the qualita- tive behaviour does not depend on orbital radius of the particle around the binary unless general relativity or tides become important (Lepp et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A circumbinary disk with a low-mass can undergo sim- ilar dynamical behaviour to a test particle (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='g Aly et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2015;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Martin & Lubow 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' If the disk is in good ra- dial communication, it can undergo solid body preces- sion at a angular momentum weighted average rate (Papaloizou & Terquem 1995;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Larwood et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For protoplanetary disks, the radial communication is wave-like (Papaloizou & Pringle 1983;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Lubow & Ogilvie 2001).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Dissipation in the disk leads to alignment either towards coplanar (Nixon et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2013;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Facchini et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2013) or polar depending on the initial tilt (Martin & Lubow 2017;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Lubow & Martin 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Zanazzi & Lai 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Cuello & Giuppone 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Several polar circumbinary disks around eccentric binaries have been observed (Kennedy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2012, 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Kenworthy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022), although none have yet been observed around a triple star.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' While polar circumbinary planets have not yet been observed, their formation may be as efficient as in a coplanar configu- ration (Childs & Martin 2021a,b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='01284v1 [astro-ph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='EP] 3 Jan 2023 2 While the evolution of circumbinary particles and disks is now fairly well understood, the inclusion of an in- ner hierarchical triple star system has not been explored in detail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In this work, for the first time, we examine the effect of an inner triple star system on the existence of polar orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In Section 2 we use n-body simulations and in Section 3 we compare to analytic models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The inner and outer binaries that compose the triple star undergo apsidal precession.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We show that this can remove the possibility of polar orbits outside of a critical radius from the triple star.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This is similar to the effects of general relativity that also causes apsidal precession of the bi- nary (Lepp et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022) but with much higher precession rates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In Section 4 we draw our conclusions and discuss implications both for observations of circumtriple disks and for the properties of planets that may form in such disks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' CIRCUMTRIPLE PARTICLE ORBITS In this Section we first consider the dynamics of a par- ticle orbiting a triple star with our standard parameters and then we consider the effect of varying different triple star parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We use the REBOUND N-body code (Rein & Liu 2012).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The simulations were integrated us- ing a combination of IAS15, a 15th order Gauss-Radau integrator (Rein & Spiegel 2015) and the WHFast, a symplectic Wisdom-Holman integrator (Rein & Tamayo 2015;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Wisdom & Holman 1991).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Triple star parameters Triple star systems are found to occur with a large range of properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Figure 3 of Tokovinin (2021) plots the outer binary period as a function of inner binary pe- riod for a sample of 1820 systems that lie within a dis- tance of 200pc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The sample is subject to strong selection effects that favor the detection of close spectroscopic bi- naries and resolved wide binaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Nonetheless, the plot suggests that for longer period inner binaries (> 1y), the outer to inner semi-major axis ratios typically range from 3 to 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For shorter period inner binaries (< 1y), the ratio range is typically from 20 to 100, and ratios of greater than 1000 also occur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Triple star systems may be unstable for a wide range of parameter space (Mardling & Aarseth 2001;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Valtonen & Karttunen 2006;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Vynatheya et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We con- sider hierarchical triple systems composed of an inner binary with an outer binary companion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The inclina- tion of the inner binary to the inclination of the binary companion must be small enough to avoid von Zeipel- Kozai-Lidov (ZKL) oscillations (von Zeipel 1910;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Kozai 1962;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Lidov 1962;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Naoz 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Hamers 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Figure 3 of Tokovinin (2021) shows evidence of the stability limit at a period ratio of about 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 predicted by Mardling & Aarseth (2001).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For the range of parameters studied, the eccentricity of all the particle orbits are relatively constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The orbits are scale free in mass and length and we adopt as our mass unit, the total mass of the triple star system, mAB, and for our length unit, the semi- major axis of the outer companion, aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For our stan- dard parameters, the inner binary has a total mass mA and is composed of an equal mass binary with mAa = mAb = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='25 mAB, semi-major axis aA = aAB/20 and eccentricity eA = 0 and an inclination of iA = 0 (coplanar) relative to outer companions orbit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The outer companion to the binary has mass mB = mA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 mAB and is in an orbit with an eccentricity of eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' More generally, we define the relative mass of inner bi- nary as fA = mAb mAa+mAb , where mAb is the smaller of the two masses and so this parameter ranges from 0 to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The relative mass of companion is fB = mB mAB .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Since the companion may be smaller or larger in mass than the inner binary this parameter ranges from 0 to 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Our standard parameters have fA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 and fB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' These parameters are in the stable region for circumtriple sys- tems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The system becomes unstable with larger aA and eAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Adopting the Multilayer Perceptron (MLP) model from Vynatheya et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (2022) and varying eAB we find it is stable for eAB ≲ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 and aA/aAB ≲ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We check our ranges of parameters with the MLP model (Vynatheya et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022) to avoid unstable regions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' How- ever, the transition between stable and unstable is grad- ual rather than abrupt (Hayashi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022) and so we have chosen our standard parameters to be well clear of unstable regions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Test particle orbits around our standard triple star We run test particle orbits at radius r around the triple star.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The test particles can have unstable orbits if they are too close to the AB binary (Holman & Wiegert 1999;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Quarles et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We only consider orbits at radii large enough to be stable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We analyse the test particle orbits in the frame of the AB binary made up of the companion star orbiting the in- ner binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We characterise the test particle orbit by its inclination and nodal phase angle relative to this binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The inclination of the orbit is given by i = cos−1(ˆlAB · ˆlp) , (1) where ˆlAB is a unit vector in the direction of the AB binary angular momentum and ˆlp is a unit vector in the direction of the particles angular momentum.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The nodal phase angle is the angle measured relative to the 3 150 100 50 0 50 100 150 i sin r=3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB r=4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB 100 0 100 i cos 150 100 50 0 50 100 150 i sin r=5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB 100 0 100 i cos r=6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='0 aAB r=3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB r=4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB r=5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB r=6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='0 aAB Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Test particle orbits around our standard triple star at orbital radii of r = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5, 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5, 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 and 6 aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Left: The (i cos φ, i sin φ) phase plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Right: Precession paths for the angular momentum vector of the test particles plotted on surface of sphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The angular momentum and eccentricity unit vectors of binary shown in green and red, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The circulating orbits are shown in green, librating orbits are red and retrograde circulating orbits are blue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' eccentricity vector of the outer binary and is given by φ = tan−1 �ˆlp · (ˆlAB × ˆeAB) ˆlp · ˆeAB � + 90◦, (2) (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019, 2020) where φ is the phase angle and ˆeAB is the eccentricity vector of the outer binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We run test particle orbits around our standard triple star that begin in circular orbits at radii of r = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5, 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5, 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 and 6 aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We start with initial inclinations in 10◦ increments from 10◦ to 170◦ and with an initial longi- tude of the ascending node of 90◦.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The resulting orbits are plotted in the (i cos φ, i sin φ) phase plane in the left panel of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The right panel shows the same infor- mation but displays the paths of the particles orbital angular momentum vector on the unit sphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For low initial inclinations, there is a circulating region shown in green, in which the particle angular momen- tum vector precesses around the binary angular momen- tum vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The retrograde circulation region is shown in blue where the particles angular momentum vector is orbiting about the negative of the binaries angular momentum vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' There is a librating region, shown in red, where the particle angular momentum vector precesses around a stationary inclination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This station- ary inclination for close in particles is at i = 90◦ and aligned with the binary eccentricity vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' As the par- ticle moves to larger orbital radii, the stationary inclina- tion moves to higher inclinations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Once the stationary inclination is > 180◦ there are no more librating or- bits and the particle has similar dynamics to one around a circular orbit binary since it nodally precesses about the binary angular momentum vector for all inclinations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This is very similar to the behavior seen in Lepp et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (2022) where we considered test particle orbits about a binary which was precessing due to the effects of general relativity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Here the behavior of the triple star system is causing a similar precession but at a timescale over an order of magnitude higher.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' All the simulations in this paper were run with zero mass test particles but to see the effects of a massive particle we ran select simulations with various mass par- ticles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The simulations are essentially unchanged by in- troducing a particle up to mAB/1000 (about a Jupiter mass if mAB ≈ 1 M⊙).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A Jupiter mass particle follows the test particle evolution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Masses significantly above this mass can change the evolution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In particular large masses in polar orbits induce a precession in the outer binary in the opposite direction of that caused by the inner binary and cause the total precession of the outer binary to be slower.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Critical radius for librating orbits In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2 we show the smallest initial inclination for a librating orbit, imin, the largest initial inclination for a librating orbit, imax, and the stationary inclination, is, where the orbit stays at a fixed inclination with no nodal precession.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The upper left panel in Figure 2 rep- 4 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) imin is imax analytic standard model 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) eA=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) eA=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) eAB=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) eAB=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) aA=1/16 aAB 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) aA=1/30 aAB 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) iA=10 0 20 40 60 80 100 120 140 160 180 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 i (°) r (aAB) iA=20 Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The minimum initial inclination (imin, magenta) and maximum initial inclination (imax, blue) for librating orbits and the stationary polar inclination (green) for varying eA, eAB, aAB or iA (in degrees) from our standard model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The analytic curve for the stationary state is from equation (8) with α = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The key in upper left panel applies to all nine panels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' resents our standard triple star parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' There are circulating orbits at low inclinations i < imin, and ret- rograde circulating orbits for i > imax.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' If imax > 180◦ then there is no retrograde circulating region and since the librating orbits occur around is, there are no librat- ing orbits when is > 180◦.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The precession of the triple star system causes the stationary inclination to move to higher inclinations with increasing test particle radius, until it becomes greater than 180◦ and then there are no librating orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We call this radius the critical radius, rc, and it represents the maximum radius at which test particles can orbit the outer binary in a polar orbit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For our standard triple star parameters, rc = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We now consider the effect of varying the triple star orbital parameters on the test particle orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The other panels in Figure 2 take our standard model and vary one of the parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In the next two panels across the top, we vary the eccentricity of the inner binary from eA = 0 to eA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 and eA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The change in eA increases the apsidal precession rate of the AB binary by about 30% for eA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 and about 70% for 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The radius rc then occurs at smaller orbital radii of 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='28 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='87 aAB, for eA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 and eA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Next, we vary from our standard case the eccentricity of the companion from its value of eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 to eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 and eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In both cases, rc is reduced, though the effect is much weaker than that seen for varying eA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This is because eAB affects both the precession rate of the ascending node of the test particle and the apsidal precession rate of the AB binary (see Section 3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We then vary the ratio of the semi-major axis of the inner binary to the companion from its standard value of aA/aAB = 1/20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For aA/aAB = 1/16 we find rc ≈ 5 aAB and for aA/aAB = 1/30, rc = 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='25 aAB which is off the range of the plot.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This again reflects the change in ap- sidal precession rate of the binary with changing geom- etry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Finally, we consider the effect of the inclination of the inner binary relative to the triple star compan- ion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We have restricted our simulations to small angle inclinations to avoid ZKL oscillations that would intro- 5 duce additional time variations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We change our stan- dard model to have the inner binary’s orbit inclined to the orbital plane of the companion and this increases rc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The critical radius gets larger as the apsidal precession rate gets smaller.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' However, we note that the inclination has a weak effect on the critical radius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Figure 3 shows the critical radius rc as a function of some of the triple star parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The crosses show the numerical determination of the radius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We vary eA, eAB, aA, iA, fA and fB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The critical radius depends on the rate of apsidal precession of the AB binary as well as on the nodal precession rate of the test parti- cle orbit (see the next section).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The faster the apsidal precession the smaller the critical radius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For typical triple star parameters (Tokovinin 2008, 2021), the criti- cal radius is in the approximate range 3−10 aAB, unless one of the stars has a much smaller mass than the oth- ers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The innermost stable orbit for a polar circumbinary test particle is typically around 2 − 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020) and so the radial range of stable polar circum- triple orbits may be quite small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' However, as discussed in Section 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1, some observed triples found with short period inner binaries have more extreme outer to inner semi-major axis ratios that allow the critical radius to extend to more than 80 aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The libration time scale increases with the semi-major axis of the test particle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For our standard model, the critical radius outside of which there are no polar or- bits is rc = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='73 aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In this case, a test particle at r = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB orbits near the stationary inclination li- brate about it with a period of about 4000PAB, where PAB is the orbital period of the outer binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' At an orbital radius of r = 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 aAB, orbits near the stationary inclination librate with a period of about 800 PAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' ANALYTICAL ESTIMATION The stationary inclination occurs where the apsidal precession rate of the binary is equal to the nodal pre- cession rate of the test particle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We follow Zanardi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (2018) to analytically find the stationary incli- nation based on the quadrupole order expansion of the Hamiltonian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' They derived it for the case where general relativity drives the apsidal precession.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The precession of the ascending node of the test particle is given by equation (4) in Zanardi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For a circular (e = 0) polar stationary orbit (Ω = 90◦), the nodal pre- cession rate is ˙Ωs = − mAmBk m3/2 ABr3/2 �aAB r �2 3 cos i(1 + 4e2 AB) 4 , (3) where k2 is the gravitational constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We equate this to the rate of change of the longitude of the periapsis for the binary, ˙ϖAB = ˙ωAB + ˙ΩAB, to find the stationary inclination for the test particle is = cos−1 � − ˙ϖAB 4 3k (mAB)3/2 mAmB r7/2 a2 AB 1 (1 + 4e2 AB) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (4) This formula is general and the apsidal precession rate for the binary could come from general relativity (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Zanardi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018), tidal interactions (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Sterne 1939) or interactions with a companion star (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Morais & Correia 2012).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The precession rate of the longitude of the periapsis of the companion in a triple in the quadrupole approxi- mation is given by ˙ϖAB = �3k 4 � �mAamAb(mAB)1/2 (mA)2 � × � a2 A a7/2 AB � � 1 (1 − e2 AB)2 � F(eA,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' iA),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (5) where F(eA,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' iA) = (1 + αe2 A) �3 cos(iA)2 − 1 2 � + 15 4 e2 A(1 − cos(iA)2) cos(2ωA) (6) (see equations 25 and 26 in Morais & Correia 2012),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' where iA is the inclination of the inner binary relative to the outer binary,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' and ωA is the argument of the periapsis of the inner binary measured relative to the outer binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The same rate may also be found for the co-planar case by adding all the quadrupole terms for ˙ω and ˙Ω for the outer binary (equations 74 and 76 in Naoz (2016)), the inclination dependence is slightly different as Morais & Correia (2012) approximate the outer binary as the fixed plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In all our configurations the outer binary carries most of the angular momentum and so this is a good approximation, as seen in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2 to 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The first term in equation (6) sets the average rate of apsidal precession and the second term causes an oscillation about this average precession rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' If ωA is an odd multiple of 45◦ then the second term is zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In practice, one can ignore the second term if one wants the average precession over long times or in a time which is centered around an odd multiple of 45◦.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The expression is valid so long as aA ≪ aAB and mB is not much less than mA meaning that the AB binary has most of the angular momentum and thus the AB binary plane is very nearly a fixed plane in the system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For an inclination of zero, the function in equation (6) simplifies to F(eAB, 0) = (1 + αe2 A).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (7) 6 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 1 rc (aAB) eA rc analytic rc vs eA 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 6 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 rc (aAB) eAB rc analytic rc vs eAB 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 7 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 8 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 9 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='07 rc (aAB) aA (aAB) rc analytic rc vs aA 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='9 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 0 5 10 15 20 25 30 rc (aAB) iA (°) rc analytic rc vs iA 4 6 8 10 12 14 16 18 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 rc (aAB) fA rc analytic rc vs fA 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 3 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 6 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 1 rc (aAB) fB rc analytic rc vs fB Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The critical radius, rc, inside of which there are polar orbits for varying eA, eAB, aA, iA, fA and fB from our standard model parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The purple crosses show numerical simulations and the green lines show the analytical estimate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The analytic lines only depends on α in the top left plot (since eA = 0 everywhere else) and there we take α = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='002 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='004 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='006 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='008 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='012 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='014 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='9 precession rate (radians/PAB) eA standard eAB=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 i=30° analytic α=1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 analytic α=2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='7 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='9 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='001 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='01 α aA/aAB eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 iA = 10° iA = 20° iA = 30° Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Left: Apsidal precession rate ( ˙ϖ) variation with eA around our standard triple star, one with eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 and one with an inclination iA = 30◦.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The analytical fits are shown with coefficients of both α = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 and α = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Right: The coefficient α in front of e2 A in the precession formula averaged over the values of eA from 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 7 Here, the α = 3 2 factor given in Morais & Correia (2012) works well in the limit of aAB ≫ aA and we find it works well for ratios of aA/aAB ≲ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' However, our stan- dard triple star has aA/aAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='05 and so higher order terms have changed this parameter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In Naoz (2016) it is clear that the octupole terms vanish for our standard case of an equal mass inner binary and so it must be due to even higher order terms (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Yokoyama et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2003;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Vinson & Chiang 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' de El´ıa et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We now consider numerical fits for this parameter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The left panel of Figure 4 shows the numerically de- termined particle precession rates around the triple star as a function of eA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We consider the standard model, the standard model with eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 and the standard model with iA = 30◦.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' To get an accurate precession rate, we average the precession rate over 40 periods of the AB binary and over a time with the relative angle of the precession of 45◦, this assures that the second term in equation (6) averages to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In the right hand panel, we show a numerical determi- nation of α as a function of aA/aAB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We take eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5 with a coplanar binary and inclinations of 10, 20 and 30◦ with eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For each point, we vary eA between 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='9 in steps of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 and find for each an exact α that would give that rate relative to the eA = 0 rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We then average all of these.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' At small ratios of aA/aAB the best fit for α is 3/2 but close to our standard model aA/aAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='05, a much better fit is α = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We now take ˙ϖAB from equation (5) and find the sta- tionary inclination with equation (4) to be is = cos−1 � −mAamAbm2 AB m3 AmB � r aAB �7/2 × � aA aAB �2 1 (1 + 4e2 AB)(1 − e2 AB)2 Fe(eA, iA) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (8) The orange lines in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2 show the analytic station- ary inclination with α = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' There is good agreement between this and the numerical solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We find the critical particle orbital radius outside of which there are no polar orbits by setting is = 180◦ and solving for r to find rc aAB = � m3 AmB mAamAbm2 AB �aAB aA �2 × (1 − e2 AB)2(1 + 4e2 AB) Fe(eA, i) �2/7 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (9) Our standard parameters have rc/aAB = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='73, in agree- ment with the top left panel of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' More generally we find rc aAB = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='73 M A E F 2/7 , (10) where M, A and E are scaling functions for the radius in terms of mass, semi-major axis, and eccentricity of the companion which have been normalized to one for our standard parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The scaling with masses is M(mAa, mAb, mB) = � m3 AmB mAamAbm2 AB �2/7 (11) = �(1 − fB)fB (1 − fA)fA �2/7 , (12) with semi-major axis is A(aA, aAB) = � aAB 20 aA �4/7 , (13) with the companion eccentricity is E(eAB) = �8(1 − e2 AB)2(1 + 4e2 AB) 9 �2/7 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' (14) The green lines in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 3 show this analytic solution for the critical radius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We see that there is good agreement between the numerical and analytic solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' DISCUSSION AND CONCLUSIONS Misaligned circumbinary test particle orbits around an eccentric binary undergo nodal precession either about the binary angular momentum vector (i = 0◦) or about the stationary polar inclination that is aligned to the binary eccentricity vector (i = 90◦).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The orbit type de- pends on the initial particle inclination and the binary eccentricity but it does not depend upon the particle semi-major axis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' With n-body simulations and analytic methods we have investigated the dynamics of circum- triple particle orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For close in particles, the polar inclination is 90◦ and the orbits around the triple star are similar to those around the outer binary with the inner binary replaced by a single star.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' However, with a hierarchical triple star, the inner and outer binaries undergo apsidal precession and this leads to an increas- ing polar stationary inclination with increasing particle semi-major axis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' There is a critical radius rc outside of which there are no polar orbits, only circulating or- bits that precess about the binary angular momentum vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We find for typical parameters that the criti- cal radius is in the approximate range 3 − 10 times the outer binary semi-major axis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Therefore, polar cir- cumtriple orbits typically exist only relatively close to a triple star.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' But for some observed shorter period inner binaries (< 1y), the ratio of the outer to inner semi- major axis is quite large (Tokovinin 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' In such cases, the circumtriple orbits can occur at relatively large dis- tances from the outer binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 8 A low-mass circumtriple disk can undergo similar be- haviour to the particles, but the radii of the disc com- municate with each other allowing solid body preces- sion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Therefore, a disk with an outer radius larger than rc could reach a polar state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' However, because rc can be only a few times the outer binary separation, even if a disk began with an outer radius smaller than rc, it may quickly spread out beyond this, depending upon the disk viscosity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This suggests that a polar circum- triple disk could form, although it may be the inner part of a broken disk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' If rc is small, communication through the disk may instead lead the outer parts to dominate the behaviour and the disc to move towards coplanar alignment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' These effects should be investigated in fu- ture work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' There are two triple star systems that may have plan- ets orbiting them, GG Tauri A (Phuong et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020b) and GW Ori (Bi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Smallwood et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The GG Tauri A system consists of three stars (Di Folco et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2014) with mB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='6 M⊙, mAa = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='38 M⊙ and mAb = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3 M⊙.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The outer binary semimajor axis is es- timated to be aAB = 36 au and inner binary semi-major axis is about aA = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1 au.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The other orbital parameters are uncertain, but we can estimate M ≈ 1 and A ≈ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='55 and rc = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 aAB = 113 au.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The disk around the triple extends from r = 180 au ≈ 5 aAB to 800 au and the proposed planet is at about 230 au.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The second system, GW Ori, has the triple star parameters mAa = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='47 M⊙, mAb = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='43 M⊙ and mAB = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='36 M⊙, aA = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2 au, aAB = 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='89 au, eA = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='069 and eAB = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='379 (Kraus et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' These give A = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='56, M = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='95, F = 1, E = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='05 and we find rc = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='0 aAB = 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4 au.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The ob- served disk is at r > 36 au ≈ 4 aAB and the proposed planet is at r = 100 au.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Again the disk and the planet are well outside the critical radius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' For both of these observed circumtriple systems, the inner edge of the disk and the orbits of the potential planets are larger than rc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This suggests that the dy- namics of the planet and disk are similar to that around a circular orbit binary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A coplanar circumbinary disc is truncated at 2 − 3 times the binary separation (Arty- mowicz & Lubow 1994) and the cavity size decreases with increasing tilt of the disc (Miranda & Lai 2015;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Lubow & Martin 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Franchini et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' There- fore the inner truncation radius of both of these disks are larger than would be predicted for a circumbinary disk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' This could be a result of the triple star effects de- scribed in this work that limit the disc to be in r > rc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' The tidal truncation of a misaligned circumtriple disk should be investigated in future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We thank an anonymous referee for useful comments that improved the manuscript.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' We acknowledge sup- port from NASA through grants 80NSSC21K0395 and 80NSSC19K0443.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' SHL thanks the Institute for Ad- vanced Study for visitor support.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' REFERENCES Aly, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Dehnen, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Nixon, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & King, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2015, MNRAS, 449, 65, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stv128 Artymowicz, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1994, ApJ, 421, 651, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1086/173679 Bate, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2012, Monthly Notices of the Royal Astronomical Society, 419, 3115 Bate, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018, MNRAS, 475, 5618, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/sty169 Bate, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lodato, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Pringle, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2010, MNRAS, 401, 1505, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='15773.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Bi, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', van der Marel, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Dong, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, ApJL, 895, L18, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/2041-8213/ab8eb4 Chen, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Franchini, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019, MNRAS, 490, 5634, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stz2948 —.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, MNRAS, 495, 141, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/staa1143 Chen, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, Monthly Notices of the Royal Astronomical Society, 494, 4645, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/staa1037 Childs, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021a, MNRAS, 507, 3461, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stab2419 —.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021b, ApJL, 920, L8, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/2041-8213/ac2957 Cuello, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Giuppone, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019, A&A, 628, A119, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/201833976 de El´ıa, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Zanardi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Dugaro, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Naoz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019, A&A, 627, A17, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/201935220 Di Folco, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Dutrey, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Le Bouquin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2014, A&A, 565, L2, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/201423675 Doolin, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Blundell, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2011, MNRAS, 418, 2656, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='19657.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Duchˆene, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Kraus, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2013, ARA&A, 51, 269, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1146/annurev-astro-081710-102602 Facchini, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lodato, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Price, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2013, MNRAS, 433, 2142, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stt877 Farago, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Laskar, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2010, MNRAS, 401, 1189, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='15711.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Franchini, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019, ApJL, 880, L18, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/2041-8213/ab2fd8 Hamers, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021, MNRAS, 500, 3481, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/staa3498 9 Hayashi, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Trani, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Suto, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022, arXiv e-prints, arXiv:2209.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='08487.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='org/abs/2209.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='08487 Holman, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Wiegert, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1999, AJ, 117, 621, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1086/300695 Kennedy, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Wyatt, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Sibthorpe, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2012, MNRAS, 421, 2264, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='20448.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Kennedy, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Matr`a, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Facchini, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2019, Nature Astronomy, 3, 230, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1038/s41550-018-0667-x Kenworthy, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Gonz´alez Picos, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Elizondo, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022, A&A, 666, A61, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/202243441 Keppler, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Penzlin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Benisty, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, A&A, 639, A62, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/202038032 Kozai, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1962, AJ, 67, 591, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1086/108790 Kraus, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Kreplin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Young, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, Science, 369, 1233, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1126/science.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='aba4633 Larwood, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Nelson, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Papaloizou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Terquem, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1996, MNRAS, 282, 597 Lepp, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Childs, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022, ApJL, 929, L5, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/2041-8213/ac61e1 Lidov, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1962, Planet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' Space Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', 9, 719, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1016/0032-0633(62)90129-0 Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018, MNRAS, 473, 3733, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stx2643 Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Ogilvie, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2001, ApJ, 560, 997, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1086/322493 Mardling, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Aarseth, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2001, MNRAS, 321, 398, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1046/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-8711.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='03974.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lepp, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022, ApJL, 927, L26, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/2041-8213/ac54b4 Martin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Lubow, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2017, ApJL, 835, L28, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/2041-8213/835/2/L28 —.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018, MNRAS, 479, 1297, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/sty1648 Miranda, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Lai, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2015, MNRAS, 452, 2396, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stv1450 Morais, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Correia, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2012, MNRAS, 419, 3447, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='19986.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Naoz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2016, ARA&A, 54, 441, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1146/annurev-astro-081915-023315 Nealon, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Cuello, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Alexander, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, MNRAS, 491, 4108, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stz3186 Nixon, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', King, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Price, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2013, MNRAS, 434, 1946, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stt1136 Offner, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Kratter, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Matzner, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Krumholz, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Klein, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2010, ApJ, 725, 1485, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1088/0004-637X/725/2/1485 Papaloizou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Pringle, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1983, MNRAS, 202, 1181, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1181 Papaloizou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Terquem, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1995, MNRAS, 274, 987 Phuong, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Dutrey, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Diep, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020a, A&A, 635, A12, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/201936173 Phuong, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Dutrey, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Di Folco, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020b, A&A, 635, L9, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/202037682 Quarles, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Li, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Kostov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Haghighipour, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2020, AJ, 159, 80, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/1538-3881/ab64fa Rein, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2012, A&A, 537, A128, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/201118085 Rein, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Spiegel, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2015, MNRAS, 446, 1424, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stu2164 Rein, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Tamayo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2015, MNRAS, 452, 376, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stv1257 Smallwood, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Nealon, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Chen, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021, MNRAS, 508, 392, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stab2624 Sterne, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1939, MNRAS, 99, 451, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='451 Tobin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Looney, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Li, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2016, The Astrophysical Journal, 818, 73, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3847/0004-637x/818/1/73 Tokovinin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2008, MNRAS, 389, 925, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='13613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x —.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2021, Universe, 7, 352, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='3390/universe7090352 Tokuda, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Onishi, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Saigo, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2014, ApJL, 789, L4, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1088/2041-8205/789/1/L4 Valtonen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Karttunen, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2006, The Three-Body Problem (Cambridge University Press) Verrier, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Evans, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2009, MNRAS, 394, 1721, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='14446.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='x Vinson, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Chiang, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018, MNRAS, 474, 4855, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stx3091 von Zeipel, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1910, Astronomische Nachrichten, 183, 345, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1002/asna.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='19091832202 Vynatheya, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Hamers, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Mardling, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Bellinger, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2022, arXiv e-prints, arXiv:2207.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='03151.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='org/abs/2207.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='03151 Wisdom, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Holman, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 1991, j, 102, 1528, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1086/115978 Yokoyama, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Santos, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Cardin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Winter, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2003, A&A, 401, 763, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361:20030174 Zanardi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', de El´ıa, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', Di Sisto, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Naoz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018, A&A, 615, A21, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1051/0004-6361/201732127 Zanazzi, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=', & Lai, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content=' 2018, MNRAS, 473, 603, doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
+page_content='1093/mnras/stx2375' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/J9AzT4oBgHgl3EQfVfw4/content/2301.01284v1.pdf'}
diff --git a/JtE0T4oBgHgl3EQfSADl/vector_store/index.pkl b/JtE0T4oBgHgl3EQfSADl/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..fa74935d1a560f989055fe17bab714e15985196e
--- /dev/null
+++ b/JtE0T4oBgHgl3EQfSADl/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9630dddf42783d6b47454b2e694091cacd982237adf8a5acdf106c0a4eb80bf5
+size 145970
diff --git a/L9FRT4oBgHgl3EQfFTdj/content/tmp_files/2301.13479v1.pdf.txt b/L9FRT4oBgHgl3EQfFTdj/content/tmp_files/2301.13479v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3d17b9e13db796da4f536610749a2912afe27451
--- /dev/null
+++ b/L9FRT4oBgHgl3EQfFTdj/content/tmp_files/2301.13479v1.pdf.txt
@@ -0,0 +1,1186 @@
+Archive TimeLine Summarization (ATLS): Conceptual Framework for
+Timeline Generation over Historical Document Collections
+Nicolas Gutehrlé
+Laboratoire CRIT,
+University of Bourgogne
+Franche-Comté, France
+nicolas.gutehrle
+@univ-fcomte.fr
+Antoine Doucet
+Laboratoire L3i,
+University of La Rochelle,
+France
+antoine.doucet
+@univ-lr.fr
+Adam Jatowt
+Dept. of Computer Science &
+Digital Science Center,
+University of Innsbruck, Austria
+adam.jatowt
+@uibk.ac.at
+Abstract
+Archive collections are nowadays mostly avail-
+able through search engines interfaces, which
+allow a user to retrieve documents by issu-
+ing queries.
+The study of these collections
+may be, however, impaired by some aspects
+of search engines, such as the overwhelming
+number of documents returned or the lack of
+contextual knowledge provided. New methods
+that could work independently or in combina-
+tion with search engines are then required to
+access these collections. In this position paper,
+we propose to extend TimeLine Summarization
+(TLS) methods on archive collections to assist
+in their studies. We provide an overview of ex-
+isting TLS methods and we describe a concep-
+tual framework for an Archive TimeLine Sum-
+marization (ATLS) system, which aims to gen-
+erate informative, readable and interpretable
+timelines.
+1
+Introduction
+1.1
+Exploring archives
+In the recent years, archives and libraries across
+the world have frequently conducted digitization
+campaigns of their collections. This first opened
+access to thousands of historical documents to a
+wider public, but also propelled the emergence of
+new research fields such as Digital Humanities and
+Digital History. These collections are usually ac-
+cessible through search engines, which return doc-
+uments relevant to a query specified by the user.
+Unfortunately, standard search engines are not fully
+suited to assist users in exploring historical collec-
+tions such as news archives where temporal aspects
+of documents play a key role. Firstly, search en-
+gines return documents by their relevance to the
+query, typically without considering the chronolog-
+ical or causal relations between them, which may
+prevent the user from understanding the interrela-
+tions between events. Furthermore when exploring
+such documents, the user might lack the contex-
+tual knowledge to understand the events that are
+mentioned in them. This is especially true when ex-
+ploring news archives coming from distant pasts or
+exploring longitudinal collections, i.e. which span
+over a long time frame such as decades or centuries.
+Search engines do not seem to consider the impor-
+tance of an event mention for a given query, thus
+less important events might be returned by the sys-
+tem, especially for broad queries. Improved search
+engines are then required to study such collections.
+1.2
+Augmenting search engines with
+timelines
+One promising method to improve the output of
+search engines operating over archival collection
+is TimeLine Summarization (TLS). TLS consists
+in summarizing multiple documents by generating
+a timeline where important events detected in the
+dataset are associated with a time unit such as a
+day. TLS is a subfield of the Multi-Document Sum-
+marization (MDS) task and has been studied exten-
+sively in the NLP community: for instance, Swan
+and Allan (2000) generate clusters of Named Enti-
+ties and noun chunks that best describe major news
+topics covered in a subset of the TDT-2 dataset
+(Allan et al., 1998), which contains text transcripts
+of broadcast news spanning from January 1, 1998,
+to June 30, 1998, in English; Nguyen et al. (2014)
+generate timelines by detecting events that are the
+most relevant to a user query. They apply their
+methodology on a dataset of newswire texts in En-
+glish covering the 2004-2011 period provided by
+the AFP French news agency; Duan et al. (2017)
+extend these methods to summarize the common
+history of similar entities such as Japanese Cities or
+French scientists. Examples of timelines generated
+by such methods are shown in Figure 1.
+Hence, TLS could serve as a distant reading tool
+and as a first step in exploring a dataset by provid-
+ing an overview of its key events. Moreover, TLS
+could be combined with search engines and used
+arXiv:2301.13479v1 [cs.CL] 31 Jan 2023
+
+Figure 1: Examples of generated timelines by Yu et al. (2021) (left) and Campos et al. (2018) (right), summarizing
+a set of documents about respectively Egyptian protests and the Syrian War. The left timeline outputs a summary
+on a day-to-day basis, whereas the right timeline lists events using uneven periods of time.
+as an interface to search results returned by issuing
+queries over large datasets, as suggested in Swan
+and Allan (2000); Alonso et al. (2021). From there,
+the user could zoom into the documents in order to
+proceed to close reading. Furthermore, these sum-
+maries would be presented in chronological order,
+thus preserving the link between events, and could
+also be contextualized by adding data from external
+knowledge bases as in Ceroni et al. (2014).
+Search engines augmented with timelines would
+be especially useful in a Digital Humanities (DH)
+context such as for facilitating the study of histori-
+cal datasets, as they would provide necessary con-
+text to understand past events and to structure the
+event landscape. They could also help the user un-
+derstand the history of a particular entity such as a
+person or a location, or even a group of such entities
+through providing a bird’s-eye view of the relevant
+data. A good example of such search engine aug-
+mented with TLS is the Conta-me Histórias (Tell
+me stories) platform1, where the user can query
+news articles from the Portuguese web archive.
+The user-friendly interface allows a distant reading
+of the documents returned by the query through
+a timeline that summarizes them, but also allows
+close reading by preserving the link to the original
+documents. To the best of our knowledge, works on
+applying TLS methods to structure archives of his-
+torical documents, or more broadly in the Digital
+Humanities field, are quite scarce.
+1.3
+Challenges of applying TLS to archives
+Unfortunately, several aspects of such archives
+make the application of TLS methods not straight-
+forward: first, these datasets are often processed
+with Optical Character Recognition (OCR). Pre-
+vious studies have shown that downstream tasks
+such as Named Entity Recognition (NER), Event
+Detection (ED) (Boros et al., 2022), Topic Mod-
+1https://contamehistorias.pt/arquivopt
+elling (TM) (Mutuvi et al., 2018) or Named En-
+tity Linking (NEL) (Linhares Pontes et al., 2019)
+are impacted by the quality of the OCR output.
+To our knowledge, there is no study on the im-
+pact of OCR on TLS, but we can assume it will
+be similar. Furthermore, archive collections may
+also differ from contemporary data because of
+their temporal context: orthographic rules may dif-
+fer, places might have changed names (Smith and
+Crane, 2001) or concepts may have acquired an-
+other meaning. Most existing annotated resources
+necessary for NLP components such as NER or ED
+are created on contemporary data. Historical docu-
+ments archives are thus harder to process because
+of this lack of suitable annotated resources.
+Most TLS methods generate timelines through
+statistical analysis of the input dataset. They also
+often require that the input corpus contains docu-
+ments of a similar type and similar content. How-
+ever, an archive collection may be heterogeneous
+and contain documents of different authors, genres,
+topics and periods. It may also be fragmentary and
+not as complete as a contemporary dataset. Finally,
+although the timelines generated by TLS systems
+are often easy to read, the process that created them
+is often not made explicit. If timelines must as-
+sist the study of historical datasets by highlighting
+important events, they must be interpretable and
+explain why these events are deemed important.
+In this position paper, we propose to extend Time-
+Line Summarization (TLS) methods to assist in the
+studies of archive collections. We first present an
+overview of existing TLS methods. We then de-
+scribe a conceptual framework for an Archive Time-
+Line Summarization (ATLS) system, which aims
+to generate informative, readable and interpretable
+timelines, before suggesting several methods to
+implement it.
+This paper is organized as follows: in Section
+2 we present an overview of existing TimeLine
+
+Date
+Summary
+2011-01-25
+Thousands of protesters spilled into the streets of Egypt on Tuesday , an unprecedented
+display of anti-government rage inspired in part by the tumult in the nearby North
+African nation of Tunisia.
+2011-01-26
+Twitter says its site is being blocked in Egypt Egyptians took to the streets in what could
+be a sequel to the recent revolution in Tunisia witter , Facebook and YouTube were
+widely used in Tunisia 's uprising and in Iran last year -LRB-.
+2011-01-28
+With parts of his capital ablaze , Mubarak said he was asking his government to resign
+and would soon announce a new one , pledging to address the concerns of thousand of
+Egyptians protesting in Cairo 's streets . Amre Moussa , the Arab League 's secretary-
+general and a veteran Egyptian diplomat , joined protesters in Cairo 's Tahrir Square on
+Friday , state-run Nile TV reported .From
+To
+Top headlines
+5/2010
+7/2011
+Syrian officials launch tear gas against protesters
+Security forces shoot at protesters
+New York Times journalist with the Pulitzer died of an Asthma attack in Syria
+8/2011
+3/2012
+Assad promises elections in February in Syria
+Us withdraws ambassador from Syria for security reasons
+NATO says goodbye to Libya and the world turns to Syria
+7/2012
+12/2012
+Meeting of senior officials in Geneva failed agreement to end violence in Syria
+Russia delivers three war helicopters to Syria
+Red Cross says Syria is in civil war
+7/2016
+11/2016
+Maternity unit among hospitals bombed in ldlib air strikes
+Russian helicopter shot down in Syria.
+Turkish army enters SyriaSummarization methods. In Section 3 and 4, we re-
+spectively describe our conceptual framework and
+discuss some of its potential applications. Finally,
+we present our conclusion in Section 5, alongside
+possibilities for future works.
+2
+Related Work
+2.1
+TimeLine Summarization
+Most TLS methods generate timelines by apply-
+ing the two following steps: the Date Selection
+step which identifies and ranks the key dates in
+the documents, and the Date Summarization step
+which generates a summary of an event occurring
+at a specific date by picking important sentences in
+the documents published on that date. To identify
+important dates in the dataset, Gholipour Ghalan-
+dari and Ifrim (2020) select the most frequent date
+mentions, Tran et al. (2015b) use a graph-ranking
+model and Kessler et al. (2012) combine a clus-
+tering model and a supervised classifier. For the
+second step, La Quatra et al. (2021) apply state-of-
+the-art methods for Text Summarization (TS) such
+as TextRank (Mihalcea and Tarau, 2004) whereas
+Martschat and Markert (2018) adapt methods from
+the Multi-Document Summarization (MDS) field.
+TLS has been generally extractive, i.e. the sum-
+mary is created by copying textual elements (e.g.,
+sentences or paragraphs) from the input data (Tran
+et al., 2015a). Other works are abstractive, i.e. the
+summary is a completely new text generated by the
+system (Steen and Markert, 2019).
+TLS methods in general tend to be applied to
+summarize datasets describing large events, such as
+the Egyptian protests or the Syrian War (Tran et al.,
+2015b; Martschat and Markert, 2018). These meth-
+ods require that the dataset covers a constrained
+period of time and is homogeneous, i.e. that the
+documents cover the same topic. Standard TLS
+methods are thus not suited to summarize heteroge-
+neous or longitudinal datasets. Some works such as
+Nguyen et al. (2014); Kessler et al. (2012); Chieu
+and Lee (2004); Pasquali et al. (2019) can be de-
+scribed as Query-based TimeLine Summarization
+(QTLS), as they apply TLS on documents related
+to a user query such as documents returned by a
+search engine.
+QTLS generally consists in the two following
+steps: Event Detection and Event Ranking. To
+detect events, Chieu and Lee (2004) select any sen-
+tence where the terms of the query appear, Nguyen
+et al. (2014) cluster by a common date every sen-
+tence returned by the query and Pasquali et al.
+(2019) detect peaks of date occurrences in the time
+span covered by the documents. Other works train a
+classifier to detect important events (Chasin, 2010)
+or rank events by their importance with a Learning-
+to-Rank model (Ge et al., 2015). However, these
+classifiers need training data, which are difficult to
+create since defining what is important is a subjec-
+tive matter. This can lead to disappointing results
+as shown in Chasin (2010). To determine the im-
+portance of events, Nguyen et al. (2014) first score
+them according to their relevancy and saliency to
+the query, then rerank them to ensure a diverse time-
+line. Chieu and Lee (2004) rank the importance of
+a sentence according to their "interest" and "bursti-
+ness", then remove duplicate sentences to ensure
+diversity. Pasquali et al. (2019) use the keyword
+extractor YAKE! (Campos et al., 2018) to weight
+the terms in the event description. Duplicate event
+descriptions are detected with the Levenshtein simi-
+larity measure and removed. Those methods finally
+select the top most important events to generate the
+timeline.
+In order to generalize the application of TLS, Yu
+et al. (2021) propose a Multiple TimeLine Summa-
+rization (MTLS) system, which generates a time-
+line for each story found in the dataset. To do so,
+it first detects events mentioned in the dataset and
+measures their saliency and consistency. An event
+linking step determines the link between these
+events in order to generate each timeline. Similarly,
+Duan et al. (2020) propose the Comparative Time-
+Line Summarization (CTLS) task, which generates
+a comparative timeline highlighting the contrast
+between two timestamped timeline documents (e.g.
+biographies, historical sections, ...) by computing
+local and global importance of events.
+There are few datasets for the TLS task such
+as 17 Timelines (T17) (Tran et al., 2013), CRISIS
+(Tran et al., 2015a), ENTITIES (Gholipour Gha-
+landari and Ifrim, 2020), CovidTLS (La Quatra
+et al., 2021) or TLS-Covid19 (Pasquali et al., 2021)
+which are constructed from contemporary news ar-
+ticles. However, datasets are often lacking in most
+projects. It is then necessary to create a dataset
+from scratch as in Minard et al. (2015); Nguyen
+et al. (2014); Ge et al. (2015); Bedi et al. (2017) or
+extend existing ones as in Yu et al. (2021).
+Due to this lack of datasets, evaluating TLS sys-
+tems is a difficult task. The date selection step can
+be evaluated with the F1-measure (La Quatra et al.,
+2021; Gholipour Ghalandari and Ifrim, 2020) or
+
+with the Mean Average Precision (MAP) metric
+(Nguyen et al., 2014). The date summary is often
+evaluated with one of the ROUGE metrics (Lin,
+2004) to compare a ground-truth timeline and a
+generated one (Nguyen et al., 2014; Duan et al.,
+2020; Yu et al., 2021; Gholipour Ghalandari and
+Ifrim, 2020). Methods relying on event detection
+such as Ge et al. (2015); Minard et al. (2015); Bedi
+et al. (2017) often evaluate their system in terms of
+Precision, Recall and F1-measure. However, most
+projects often lack datasets and must then resort to
+human evaluation as in Duan et al. (2017); Swan
+and Allan (2000); Tran et al. (2015a).
+2.2
+TLS Variants
+We present below formal definitions of several ex-
+isting TLS variants:
+TLS: takes as input a standalone homogeneous
+dataset of timestamped documents D
+=
+{d1, d2, ..., d|D|} and generates a timeline
+T = {p1, p2, ..., p|T|} of time-summary pairs
+pi = (ti, si), where si summarizes important
+events happening at time ti;
+QTLS: outputs a timeline T = {p1, p2, ..., p|T|}
+as a sequence of time-summary pairs pi =
+(ti, si) from a set of timestamped documents
+{d1, d2, ..., d|D|} based on a query Q
+=
+{w1, w2, ..., wk} where wi denotes a word be-
+longing to the query;
+MTLS: takes as input a dataset of timestamped
+documents D = {d1, d2, ..., d|D|} that can be
+standalone or returned using a query Q =
+{w1, w2, ..., wk}, and outputs a set of time-
+lines T
+= {T1, T2, ..., Tm} for each story
+or topic detected in D, where each time-
+line Ti is a sequence of time-summary pairs
+pi = (ti, si);
+CTLS: takes as input two datasets of timestamped
+documents DA
+= {d1, d2, ..., d|DA|} and
+DB = {d1, d2, ..., d|DB|} and outputs two
+timelines TA and TB made of contrasting
+events detected in DA and DB, each as a se-
+quence of time-summary pairs pi = (ti, si);
+3
+Framework
+In this section, we present a conceptual framework
+for an Archive TimeLine Summarization (ATLS)
+which addresses the challenges raised by archive
+collections such as the sparsity of data, OCR prob-
+lems, context shifts and linguistic changes over
+time in order to generate timelines based on these
+datasets. We first provide a definition of ATLS
+and describe the type of dataset expected before
+presenting the framework and discussing how to
+evaluate its output.
+3.1
+Overview
+The framework consists of the two key steps: Time-
+line Generation and Timeline Presentation. The
+first step extracts textual elements describing an
+event and attributes them an importance score.
+The second one generates the timeline by filtering
+events and selecting their description.
+The processing stages of the framework are
+shown in Figure 2. The first step has to run only
+once over the processed dataset, since it aims to
+detect the elements composing the timeline to be
+generated. In contrast, the second step can be run
+multiple times to update the timeline.
+3.2
+Problem Definition
+We define ATLS as follows:
+Input: A longitudinal dataset of timestamped doc-
+uments D = {d1, d2, ..., d|D|} taken from an
+archival collection, either standalone or re-
+turned by a query Q = {w1, w2, ..., wk}. The
+period of time covered by D is usually much
+longer than the one typically used in TLS.
+Output: A timeline T generated from D as a se-
+quence of time-summary pairs pi = (ti, si),
+where si summarizes important events hap-
+pening at time ti.
+We compare the key characteristics of TLS and
+ATLS in Tab. 1.
+3.3
+Expected Dataset
+The framework takes as input a longitudinal dataset
+composed of timestamped documents, such as
+news articles from a historical newspaper collec-
+tion. This dataset can be standalone or made of
+documents returned by a search engine for a given
+query Q.
+The dataset could be in raw format
+or have been pre-processed. We would suggest
+at least the two following pre-processing steps:
+first, we recommend to clean the dataset if it has
+been processed with OCR, either manually or semi-
+automatically, since the OCR quality will impact
+further steps (Nguyen et al., 2021). Secondly, we
+recommend to detect temporal expressions, as they
+are a good indicator of event mentions. Tempo-
+ral expressions are either explicit (e.g. February
+17, 1995) or implicit (e.g. yesterday, next month).
+One can use tools such as HeidelTime (Strötgen
+
+Figure 2: Conceptual pipeline for building the ATLS system
+and Gertz, 2010) or SUTime (Chang and Manning,
+2012) to detect temporal expressions in text and re-
+solve them to an absolute date format, simplifying
+their use in the TLS process. However, we must
+keep in mind that the detection of temporal expres-
+sions, especially implicit ones, is still a challenging
+task. Moreover, available tools such as these were
+mainly conceived for contemporary data, and thus
+may not work as properly on historical data.
+The input dataset could be pre-processed further
+by applying NLP components such as Name En-
+tity Recognition (NER), Topic Modelling (TM),
+Event Extraction (EE), Relation Extraction (RE),
+Keyword Extraction (KE), or Keyword Generation
+(KG). Such annotations could be used to index
+the dataset and allow the user to query documents
+about a specific Named Entity or topic, as in the
+impresso2 or the NewsEye3 platforms.
+3.4
+Timeline Generation
+In this section, we present the first main step of the
+framework, which extracts mentions of events and
+attributes them an importance score.
+3.4.1
+Event Detection
+Although events can be defined in many ways, a
+commonly accepted definition is "something that
+is happening or that is holding true in a given cir-
+cumstance", as stated in the TimeML guidelines
+Saurí et al. (2006). Events can be detected in mul-
+tiple ways: one could detect them through statis-
+tical analysis of the corpus. For instance, Chieu
+and Lee (2004) measure the occurrences of similar
+sentences associated with the same date, whereas
+Pasquali et al. (2019) measure the occurrences of
+articles in atomic time intervals to later aggregate
+them and determine the bursty time periods. These
+statistical methods are especially suited for homo-
+geneous datasets, but may not work as well on
+heterogeneous or fragmentary datasets. One could
+2https://impresso-project.ch/app/
+3https://www.newseye.eu/
+also train a Learning-to-Rank model on summaries
+created by experts in order to detect important sen-
+tences as in (Tran et al., 2013). This would, how-
+ever, require training data which tend to be scarce,
+even when for contemporary data.
+Alternatively, one could use an Event Detection
+model to detect and annotate events in the dataset as
+in Chasin (2010). Event Detection is another task
+in the NLP community that has been extensively
+studied, and some previous works such as Nguyen
+et al. (2020) have already applied these methods in
+humanities contexts. However, we need to keep in
+mind that training such a model requires annotated
+resources that are often lacking, especially for his-
+torical data, and that the OCR quality of documents
+impacts the output of these models.
+Finally, we could select as event any sentence
+containing at least a time expression, either explicit
+or implicit as in Duan et al. (2019); Nguyen et al.
+(2014). This selection could be made even finer by
+taking sentences that also contain a Named Entity
+as in Abujabal and Berberich (2015); Bedi et al.
+(2017). One can then apply algorithms such as
+Affinity Propagation (Frey and Dueck, 2007) or
+Chinese Whispers (Biemann, 2006) to gather sen-
+tences describing the same event as in Rusu et al.
+(2014); Yu et al. (2021); Steen and Markert (2019).
+Regardless of the method used to detect them,
+events should all be associated with time. These
+could be the time expressions occurring with the
+event mentions, or the Document Creation Date
+(DCD) if no time expressions are present. Alterna-
+tively, approaches for estimating the focus time of
+text (Jatowt et al., 2015), in absence of any tempo-
+ral expressions can be applied to associate event-
+related sentences with particular points of time.
+3.4.2
+Event Importance Estimation
+As mentioned in Section 2, the importance of an
+event can be measured in a supervised or semi-
+supervised manner with a classifier (Chasin, 2010;
+
+Timeline Generation step
+TimelinePresentation step
+Assign
+Select filters,
+Selectevents
+importance
+event descriptions,
+mentioned
+scoreto each
+data augmentation
+Update
+event E
+to generate timeline (T)
+the parameters
+in D
+of the timeline
+2
+3
+Input: Dataset (D)
+Set of events (E)
+present in D
+Setofscoredevents (SE)
+Output: Timeline (T)Ge et al., 2015). This method, however, requires
+training data that are difficult to obtain or produce.
+Furthermore, the process leading a classifier to a
+prediction is generally not explained. Since the
+goal of this framework is to assist in the study of
+longitudinal datasets, it is necessary that the pro-
+cess of generating a timeline is interpretable. Thus,
+we would suggest to measure the importance score
+in an unsupervised manner by extracting features
+from the dataset as in (Nguyen et al., 2014; Chieu
+and Lee, 2004; Campos et al., 2018). Some of the
+features that we think could help measure this im-
+portance score are listed below, with suggestions
+on how to compute them:
+Redundancy: The more frequently an event is
+mentioned, the more important it should be.
+One can then simply count the occurrences of
+events, or as an alternative, assign them im-
+portance weights by calculating their TF-IDF
+scores over all the time units. However, as the
+data might be fragmentary in archive datasets,
+this feature should rather not be used alone;
+Contemporary references: an event may be im-
+portant at a given time if other events occur-
+ring around the same period of time refer to it.
+Thus, to evaluate this feature, we could count
+how often an event is referred to from the
+descriptions of other events in a given short
+period of time around that event;
+Retrospective references: Similarly, an event is
+likely to be important if documents keep men-
+tioning it some time after it occurred. To as-
+sess this kind of across-time reference to the
+event, one could count how often (and perhaps
+for how long) an event is mentioned by other
+events that occurred after a given period of
+time. Other solutions may rely on computing
+random walks over graphs composed of times-
+tamped events and/or entities to measure the
+amount of signal propagation from the past to-
+wards "the recent times" (Jatowt et al., 2016);
+Causality: an event is likely to be important if it
+is the cause of other events that occurred after
+it. To evaluate the causality of an event, one
+could use date reference graphs as in Tran
+et al. (2015b), which measure the frequency
+of references, the topical influence and tempo-
+ral influence between two events to determine
+a causal link. It is also possible to use Causal
+Relation Extraction (CRE) methods as pre-
+sented by Gao et al. for example. However,
+the CRE task is far from solved and may re-
+quire much more dataset pre-processing;
+Common sense: some events are clearly more im-
+portant than other, e.g. the birth of a child or
+marrying a partner are usually more impor-
+tant events in a family history than repainting
+a house. To represent that kind of common
+sense knowledge and compute this feature, it
+may be necessary to create a dataset of events
+that are deemed important to train a 1-class
+classifier (1CC) as in Duan et al. (2019) or a
+Learning-to-Rank model as in Ge et al. (2015).
+Note that while important events can be col-
+lected from historical textbooks or history-
+related content, gathering unimportant events
+may be less easy and more problematic; hence
+the solution could be to rely on a 1CC task.
+Using these features, a straightforward formula
+to calculate the importance of an event could be:
+α · F1 + β · F2 + γ · F3 + δ · F4 + ϵ · F5
+where F1, F2, F3, F4, F5 are the scaled values
+of the features described above and α, β, γ, δ, ϵ are
+hyper-parameters of which value is defined by the
+user or document archive custodians. Similarly to
+event detection, the user could be asked to select
+any of these features to compute this score.
+Some periods may contain much more docu-
+ments than others. For instance, fewer documents
+may be available during a war time because of
+censorship or paper restriction. This lack of doc-
+uments may lead to events that are far more or
+far less mentioned than others, and bias frequency-
+based features such as redundancy, contemporary
+and retrospective references. Thus, these features
+should be normalized before being incorporated.
+Furthermore, we suggest these features since
+they are easy to compute, but we also acknowledge
+that they may not be sufficient to measure the im-
+portance of an event from the perspective of an
+expert such as a historian. Because the formula
+to compute the importance score is modular, one
+could incorporate more features in collaboration
+with experts.
+3.5
+Timeline Presentation
+In this section, we describe the second main step of
+the framework, which generates the timeline from
+events scored in the previous step. We present sets
+of filters to select which events should appear on
+the timeline and how they should be presented. We
+
+also describe an optional step of timeline augmen-
+tation using external data.
+3.5.1
+Event Filtering
+A dataset may contain hundreds or thousands of
+mentioned events. It is necessary to select those
+that will be added to the timeline. To do so, we can
+use filters such as described below. The weight of
+these filters could be changed on the user interface,
+thus allowing users to instantly update the timeline.
+Top N: top N most important events are retained;
+Importance Threshold (IT): only
+events
+of
+which the importance score is superior to a
+pre-fixed threshold IT are taken. Individual
+thresholds for the features described in
+Section 3.4 that make up the importance score
+can also be set;
+Topical Diversity Threshold (TopDT):
+removes redundant event mentions and
+ensures the timeline is topically diverse.
+Topical diversity can be simply measured
+using Maximal Marginal Relevance (MMR)
+(Goldstein-Stewart and Carbonell, 1998) or
+the n-gram blocking metric as in Liu (2019);
+Temporal Diversity Threshold (TempDT) :
+ensures every time unit on the generated
+timeline is evenly represented by setting a
+minimum and maximum number of events
+that can appear at each time unit.
+3.5.2
+Event Description Selection
+There are multiple ways to represent an event on a
+timeline. One could select a sentence that describes
+the event. If this sentence is too long, one could
+use sentence compression methods (Filippova and
+Strube, 2008) to only keep its most important part.
+As mentioned earlier, an event might be represented
+by a cluster of sentences. The user can thus select
+one sentence among this cluster or generate a cloud
+of terms of all sentences contained in it, as in Duan
+et al. (2019). One could also use headlines if the tar-
+get documents are articles as in Tran et al. (2015a);
+Pasquali et al. (2019).
+Finally, we could also use a Natural Language
+Generation (NLG) system as in Steen and Markert
+(2019), as these generated texts are often easier
+to understand than text extracted from the docu-
+ments. However, abstractive methods such as these
+may suffer from inaccuracies or hallucinations, i.e.
+generate information that is not present in the orig-
+inal documents. Thus, abstractive methods might
+generate improper event descriptions and lose the
+connection with the original documents. On the
+other hand, a common drawback of purely extrac-
+tive methods is that selected sentences may require
+some context or at least post-processing for users to
+be able to properly understand them (e.g. pronouns
+may need to be resolved or we need to add defini-
+tions or descriptions of some entities or events).
+3.5.3
+Timeline Augmentation
+To properly understand them, some events may re-
+quire contextual knowledge that is missing from
+the processed dataset. This can especially happen
+if the user is not a domain expert. Such contextual
+knowledge may be found in knowledge bases such
+as Wikidata or Wikipedia Year pages (see for exam-
+ple (Tran et al., 2015c)). Thus, timelines generated
+by an ATLS system could be augmented with con-
+textual data provided by external knowledge bases
+as in (Ceroni et al., 2014). These augmented time-
+lines could help in explaining a dataset by summa-
+rizing it and providing the user with the necessary
+knowledge to understand it. Unfortunately, most
+resources created by experts are not in a machine-
+readable format (Gutehrlé et al., 2021). Hence, this
+step may require more effort.
+3.6
+Timeline Evaluation
+As mentioned earlier, the evaluation of a TLS sys-
+tem is a difficult task because of the lack of evalu-
+ation datasets and the inherent subjectivity of the
+task. In order to evaluate the output, we would
+suggest to manually assess the produced timelines,
+either by following some evaluation criteria as in
+Duan et al. (2017), or by comparing them with
+resources created by experts such as timelines de-
+rived from history books as in Bedi et al. (2017).
+One could also use this framework to bootstrap
+an evaluation dataset specific to the given corpus,
+towards an automatic evaluation.
+4
+Discussion
+In this section, we describe two hypothetical use
+cases comparing the application of TLS and ATLS
+systems, and compare in Table 1 the types of
+datasets and timelines both methods can process.
+Finally, we discuss potential extensions of ATLS.
+4.1
+Use cases
+In the first hypothetical use case, a user has curated
+a homogeneous dataset of timestamped documents
+from Web archives. This dataset is made of news
+articles related to a story spanning over a year. It
+has been pre-processed to remove HTML tags and
+
+TLS
+ATLS
+Covered period
+Shorter
+Longer
+Input Data Size
+Small / Medium
+Large
+Documents type
+Timestamped documents (e.g. news articles)
+Document Format
+Usually born digital
+Often digitized
+Data Integrity
+Usually complete
+Can be fragmentary
+Presence of noise
+Less likely
+Depends on OCR quality
+Semantic evolution
+Less common
+Possible (esp. over long time)
+Need for query-based filtering
+Optional (depends on data size and heterogeneity)
+Need for contextualization
+Less likely
+More likely (esp. over long time)
+Need for interpretable output
+Yes
+Table 1: Comparison of TLS and ATLS tasks
+extract temporal expressions. To generate the time-
+line, the user applies the TLS method: important
+dates are first selected before generating a sum-
+mary of events occurring at each date. The user
+can select the sentence mentioning the event, the
+headline of the article or apply abstractive methods
+to generate its description.
+In the second hypothetical use case, a user has
+curated a heterogeneous corpus to study the eco-
+nomical life of a certain French region in the 20th
+century. This corpus is composed of periodicals,
+newspapers and magazines from different sources
+(parishes, libraries, etc.) published over a cen-
+tury and processed with OCR. This dataset has
+also been pre-processed: the documents have been
+first cleaned of OCR errors, then automatically an-
+notated with Temporal Expression Extraction and
+Named Entity Recognition components. Further-
+more, the dataset has been indexed so as to allow
+query-based searching. To generate the timeline,
+the user applies some of the ATLS approaches men-
+tioned in this paper: events are first detected by
+clustering similar sentences that contain a temporal
+expression and a Named Entity. The importance
+of these events is then scored using the formula
+described in Section 3.4.2. The timeline is gen-
+erated by setting high values to the topical and
+temporal diversity thresholds, and augmented with
+external data from Wikidata, so as to ensure a com-
+prehensive and contextualized timeline. Similarly,
+the user can select from the user interface to use
+a cloud of terms or a sentence from the cluster to
+generate event descriptions.
+4.2
+Extensions of ATLS systems
+Timelines are usually represented linearly, where
+each time unit is of the same size (usually a day or
+a year). However, the optimal granularity of tem-
+poral units might vary when generating a timeline
+over a long period of time. For example, when
+referring to a distant past, humans tend to often de-
+scribe entire decades or years rather than discussing
+each day or month which is more common for the
+recent past. Furthermore, events mentioned in his-
+torical documents might not always be recorded
+with the same temporal precision (e.g., some events
+may have missing dates, the dates can be impre-
+cise or difficult to be inferred). A possible solution
+would be to generate logarithmic timelines, where
+the granularity of the time unit changes over time,
+as suggested in Jatowt and Au Yeung (2011).
+If the documents in the datasets are annotated
+with Named Entities, one could generate entity-
+based timelines. This could help understand the
+history of a specific entity such as a person or a
+location as in Duan et al. (2019). This idea could
+be extended by generating aggregate timelines for
+multiple entities at the same time. These timelines
+could be agglomerative or contrastive and respec-
+tively show the similarities and differences between
+the history of multiple entities of the same type
+(e.g., cities in the same region or country, scientists
+of the same area). Similar to Duan et al. (2020),
+such comparative timelines would allow to study
+the history of entities of the same or similar type,
+e.g. Berlin vs. Paris or even entities of different
+types, e.g. Paris and the writer Victor Hugo.
+5
+Conclusion
+TimeLine Summarization can be a useful tool for
+getting an overview of historical collections as well
+as it can serve as a novel information access means
+to news article archives. In this position paper,
+we have presented an overview of existing TLS
+methods and described a conceptual framework for
+Archive TimeLine Summarization systems.
+The implementation of the framework outlined
+in this paper will be the subject of our future work.
+We also intend to ask humanities scholars (his-
+torians, archivists, ...) to evaluate the quality of
+generated timelines and the effectiveness of our
+framework for the study of archive collections.
+
+References
+Abdalghani Abujabal and Klaus Berberich. 2015. Im-
+portant events in the past, present, and future. pages
+1315–1320.
+James Allan, Ron Papka, and Victor Lavrenko. 1998.
+On-line new event detection and tracking. In Pro-
+ceedings of the 21st Annual International ACM SI-
+GIR Conference on Research and Development in
+Information Retrieval, SIGIR ’98, page 37–45, New
+York, NY, USA. Association for Computing Machin-
+ery.
+Omar Alonso, Stefano Marchesin, Marc Najork, and
+Gianmaria Silvello, editors. 2021. Proceedings of
+the Second International Conference on Design of
+Experimental Search & Information REtrieval Sys-
+tems, Padova, Italy, September 15-18, 2021, vol-
+ume 2950 of CEUR Workshop Proceedings. CEUR-
+WS.org.
+Harsimran Bedi, Sangameshwar Patil, Swapnil Hing-
+mire, and Girish Palshikar. 2017.
+Event time-
+line generation from history textbooks.
+In Pro-
+ceedings of the 4th Workshop on Natural Lan-
+guage Processing Techniques for Educational Appli-
+cations (NLPTEA 2017), pages 69–77, Taipei, Tai-
+wan. Asian Federation of Natural Language Process-
+ing.
+Chris Biemann. 2006. Chinese whispers - an efficient
+graph clustering algorithm and its application to nat-
+ural language processing problems. In Proceedings
+of TextGraphs: the First Workshop on Graph Based
+Methods for Natural Language Processing, pages
+73–80, New York City. Association for Computa-
+tional Linguistics.
+Emanuela Boros, Nhu Khoa Nguyen, Gaël Lejeune,
+and Antoine Doucet. 2022.
+Assessing the impact
+of OCR noise on multilingual event detection over
+digitised documents. International Journal on Digi-
+tal Libraries.
+Ricardo Campos, Vítor Mangaravite, Arian Pasquali,
+Alípio Mário Jorge, Célia Nunes, and Adam Ja-
+towt. 2018.
+Yake!
+collection-independent auto-
+matic keyword extractor. In Advances in Informa-
+tion Retrieval, pages 806–810, Cham. Springer In-
+ternational Publishing.
+Andrea Ceroni, Nam Khanh Tran, Nattiya Kanhabua,
+and Claudia Niederée. 2014. Bridging temporal con-
+text gaps using time-aware re-contextualization. In
+Proceedings of the 37th International ACM SIGIR
+Conference on Research & Development in Informa-
+tion Retrieval, SIGIR ’14, page 1127–1130, New
+York, NY, USA. Association for Computing Machin-
+ery.
+Angel X. Chang and Christopher Manning. 2012. SU-
+Time: A library for recognizing and normalizing
+time expressions. In Proceedings of the Eighth In-
+ternational Conference on Language Resources and
+Evaluation (LREC’12), pages 3735–3740, Istanbul,
+Turkey. European Language Resources Association
+(ELRA).
+Rachel Chasin. 2010. Event and temporal information
+extraction towards timelines of wikipedia articles.
+Hai Leong Chieu and Yoong Keok Lee. 2004. Query
+based event extraction along a timeline. In Proceed-
+ings of the 27th Annual International ACM SIGIR
+Conference on Research and Development in Infor-
+mation Retrieval, SIGIR ’04, page 425–432, New
+York, NY, USA. Association for Computing Machin-
+ery.
+Yijun Duan, Adam Jatowt, and Katsumi Tanaka. 2017.
+Discovering typical histories of entities by multi-
+timeline summarization. In Proceedings of the 28th
+ACM Conference on Hypertext and Social Media,
+HT ’17, page 105–114, New York, NY, USA. As-
+sociation for Computing Machinery.
+Yijun Duan, Adam Jatowt, and Katsumi Tanaka. 2019.
+History-driven entity categorization. In Web and Big
+Data, pages 349–364, Cham. Springer International
+Publishing.
+Yijun Duan, Adam Jatowt, and Masatoshi Yoshikawa.
+2020. Comparative timeline summarization via dy-
+namic affinity-preserving random walk. In ECAI.
+Katja Filippova and Michael Strube. 2008.
+Depen-
+dency tree based sentence compression. In INLG.
+Brendan J. Frey and Delbert Dueck. 2007. Clustering
+by passing messages between data points. Science,
+315(5814):972–976.
+Lei Gao, Prafulla Kumar Choubey, and Ruihong
+Huang. Modeling document-level causal structures
+for event causal relation identification. Proceedings
+of the 2019 Conference of the North American Chap-
+ter of the Association for Computational Linguistics:
+Human Language Technologies, Volume 1 (Long Pa-
+pers).
+Tao Ge, Wenzhe Pei, Heng Ji, Sujian Li, Baobao
+Chang, and Zhifang Sui. 2015.
+Bring you to the
+past:
+Automatic generation of topically relevant
+event chronicles. In Proceedings of the 53rd Annual
+Meeting of the Association for Computational Lin-
+guistics and the 7th International Joint Conference
+on Natural Language Processing (Volume 1: Long
+Papers), pages 575–585, Beijing, China. Associa-
+tion for Computational Linguistics.
+Demian Gholipour Ghalandari and Georgiana Ifrim.
+2020. Examining the state-of-the-art in news time-
+line summarization. In Proceedings of the 58th An-
+nual Meeting of the Association for Computational
+Linguistics, pages 1322–1334, Online. Association
+for Computational Linguistics.
+Jade Goldstein-Stewart and Jaime G. Carbonell. 1998.
+Summarization:
+(1) using MMR for Diversity-
+Based Reranking and (2) Evaluating Summaries. In
+TIPSTER.
+
+Nicolas Gutehrlé, Oleg Harlamov, Farimah Karimi,
+Haoyu Wei, Axel Jean-Caurant, and Lidia Pivo-
+varova. 2021.
+SpaceWars: A Web Interface for
+Exploring the Spatio-temporal Dimensions of WWI
+Newspaper Reporting.
+CEUR Workshop Proceed-
+ings.
+Adam Jatowt and Ching-man Au Yeung. 2011.
+Ex-
+tracting collective expectations about the future
+from large text collections. In Proceedings of the
+20th ACM International Conference on Informa-
+tion and Knowledge Management, CIKM ’11, page
+1259–1264, New York, NY, USA. Association for
+Computing Machinery.
+Adam Jatowt, Daisuke Kawai, and Katsumi Tanaka.
+2016.
+Predicting importance of historical per-
+sons using wikipedia. In Proceedings of the 25th
+ACM International Conference on Information and
+Knowledge Management, CIKM 2016, Indianapolis,
+IN, USA, October 24-28, 2016, pages 1909–1912.
+ACM.
+Adam Jatowt, Ching-man Au Yeung, and Katsumi
+Tanaka. 2015.
+Generic method for detecting fo-
+cus time of documents.
+Inf. Process. Manag.,
+51(6):851–868.
+Rémy Kessler,
+Xavier Tannier,
+Caroline Hagège,
+Véronique Moriceau, and André Bittar. 2012. Find-
+ing salient dates for building thematic timelines. In
+ACL.
+Moreno La Quatra, Luca Cagliero, Elena Baralis, Al-
+berto Messina, and Maurizio Montagnuolo. 2021.
+Summarize Dates First: A Paradigm Shift in Time-
+line Summarization, page 418–427. Association for
+Computing Machinery, New York, NY, USA.
+Chin-Yew Lin. 2004.
+ROUGE: A package for auto-
+matic evaluation of summaries. In Text Summariza-
+tion Branches Out, pages 74–81, Barcelona, Spain.
+Association for Computational Linguistics.
+Elvys Linhares Pontes, Ahmed Hamdi, Nicolas Sidère,
+and Antoine Doucet. 2019. Impact of OCR Quality
+on Named Entity Linking. In International Confer-
+ence on Asia-Pacific Digital Libraries 2019, Kuala
+Lumpur, Malaysia.
+Yang Liu. 2019. Fine-tune BERT for extractive sum-
+marization.
+Sebastian Martschat and Katja Markert. 2018. A tem-
+porally sensitive submodularity framework for time-
+line summarization.
+In Proceedings of the 22nd
+Conference on Computational Natural Language
+Learning, pages 230–240, Brussels, Belgium. Asso-
+ciation for Computational Linguistics.
+Rada Mihalcea and Paul Tarau. 2004.
+TextRank:
+Bringing order into text. In Proceedings of the 2004
+Conference on Empirical Methods in Natural Lan-
+guage Processing, pages 404–411, Barcelona, Spain.
+Association for Computational Linguistics.
+Anne-Lyse Minard, Manuela Speranza, Eneko Agirre,
+Itziar Aldabe, Marieke van Erp, Bernardo Magnini,
+German Rigau, and Rubén Urizar. 2015. SemEval-
+2015 task 4: TimeLine: Cross-document event or-
+dering.
+In Proceedings of the 9th International
+Workshop on Semantic Evaluation (SemEval 2015),
+pages 778–786, Denver, Colorado. Association for
+Computational Linguistics.
+Stephen Mutuvi, Antoine Doucet, Moses Odeo, and
+Adam Jatowt. 2018. Evaluating the Impact of OCR
+Errors on Topic Modeling. In Maturity and Innova-
+tion in Digital Libraries. 20th International Confer-
+ence on Asia-Pacific Digital Libraries, ICADL 2018,
+Hamilton, New Zealand, November 19-22, 2018,
+Proceedings, pages 3 – 14.
+Kiem-Hieu Nguyen, Xavier Tannier, and Veronique
+Moriceau. 2014. Ranking multidocument event de-
+scriptions for building thematic timelines. In Pro-
+ceedings of COLING 2014, the 25th International
+Conference on Computational Linguistics:
+Tech-
+nical Papers, pages 1208–1217, Dublin, Ireland.
+Dublin City University and Association for Compu-
+tational Linguistics.
+Nhu Khoa Nguyen, Emanuela Boros, Gaël Lejeune,
+and Antoine Doucet. 2020.
+Impact analysis of
+document digitization on event extraction. In Pro-
+ceedings of the 4th Workshop on Natural Language
+for Artificial Intelligence (NL4AI 2020) co-located
+with the 19th International Conference of the Ital-
+ian Association for Artificial Intelligence (AI*IA
+2020), Anywhere, November 25th-27th, 2020, vol-
+ume 2735 of CEUR Workshop Proceedings, pages
+17–28. CEUR-WS.org.
+Thi Tuyet Hai Nguyen, Adam Jatowt, Mickaël Cous-
+taty, and Antoine Doucet. 2021.
+Survey of Post-
+OCR processing approaches. ACM Comput. Surv.,
+54(6):124:1–124:37.
+Arian Pasquali, Ricardo Campos, Alexandre Ribeiro,
+Brenda Salenave Santana, Alípio Mário Jorge, and
+Adam Jatowt. 2021. Tls-covid19: A new annotated
+corpus for timeline summarization. In ECIR.
+Arian Pasquali, Vítor Mangaravite, Ricardo Campos,
+Alípio Mário Jorge, and Adam Jatowt. 2019. Inter-
+active system for automatically generating temporal
+narratives.
+In Advances in Information Retrieval,
+pages 251–255, Cham. Springer International Pub-
+lishing.
+Delia Rusu, James Hodson, and Anthony Kimball.
+2014. Unsupervised techniques for extracting and
+clustering complex events in news.
+In Proceed-
+ings of the Second Workshop on EVENTS: Definition,
+Detection, Coreference, and Representation, pages
+26–34, Baltimore, Maryland, USA. Association for
+Computational Linguistics.
+Roser Saurí, Jessica Moszkowicz, Bob Knippen, Rob
+Gaizauskas, Andrea Setzer, and James Pustejovsky.
+2006. Timeml annotation guidelines version 1.2.1.
+
+David A. Smith and Gregory Crane. 2001.
+Disam-
+biguating geographic names in a historical digital
+library.
+In Proceedings of the 5th European Con-
+ference on Research and Advanced Technology for
+Digital Libraries, ECDL ’01, page 127–136, Berlin,
+Heidelberg. Springer-Verlag.
+Julius Steen and Katja Markert. 2019. Abstractive time-
+line summarization. In Proceedings of the 2nd Work-
+shop on New Frontiers in Summarization, pages 21–
+31, Hong Kong, China. Association for Computa-
+tional Linguistics.
+Jannik Strötgen and Michael Gertz. 2010. HeidelTime:
+High quality rule-based extraction and normaliza-
+tion of temporal expressions. In Proceedings of the
+5th International Workshop on Semantic Evaluation,
+pages 321–324, Uppsala, Sweden. Association for
+Computational Linguistics.
+Russell Swan and James Allan. 2000. Automatic gen-
+eration of overview timelines.
+In Proceedings of
+the 23rd Annual International ACM SIGIR Confer-
+ence on Research and Development in Information
+Retrieval, SIGIR ’00, page 49–56, New York, NY,
+USA. Association for Computing Machinery.
+Giang Tran, Mohammad Alrifai, and Eelco Herder.
+2015a. Timeline summarization from relevant head-
+lines. In Advances in Information Retrieval, pages
+245–256, Cham. Springer International Publishing.
+Giang Tran, Eelco Herder, and Katja Markert. 2015b.
+Joint graphical models for date selection in timeline
+summarization. In Proceedings of the 53rd Annual
+Meeting of the Association for Computational Lin-
+guistics and the 7th International Joint Conference
+on Natural Language Processing (Volume 1: Long
+Papers), pages 1598–1607, Beijing, China. Associa-
+tion for Computational Linguistics.
+Giang Binh Tran, Tuan Tran, Nam Khanh Tran,
+Mohammad Alrifai, and Nattiya Kanhabua. 2013.
+Leveraging learning to rank in an optimization
+framework for timeline summarization.
+Nam Khanh Tran, Andrea Ceroni, Nattiya Kanhabua,
+and Claudia Niederée. 2015c.
+Back to the past:
+Supporting interpretations of forgotten stories by
+time-aware re-contextualization. In Proceedings of
+the Eighth ACM International Conference on Web
+Search and Data Mining, WSDM 2015, Shanghai,
+China, February 2-6, 2015, pages 339–348. ACM.
+Yi Yu, Adam Jatowt, Antoine Doucet, Kazunari
+Sugiyama, and Masatoshi Yoshikawa. 2021. Multi-
+TimeLine summarization (MTLS): Improving time-
+line summarization by generating multiple sum-
+maries.
+In Proceedings of the 59th Annual Meet-
+ing of the Association for Computational Linguistics
+and the 11th International Joint Conference on Nat-
+ural Language Processing (Volume 1: Long Papers),
+pages 377–387, Online. Association for Computa-
+tional Linguistics.
+
diff --git a/L9FRT4oBgHgl3EQfFTdj/content/tmp_files/load_file.txt b/L9FRT4oBgHgl3EQfFTdj/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..18178e02cc2f91be13ed8674a8862eb7a80f68ee
--- /dev/null
+++ b/L9FRT4oBgHgl3EQfFTdj/content/tmp_files/load_file.txt
@@ -0,0 +1,681 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf,len=680
+page_content='Archive TimeLine Summarization (ATLS): Conceptual Framework for Timeline Generation over Historical Document Collections Nicolas Gutehrlé Laboratoire CRIT, University of Bourgogne Franche-Comté, France nicolas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='gutehrle @univ-fcomte.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='fr Antoine Doucet Laboratoire L3i, University of La Rochelle, France antoine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='doucet @univ-lr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='fr Adam Jatowt Dept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' of Computer Science & Digital Science Center, University of Innsbruck, Austria adam.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='jatowt @uibk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='at Abstract Archive collections are nowadays mostly avail- able through search engines interfaces, which allow a user to retrieve documents by issu- ing queries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The study of these collections may be, however, impaired by some aspects of search engines, such as the overwhelming number of documents returned or the lack of contextual knowledge provided.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' New methods that could work independently or in combina- tion with search engines are then required to access these collections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In this position paper, we propose to extend TimeLine Summarization (TLS) methods on archive collections to assist in their studies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We provide an overview of ex- isting TLS methods and we describe a concep- tual framework for an Archive TimeLine Sum- marization (ATLS) system, which aims to gen- erate informative, readable and interpretable timelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 1 Introduction 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1 Exploring archives In the recent years, archives and libraries across the world have frequently conducted digitization campaigns of their collections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This first opened access to thousands of historical documents to a wider public, but also propelled the emergence of new research fields such as Digital Humanities and Digital History.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' These collections are usually ac- cessible through search engines, which return doc- uments relevant to a query specified by the user.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Unfortunately, standard search engines are not fully suited to assist users in exploring historical collec- tions such as news archives where temporal aspects of documents play a key role.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Firstly, search en- gines return documents by their relevance to the query, typically without considering the chronolog- ical or causal relations between them, which may prevent the user from understanding the interrela- tions between events.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Furthermore when exploring such documents, the user might lack the contex- tual knowledge to understand the events that are mentioned in them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This is especially true when ex- ploring news archives coming from distant pasts or exploring longitudinal collections, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' which span over a long time frame such as decades or centuries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Search engines do not seem to consider the impor- tance of an event mention for a given query, thus less important events might be returned by the sys- tem, especially for broad queries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Improved search engines are then required to study such collections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2 Augmenting search engines with timelines One promising method to improve the output of search engines operating over archival collection is TimeLine Summarization (TLS).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' TLS consists in summarizing multiple documents by generating a timeline where important events detected in the dataset are associated with a time unit such as a day.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' TLS is a subfield of the Multi-Document Sum- marization (MDS) task and has been studied exten- sively in the NLP community: for instance, Swan and Allan (2000) generate clusters of Named Enti- ties and noun chunks that best describe major news topics covered in a subset of the TDT-2 dataset (Allan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 1998), which contains text transcripts of broadcast news spanning from January 1, 1998, to June 30, 1998, in English;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014) generate timelines by detecting events that are the most relevant to a user query.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' They apply their methodology on a dataset of newswire texts in En- glish covering the 2004-2011 period provided by the AFP French news agency;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017) extend these methods to summarize the common history of similar entities such as Japanese Cities or French scientists.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Examples of timelines generated by such methods are shown in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Hence, TLS could serve as a distant reading tool and as a first step in exploring a dataset by provid- ing an overview of its key events.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Moreover, TLS could be combined with search engines and used arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='13479v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='CL] 31 Jan 2023 Figure 1: Examples of generated timelines by Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2021) (left) and Campos et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2018) (right), summarizing a set of documents about respectively Egyptian protests and the Syrian War.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The left timeline outputs a summary on a day-to-day basis, whereas the right timeline lists events using uneven periods of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' as an interface to search results returned by issuing queries over large datasets, as suggested in Swan and Allan (2000);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Alonso et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' From there, the user could zoom into the documents in order to proceed to close reading.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Furthermore, these sum- maries would be presented in chronological order, thus preserving the link between events, and could also be contextualized by adding data from external knowledge bases as in Ceroni et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Search engines augmented with timelines would be especially useful in a Digital Humanities (DH) context such as for facilitating the study of histori- cal datasets, as they would provide necessary con- text to understand past events and to structure the event landscape.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' They could also help the user un- derstand the history of a particular entity such as a person or a location, or even a group of such entities through providing a bird’s-eye view of the relevant data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' A good example of such search engine aug- mented with TLS is the Conta-me Histórias (Tell me stories) platform1, where the user can query news articles from the Portuguese web archive.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The user-friendly interface allows a distant reading of the documents returned by the query through a timeline that summarizes them, but also allows close reading by preserving the link to the original documents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To the best of our knowledge, works on applying TLS methods to structure archives of his- torical documents, or more broadly in the Digital Humanities field, are quite scarce.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='3 Challenges of applying TLS to archives Unfortunately, several aspects of such archives make the application of TLS methods not straight- forward: first, these datasets are often processed with Optical Character Recognition (OCR).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Pre- vious studies have shown that downstream tasks such as Named Entity Recognition (NER), Event Detection (ED) (Boros et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2022), Topic Mod- 1https://contamehistorias.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='pt/arquivopt elling (TM) (Mutuvi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2018) or Named En- tity Linking (NEL) (Linhares Pontes et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2019) are impacted by the quality of the OCR output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To our knowledge, there is no study on the im- pact of OCR on TLS, but we can assume it will be similar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Furthermore, archive collections may also differ from contemporary data because of their temporal context: orthographic rules may dif- fer, places might have changed names (Smith and Crane, 2001) or concepts may have acquired an- other meaning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Most existing annotated resources necessary for NLP components such as NER or ED are created on contemporary data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Historical docu- ments archives are thus harder to process because of this lack of suitable annotated resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Most TLS methods generate timelines through statistical analysis of the input dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' They also often require that the input corpus contains docu- ments of a similar type and similar content.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' How- ever, an archive collection may be heterogeneous and contain documents of different authors, genres, topics and periods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' It may also be fragmentary and not as complete as a contemporary dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Finally, although the timelines generated by TLS systems are often easy to read, the process that created them is often not made explicit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' If timelines must as- sist the study of historical datasets by highlighting important events, they must be interpretable and explain why these events are deemed important.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In this position paper, we propose to extend Time- Line Summarization (TLS) methods to assist in the studies of archive collections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We first present an overview of existing TLS methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We then de- scribe a conceptual framework for an Archive Time- Line Summarization (ATLS) system, which aims to generate informative, readable and interpretable timelines, before suggesting several methods to implement it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This paper is organized as follows: in Section 2 we present an overview of existing TimeLine Date Summary 2011-01-25 Thousands of protesters spilled into the streets of Egypt on Tuesday , an unprecedented display of anti-government rage inspired in part by the tumult in the nearby North African nation of Tunisia.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=" 2011-01-26 Twitter says its site is being blocked in Egypt Egyptians took to the streets in what could be a sequel to the recent revolution in Tunisia witter , Facebook and YouTube were widely used in Tunisia 's uprising and in Iran last year -LRB-." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=" 2011-01-28 With parts of his capital ablaze , Mubarak said he was asking his government to resign and would soon announce a new one , pledging to address the concerns of thousand of Egyptians protesting in Cairo 's streets ." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=" Amre Moussa , the Arab League 's secretary- general and a veteran Egyptian diplomat , joined protesters in Cairo 's Tahrir Square on Friday , state-run Nile TV reported ." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='From ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='To ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Top headlines ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='5/2010 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='7/2011 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Syrian officials launch tear gas against protesters ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Security forces shoot at protesters ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='New York Times journalist with the Pulitzer died of an Asthma attack in Syria ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='8/2011 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='3/2012 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Assad promises elections in February in Syria ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Us withdraws ambassador from Syria for security reasons ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='NATO says goodbye to Libya and the world turns to Syria ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='7/2012 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='12/2012 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Meeting of senior officials in Geneva failed agreement to end violence in Syria ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Russia delivers three war helicopters to Syria ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Red Cross says Syria is in civil war ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='7/2016 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='11/2016 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Maternity unit among hospitals bombed in ldlib air strikes ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='Russian helicopter shot down in Syria.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Turkish army enters SyriaSummarization methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Section 3 and 4, we re- spectively describe our conceptual framework and discuss some of its potential applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Finally, we present our conclusion in Section 5, alongside possibilities for future works.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2 Related Work 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1 TimeLine Summarization Most TLS methods generate timelines by apply- ing the two following steps: the Date Selection step which identifies and ranks the key dates in the documents, and the Date Summarization step which generates a summary of an event occurring at a specific date by picking important sentences in the documents published on that date.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To identify important dates in the dataset, Gholipour Ghalan- dari and Ifrim (2020) select the most frequent date mentions, Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015b) use a graph-ranking model and Kessler et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2012) combine a clus- tering model and a supervised classifier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' For the second step, La Quatra et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2021) apply state-of- the-art methods for Text Summarization (TS) such as TextRank (Mihalcea and Tarau, 2004) whereas Martschat and Markert (2018) adapt methods from the Multi-Document Summarization (MDS) field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' TLS has been generally extractive, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' the sum- mary is created by copying textual elements (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', sentences or paragraphs) from the input data (Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Other works are abstractive, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' the summary is a completely new text generated by the system (Steen and Markert, 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' TLS methods in general tend to be applied to summarize datasets describing large events, such as the Egyptian protests or the Syrian War (Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Martschat and Markert, 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' These meth- ods require that the dataset covers a constrained period of time and is homogeneous, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' that the documents cover the same topic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Standard TLS methods are thus not suited to summarize heteroge- neous or longitudinal datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Some works such as Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Kessler et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2012);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chieu and Lee (2004);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Pasquali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019) can be de- scribed as Query-based TimeLine Summarization (QTLS), as they apply TLS on documents related to a user query such as documents returned by a search engine.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' QTLS generally consists in the two following steps: Event Detection and Event Ranking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To detect events, Chieu and Lee (2004) select any sen- tence where the terms of the query appear, Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014) cluster by a common date every sen- tence returned by the query and Pasquali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019) detect peaks of date occurrences in the time span covered by the documents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Other works train a classifier to detect important events (Chasin, 2010) or rank events by their importance with a Learning- to-Rank model (Ge et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, these classifiers need training data, which are difficult to create since defining what is important is a subjec- tive matter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This can lead to disappointing results as shown in Chasin (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To determine the im- portance of events, Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014) first score them according to their relevancy and saliency to the query, then rerank them to ensure a diverse time- line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chieu and Lee (2004) rank the importance of a sentence according to their "interest" and "bursti- ness", then remove duplicate sentences to ensure diversity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Pasquali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019) use the keyword extractor YAKE!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (Campos et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2018) to weight the terms in the event description.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Duplicate event descriptions are detected with the Levenshtein simi- larity measure and removed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Those methods finally select the top most important events to generate the timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In order to generalize the application of TLS, Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2021) propose a Multiple TimeLine Summa- rization (MTLS) system, which generates a time- line for each story found in the dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To do so, it first detects events mentioned in the dataset and measures their saliency and consistency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' An event linking step determines the link between these events in order to generate each timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Similarly, Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2020) propose the Comparative Time- Line Summarization (CTLS) task, which generates a comparative timeline highlighting the contrast between two timestamped timeline documents (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' biographies, historical sections, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=') by computing local and global importance of events.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' There are few datasets for the TLS task such as 17 Timelines (T17) (Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2013), CRISIS (Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015a), ENTITIES (Gholipour Gha- landari and Ifrim, 2020), CovidTLS (La Quatra et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2021) or TLS-Covid19 (Pasquali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2021) which are constructed from contemporary news ar- ticles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, datasets are often lacking in most projects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' It is then necessary to create a dataset from scratch as in Minard et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Ge et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Bedi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017) or extend existing ones as in Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Due to this lack of datasets, evaluating TLS sys- tems is a difficult task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The date selection step can be evaluated with the F1-measure (La Quatra et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Gholipour Ghalandari and Ifrim, 2020) or with the Mean Average Precision (MAP) metric (Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The date summary is often evaluated with one of the ROUGE metrics (Lin, 2004) to compare a ground-truth timeline and a generated one (Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2014;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Gholipour Ghalandari and Ifrim, 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Methods relying on event detection such as Ge et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Minard et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Bedi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017) often evaluate their system in terms of Precision, Recall and F1-measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, most projects often lack datasets and must then resort to human evaluation as in Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Swan and Allan (2000);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2 TLS Variants We present below formal definitions of several ex- isting TLS variants: TLS: takes as input a standalone homogeneous dataset of timestamped documents D = {d1, d2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', d|D|} and generates a timeline T = {p1, p2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', p|T|} of time-summary pairs pi = (ti, si), where si summarizes important events happening at time ti;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' QTLS: outputs a timeline T = {p1, p2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', p|T|} as a sequence of time-summary pairs pi = (ti, si) from a set of timestamped documents {d1, d2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', d|D|} based on a query Q = {w1, w2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', wk} where wi denotes a word be- longing to the query;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' MTLS: takes as input a dataset of timestamped documents D = {d1, d2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', d|D|} that can be standalone or returned using a query Q = {w1, w2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', wk}, and outputs a set of time- lines T = {T1, T2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', Tm} for each story or topic detected in D, where each time- line Ti is a sequence of time-summary pairs pi = (ti, si);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' CTLS: takes as input two datasets of timestamped documents DA = {d1, d2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', d|DA|} and DB = {d1, d2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', d|DB|} and outputs two timelines TA and TB made of contrasting events detected in DA and DB, each as a se- quence of time-summary pairs pi = (ti, si);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3 Framework In this section, we present a conceptual framework for an Archive TimeLine Summarization (ATLS) which addresses the challenges raised by archive collections such as the sparsity of data, OCR prob- lems, context shifts and linguistic changes over time in order to generate timelines based on these datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We first provide a definition of ATLS and describe the type of dataset expected before presenting the framework and discussing how to evaluate its output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1 Overview The framework consists of the two key steps: Time- line Generation and Timeline Presentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The first step extracts textual elements describing an event and attributes them an importance score.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The second one generates the timeline by filtering events and selecting their description.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The processing stages of the framework are shown in Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The first step has to run only once over the processed dataset, since it aims to detect the elements composing the timeline to be generated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In contrast, the second step can be run multiple times to update the timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2 Problem Definition We define ATLS as follows: Input: A longitudinal dataset of timestamped doc- uments D = {d1, d2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', d|D|} taken from an archival collection, either standalone or re- turned by a query Q = {w1, w2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', wk}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The period of time covered by D is usually much longer than the one typically used in TLS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Output: A timeline T generated from D as a se- quence of time-summary pairs pi = (ti, si), where si summarizes important events hap- pening at time ti.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We compare the key characteristics of TLS and ATLS in Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='3 Expected Dataset The framework takes as input a longitudinal dataset composed of timestamped documents, such as news articles from a historical newspaper collec- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This dataset can be standalone or made of documents returned by a search engine for a given query Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The dataset could be in raw format or have been pre-processed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We would suggest at least the two following pre-processing steps: first, we recommend to clean the dataset if it has been processed with OCR, either manually or semi- automatically, since the OCR quality will impact further steps (Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Secondly, we recommend to detect temporal expressions, as they are a good indicator of event mentions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Tempo- ral expressions are either explicit (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' February 17, 1995) or implicit (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' yesterday, next month).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One can use tools such as HeidelTime (Strötgen Figure 2: Conceptual pipeline for building the ATLS system and Gertz, 2010) or SUTime (Chang and Manning, 2012) to detect temporal expressions in text and re- solve them to an absolute date format, simplifying their use in the TLS process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, we must keep in mind that the detection of temporal expres- sions, especially implicit ones, is still a challenging task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Moreover, available tools such as these were mainly conceived for contemporary data, and thus may not work as properly on historical data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The input dataset could be pre-processed further by applying NLP components such as Name En- tity Recognition (NER), Topic Modelling (TM), Event Extraction (EE), Relation Extraction (RE), Keyword Extraction (KE), or Keyword Generation (KG).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Such annotations could be used to index the dataset and allow the user to query documents about a specific Named Entity or topic, as in the impresso2 or the NewsEye3 platforms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='4 Timeline Generation In this section, we present the first main step of the framework, which extracts mentions of events and attributes them an importance score.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1 Event Detection Although events can be defined in many ways, a commonly accepted definition is "something that is happening or that is holding true in a given cir- cumstance", as stated in the TimeML guidelines Saurí et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Events can be detected in mul- tiple ways: one could detect them through statis- tical analysis of the corpus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' For instance, Chieu and Lee (2004) measure the occurrences of similar sentences associated with the same date, whereas Pasquali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019) measure the occurrences of articles in atomic time intervals to later aggregate them and determine the bursty time periods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' These statistical methods are especially suited for homo- geneous datasets, but may not work as well on heterogeneous or fragmentary datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One could 2https://impresso-project.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='ch/app/ 3https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='newseye.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='eu/ also train a Learning-to-Rank model on summaries created by experts in order to detect important sen- tences as in (Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This would, how- ever, require training data which tend to be scarce, even when for contemporary data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Alternatively, one could use an Event Detection model to detect and annotate events in the dataset as in Chasin (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Event Detection is another task in the NLP community that has been extensively studied, and some previous works such as Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2020) have already applied these methods in humanities contexts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, we need to keep in mind that training such a model requires annotated resources that are often lacking, especially for his- torical data, and that the OCR quality of documents impacts the output of these models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Finally, we could select as event any sentence containing at least a time expression, either explicit or implicit as in Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This selection could be made even finer by taking sentences that also contain a Named Entity as in Abujabal and Berberich (2015);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Bedi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One can then apply algorithms such as Affinity Propagation (Frey and Dueck, 2007) or Chinese Whispers (Biemann, 2006) to gather sen- tences describing the same event as in Rusu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2014);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2021);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Steen and Markert (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Regardless of the method used to detect them, events should all be associated with time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' These could be the time expressions occurring with the event mentions, or the Document Creation Date (DCD) if no time expressions are present.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Alterna- tively, approaches for estimating the focus time of text (Jatowt et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015), in absence of any tempo- ral expressions can be applied to associate event- related sentences with particular points of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2 Event Importance Estimation As mentioned in Section 2, the importance of an event can be measured in a supervised or semi- supervised manner with a classifier (Chasin, 2010;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Timeline Generation step TimelinePresentation step Assign Select filters, Selectevents importance event descriptions, mentioned scoreto each data augmentation Update event E to generate timeline (T) the parameters in D of the timeline 2 3 Input: Dataset (D) Set of events (E) present in D Setofscoredevents (SE) Output: Timeline (T)Ge et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This method, however, requires training data that are difficult to obtain or produce.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Furthermore, the process leading a classifier to a prediction is generally not explained.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Since the goal of this framework is to assist in the study of longitudinal datasets, it is necessary that the pro- cess of generating a timeline is interpretable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Thus, we would suggest to measure the importance score in an unsupervised manner by extracting features from the dataset as in (Nguyen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2014;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chieu and Lee, 2004;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Campos et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Some of the features that we think could help measure this im- portance score are listed below, with suggestions on how to compute them: Redundancy: The more frequently an event is mentioned, the more important it should be.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One can then simply count the occurrences of events, or as an alternative, assign them im- portance weights by calculating their TF-IDF scores over all the time units.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, as the data might be fragmentary in archive datasets, this feature should rather not be used alone;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Contemporary references: an event may be im- portant at a given time if other events occur- ring around the same period of time refer to it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Thus, to evaluate this feature, we could count how often an event is referred to from the descriptions of other events in a given short period of time around that event;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Retrospective references: Similarly, an event is likely to be important if documents keep men- tioning it some time after it occurred.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To as- sess this kind of across-time reference to the event, one could count how often (and perhaps for how long) an event is mentioned by other events that occurred after a given period of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Other solutions may rely on computing random walks over graphs composed of times- tamped events and/or entities to measure the amount of signal propagation from the past to- wards "the recent times" (Jatowt et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2016);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Causality: an event is likely to be important if it is the cause of other events that occurred after it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To evaluate the causality of an event, one could use date reference graphs as in Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015b), which measure the frequency of references, the topical influence and tempo- ral influence between two events to determine a causal link.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' It is also possible to use Causal Relation Extraction (CRE) methods as pre- sented by Gao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' for example.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, the CRE task is far from solved and may re- quire much more dataset pre-processing;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Common sense: some events are clearly more im- portant than other, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' the birth of a child or marrying a partner are usually more impor- tant events in a family history than repainting a house.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To represent that kind of common sense knowledge and compute this feature, it may be necessary to create a dataset of events that are deemed important to train a 1-class classifier (1CC) as in Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019) or a Learning-to-Rank model as in Ge et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Note that while important events can be col- lected from historical textbooks or history- related content, gathering unimportant events may be less easy and more problematic;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' hence the solution could be to rely on a 1CC task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Using these features, a straightforward formula to calculate the importance of an event could be: α · F1 + β · F2 + γ · F3 + δ · F4 + ϵ · F5 where F1, F2, F3, F4, F5 are the scaled values of the features described above and α, β, γ, δ, ϵ are hyper-parameters of which value is defined by the user or document archive custodians.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Similarly to event detection, the user could be asked to select any of these features to compute this score.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Some periods may contain much more docu- ments than others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' For instance, fewer documents may be available during a war time because of censorship or paper restriction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This lack of doc- uments may lead to events that are far more or far less mentioned than others, and bias frequency- based features such as redundancy, contemporary and retrospective references.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Thus, these features should be normalized before being incorporated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Furthermore, we suggest these features since they are easy to compute, but we also acknowledge that they may not be sufficient to measure the im- portance of an event from the perspective of an expert such as a historian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Because the formula to compute the importance score is modular, one could incorporate more features in collaboration with experts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='5 Timeline Presentation In this section, we describe the second main step of the framework, which generates the timeline from events scored in the previous step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We present sets of filters to select which events should appear on the timeline and how they should be presented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We also describe an optional step of timeline augmen- tation using external data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1 Event Filtering A dataset may contain hundreds or thousands of mentioned events.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' It is necessary to select those that will be added to the timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To do so, we can use filters such as described below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The weight of these filters could be changed on the user interface, thus allowing users to instantly update the timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Top N: top N most important events are retained;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Importance Threshold (IT): only events of which the importance score is superior to a pre-fixed threshold IT are taken.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Individual thresholds for the features described in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='4 that make up the importance score can also be set;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Topical Diversity Threshold (TopDT): removes redundant event mentions and ensures the timeline is topically diverse.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Topical diversity can be simply measured using Maximal Marginal Relevance (MMR) (Goldstein-Stewart and Carbonell, 1998) or the n-gram blocking metric as in Liu (2019);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Temporal Diversity Threshold (TempDT) : ensures every time unit on the generated timeline is evenly represented by setting a minimum and maximum number of events that can appear at each time unit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2 Event Description Selection There are multiple ways to represent an event on a timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One could select a sentence that describes the event.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' If this sentence is too long, one could use sentence compression methods (Filippova and Strube, 2008) to only keep its most important part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' As mentioned earlier, an event might be represented by a cluster of sentences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The user can thus select one sentence among this cluster or generate a cloud of terms of all sentences contained in it, as in Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One could also use headlines if the tar- get documents are articles as in Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2015a);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Pasquali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Finally, we could also use a Natural Language Generation (NLG) system as in Steen and Markert (2019), as these generated texts are often easier to understand than text extracted from the docu- ments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, abstractive methods such as these may suffer from inaccuracies or hallucinations, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' generate information that is not present in the orig- inal documents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Thus, abstractive methods might generate improper event descriptions and lose the connection with the original documents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' On the other hand, a common drawback of purely extrac- tive methods is that selected sentences may require some context or at least post-processing for users to be able to properly understand them (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' pronouns may need to be resolved or we need to add defini- tions or descriptions of some entities or events).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='3 Timeline Augmentation To properly understand them, some events may re- quire contextual knowledge that is missing from the processed dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This can especially happen if the user is not a domain expert.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Such contextual knowledge may be found in knowledge bases such as Wikidata or Wikipedia Year pages (see for exam- ple (Tran et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2015c)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Thus, timelines generated by an ATLS system could be augmented with con- textual data provided by external knowledge bases as in (Ceroni et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' These augmented time- lines could help in explaining a dataset by summa- rizing it and providing the user with the necessary knowledge to understand it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Unfortunately, most resources created by experts are not in a machine- readable format (Gutehrlé et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Hence, this step may require more effort.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='6 Timeline Evaluation As mentioned earlier, the evaluation of a TLS sys- tem is a difficult task because of the lack of evalu- ation datasets and the inherent subjectivity of the task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In order to evaluate the output, we would suggest to manually assess the produced timelines, either by following some evaluation criteria as in Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017), or by comparing them with resources created by experts such as timelines de- rived from history books as in Bedi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' One could also use this framework to bootstrap an evaluation dataset specific to the given corpus, towards an automatic evaluation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 4 Discussion In this section, we describe two hypothetical use cases comparing the application of TLS and ATLS systems, and compare in Table 1 the types of datasets and timelines both methods can process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Finally, we discuss potential extensions of ATLS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1 Use cases In the first hypothetical use case, a user has curated a homogeneous dataset of timestamped documents from Web archives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This dataset is made of news articles related to a story spanning over a year.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' It has been pre-processed to remove HTML tags and TLS ATLS Covered period Shorter Longer Input Data Size Small / Medium Large Documents type Timestamped documents (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' news articles) Document Format Usually born digital Often digitized Data Integrity Usually complete Can be fragmentary Presence of noise Less likely Depends on OCR quality Semantic evolution Less common Possible (esp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' over long time) Need for query-based filtering Optional (depends on data size and heterogeneity) Need for contextualization Less likely More likely (esp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' over long time) Need for interpretable output Yes Table 1: Comparison of TLS and ATLS tasks extract temporal expressions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To generate the time- line, the user applies the TLS method: important dates are first selected before generating a sum- mary of events occurring at each date.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The user can select the sentence mentioning the event, the headline of the article or apply abstractive methods to generate its description.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In the second hypothetical use case, a user has curated a heterogeneous corpus to study the eco- nomical life of a certain French region in the 20th century.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This corpus is composed of periodicals, newspapers and magazines from different sources (parishes, libraries, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=') published over a cen- tury and processed with OCR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This dataset has also been pre-processed: the documents have been first cleaned of OCR errors, then automatically an- notated with Temporal Expression Extraction and Named Entity Recognition components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Further- more, the dataset has been indexed so as to allow query-based searching.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' To generate the timeline, the user applies some of the ATLS approaches men- tioned in this paper: events are first detected by clustering similar sentences that contain a temporal expression and a Named Entity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The importance of these events is then scored using the formula described in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The timeline is gen- erated by setting high values to the topical and temporal diversity thresholds, and augmented with external data from Wikidata, so as to ensure a com- prehensive and contextualized timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Similarly, the user can select from the user interface to use a cloud of terms or a sentence from the cluster to generate event descriptions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2 Extensions of ATLS systems Timelines are usually represented linearly, where each time unit is of the same size (usually a day or a year).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' However, the optimal granularity of tem- poral units might vary when generating a timeline over a long period of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' For example, when referring to a distant past, humans tend to often de- scribe entire decades or years rather than discussing each day or month which is more common for the recent past.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Furthermore, events mentioned in his- torical documents might not always be recorded with the same temporal precision (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', some events may have missing dates, the dates can be impre- cise or difficult to be inferred).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' A possible solution would be to generate logarithmic timelines, where the granularity of the time unit changes over time, as suggested in Jatowt and Au Yeung (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' If the documents in the datasets are annotated with Named Entities, one could generate entity- based timelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This could help understand the history of a specific entity such as a person or a location as in Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' This idea could be extended by generating aggregate timelines for multiple entities at the same time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' These timelines could be agglomerative or contrastive and respec- tively show the similarities and differences between the history of multiple entities of the same type (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', cities in the same region or country, scientists of the same area).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Similar to Duan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' (2020), such comparative timelines would allow to study the history of entities of the same or similar type, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Berlin vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Paris or even entities of different types, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Paris and the writer Victor Hugo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 5 Conclusion TimeLine Summarization can be a useful tool for getting an overview of historical collections as well as it can serve as a novel information access means to news article archives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In this position paper, we have presented an overview of existing TLS methods and described a conceptual framework for Archive TimeLine Summarization systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' The implementation of the framework outlined in this paper will be the subject of our future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' We also intend to ask humanities scholars (his- torians, archivists, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=') to evaluate the quality of generated timelines and the effectiveness of our framework for the study of archive collections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' References Abdalghani Abujabal and Klaus Berberich.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Im- portant events in the past, present, and future.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' pages 1315–1320.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' James Allan, Ron Papka, and Victor Lavrenko.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' On-line new event detection and tracking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Pro- ceedings of the 21st Annual International ACM SI- GIR Conference on Research and Development in Information Retrieval, SIGIR ’98, page 37–45, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computing Machin- ery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Omar Alonso, Stefano Marchesin, Marc Najork, and Gianmaria Silvello, editors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Proceedings of the Second International Conference on Design of Experimental Search & Information REtrieval Sys- tems, Padova, Italy, September 15-18, 2021, vol- ume 2950 of CEUR Workshop Proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' CEUR- WS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='org.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Harsimran Bedi, Sangameshwar Patil, Swapnil Hing- mire, and Girish Palshikar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Event time- line generation from history textbooks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Pro- ceedings of the 4th Workshop on Natural Lan- guage Processing Techniques for Educational Appli- cations (NLPTEA 2017), pages 69–77, Taipei, Tai- wan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Asian Federation of Natural Language Process- ing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chris Biemann.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chinese whispers - an efficient graph clustering algorithm and its application to nat- ural language processing problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of TextGraphs: the First Workshop on Graph Based Methods for Natural Language Processing, pages 73–80, New York City.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computa- tional Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Emanuela Boros, Nhu Khoa Nguyen, Gaël Lejeune, and Antoine Doucet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Assessing the impact of OCR noise on multilingual event detection over digitised documents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' International Journal on Digi- tal Libraries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Ricardo Campos, Vítor Mangaravite, Arian Pasquali, Alípio Mário Jorge, Célia Nunes, and Adam Ja- towt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yake!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' collection-independent auto- matic keyword extractor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Advances in Informa- tion Retrieval, pages 806–810, Cham.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Springer In- ternational Publishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Andrea Ceroni, Nam Khanh Tran, Nattiya Kanhabua, and Claudia Niederée.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Bridging temporal con- text gaps using time-aware re-contextualization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 37th International ACM SIGIR Conference on Research & Development in Informa- tion Retrieval, SIGIR ’14, page 1127–1130, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computing Machin- ery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Angel X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chang and Christopher Manning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' SU- Time: A library for recognizing and normalizing time expressions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the Eighth In- ternational Conference on Language Resources and Evaluation (LREC’12), pages 3735–3740, Istanbul, Turkey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' European Language Resources Association (ELRA).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Rachel Chasin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Event and temporal information extraction towards timelines of wikipedia articles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Hai Leong Chieu and Yoong Keok Lee.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Query based event extraction along a timeline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceed- ings of the 27th Annual International ACM SIGIR Conference on Research and Development in Infor- mation Retrieval, SIGIR ’04, page 425–432, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computing Machin- ery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yijun Duan, Adam Jatowt, and Katsumi Tanaka.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Discovering typical histories of entities by multi- timeline summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 28th ACM Conference on Hypertext and Social Media, HT ’17, page 105–114, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' As- sociation for Computing Machinery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yijun Duan, Adam Jatowt, and Katsumi Tanaka.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' History-driven entity categorization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Web and Big Data, pages 349–364, Cham.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Springer International Publishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yijun Duan, Adam Jatowt, and Masatoshi Yoshikawa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Comparative timeline summarization via dy- namic affinity-preserving random walk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In ECAI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Katja Filippova and Michael Strube.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Depen- dency tree based sentence compression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In INLG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Brendan J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Frey and Delbert Dueck.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Clustering by passing messages between data points.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Science, 315(5814):972–976.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Lei Gao, Prafulla Kumar Choubey, and Ruihong Huang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Modeling document-level causal structures for event causal relation identification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Proceedings of the 2019 Conference of the North American Chap- ter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Pa- pers).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Tao Ge, Wenzhe Pei, Heng Ji, Sujian Li, Baobao Chang, and Zhifang Sui.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Bring you to the past: Automatic generation of topically relevant event chronicles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 53rd Annual Meeting of the Association for Computational Lin- guistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 575–585, Beijing, China.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Associa- tion for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Demian Gholipour Ghalandari and Georgiana Ifrim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Examining the state-of-the-art in news time- line summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 58th An- nual Meeting of the Association for Computational Linguistics, pages 1322–1334, Online.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Jade Goldstein-Stewart and Jaime G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Carbonell.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Summarization: (1) using MMR for Diversity- Based Reranking and (2) Evaluating Summaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In TIPSTER.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Nicolas Gutehrlé, Oleg Harlamov, Farimah Karimi, Haoyu Wei, Axel Jean-Caurant, and Lidia Pivo- varova.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' SpaceWars: A Web Interface for Exploring the Spatio-temporal Dimensions of WWI Newspaper Reporting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' CEUR Workshop Proceed- ings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Adam Jatowt and Ching-man Au Yeung.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Ex- tracting collective expectations about the future from large text collections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 20th ACM International Conference on Informa- tion and Knowledge Management, CIKM ’11, page 1259–1264, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computing Machinery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Adam Jatowt, Daisuke Kawai, and Katsumi Tanaka.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Predicting importance of historical per- sons using wikipedia.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 25th ACM International Conference on Information and Knowledge Management, CIKM 2016, Indianapolis, IN, USA, October 24-28, 2016, pages 1909–1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' ACM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Adam Jatowt, Ching-man Au Yeung, and Katsumi Tanaka.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Generic method for detecting fo- cus time of documents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Manag.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 51(6):851–868.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Rémy Kessler, Xavier Tannier, Caroline Hagège, Véronique Moriceau, and André Bittar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Find- ing salient dates for building thematic timelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In ACL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Moreno La Quatra, Luca Cagliero, Elena Baralis, Al- berto Messina, and Maurizio Montagnuolo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Summarize Dates First: A Paradigm Shift in Time- line Summarization, page 418–427.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computing Machinery, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Chin-Yew Lin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' ROUGE: A package for auto- matic evaluation of summaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Text Summariza- tion Branches Out, pages 74–81, Barcelona, Spain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Elvys Linhares Pontes, Ahmed Hamdi, Nicolas Sidère, and Antoine Doucet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Impact of OCR Quality on Named Entity Linking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In International Confer- ence on Asia-Pacific Digital Libraries 2019, Kuala Lumpur, Malaysia.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yang Liu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Fine-tune BERT for extractive sum- marization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Sebastian Martschat and Katja Markert.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' A tem- porally sensitive submodularity framework for time- line summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 22nd Conference on Computational Natural Language Learning, pages 230–240, Brussels, Belgium.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Asso- ciation for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Rada Mihalcea and Paul Tarau.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' TextRank: Bringing order into text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 2004 Conference on Empirical Methods in Natural Lan- guage Processing, pages 404–411, Barcelona, Spain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Anne-Lyse Minard, Manuela Speranza, Eneko Agirre, Itziar Aldabe, Marieke van Erp, Bernardo Magnini, German Rigau, and Rubén Urizar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' SemEval- 2015 task 4: TimeLine: Cross-document event or- dering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 778–786, Denver, Colorado.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Stephen Mutuvi, Antoine Doucet, Moses Odeo, and Adam Jatowt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Evaluating the Impact of OCR Errors on Topic Modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Maturity and Innova- tion in Digital Libraries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 20th International Confer- ence on Asia-Pacific Digital Libraries, ICADL 2018, Hamilton, New Zealand, November 19-22, 2018, Proceedings, pages 3 – 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Kiem-Hieu Nguyen, Xavier Tannier, and Veronique Moriceau.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Ranking multidocument event de- scriptions for building thematic timelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Pro- ceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Tech- nical Papers, pages 1208–1217, Dublin, Ireland.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Dublin City University and Association for Compu- tational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Nhu Khoa Nguyen, Emanuela Boros, Gaël Lejeune, and Antoine Doucet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Impact analysis of document digitization on event extraction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Pro- ceedings of the 4th Workshop on Natural Language for Artificial Intelligence (NL4AI 2020) co-located with the 19th International Conference of the Ital- ian Association for Artificial Intelligence (AI*IA 2020), Anywhere, November 25th-27th, 2020, vol- ume 2735 of CEUR Workshop Proceedings, pages 17–28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' CEUR-WS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='org.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Thi Tuyet Hai Nguyen, Adam Jatowt, Mickaël Cous- taty, and Antoine Doucet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Survey of Post- OCR processing approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' ACM Comput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Surv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=', 54(6):124:1–124:37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Arian Pasquali, Ricardo Campos, Alexandre Ribeiro, Brenda Salenave Santana, Alípio Mário Jorge, and Adam Jatowt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Tls-covid19: A new annotated corpus for timeline summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In ECIR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Arian Pasquali, Vítor Mangaravite, Ricardo Campos, Alípio Mário Jorge, and Adam Jatowt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Inter- active system for automatically generating temporal narratives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Advances in Information Retrieval, pages 251–255, Cham.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Springer International Pub- lishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Delia Rusu, James Hodson, and Anthony Kimball.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Unsupervised techniques for extracting and clustering complex events in news.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceed- ings of the Second Workshop on EVENTS: Definition, Detection, Coreference, and Representation, pages 26–34, Baltimore, Maryland, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Roser Saurí, Jessica Moszkowicz, Bob Knippen, Rob Gaizauskas, Andrea Setzer, and James Pustejovsky.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Timeml annotation guidelines version 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' David A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Smith and Gregory Crane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Disam- biguating geographic names in a historical digital library.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 5th European Con- ference on Research and Advanced Technology for Digital Libraries, ECDL ’01, page 127–136, Berlin, Heidelberg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Springer-Verlag.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Julius Steen and Katja Markert.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Abstractive time- line summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 2nd Work- shop on New Frontiers in Summarization, pages 21– 31, Hong Kong, China.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computa- tional Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Jannik Strötgen and Michael Gertz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' HeidelTime: High quality rule-based extraction and normaliza- tion of temporal expressions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 5th International Workshop on Semantic Evaluation, pages 321–324, Uppsala, Sweden.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Russell Swan and James Allan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Automatic gen- eration of overview timelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 23rd Annual International ACM SIGIR Confer- ence on Research and Development in Information Retrieval, SIGIR ’00, page 49–56, New York, NY, USA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computing Machinery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Giang Tran, Mohammad Alrifai, and Eelco Herder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Timeline summarization from relevant head- lines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Advances in Information Retrieval, pages 245–256, Cham.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Springer International Publishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Giang Tran, Eelco Herder, and Katja Markert.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Joint graphical models for date selection in timeline summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 53rd Annual Meeting of the Association for Computational Lin- guistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1598–1607, Beijing, China.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Associa- tion for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Giang Binh Tran, Tuan Tran, Nam Khanh Tran, Mohammad Alrifai, and Nattiya Kanhabua.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Leveraging learning to rank in an optimization framework for timeline summarization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Nam Khanh Tran, Andrea Ceroni, Nattiya Kanhabua, and Claudia Niederée.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2015c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Back to the past: Supporting interpretations of forgotten stories by time-aware re-contextualization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the Eighth ACM International Conference on Web Search and Data Mining, WSDM 2015, Shanghai, China, February 2-6, 2015, pages 339–348.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' ACM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Yi Yu, Adam Jatowt, Antoine Doucet, Kazunari Sugiyama, and Masatoshi Yoshikawa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Multi- TimeLine summarization (MTLS): Improving time- line summarization by generating multiple sum- maries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' In Proceedings of the 59th Annual Meet- ing of the Association for Computational Linguistics and the 11th International Joint Conference on Nat- ural Language Processing (Volume 1: Long Papers), pages 377–387, Online.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
+page_content=' Association for Computa- tional Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/L9FRT4oBgHgl3EQfFTdj/content/2301.13479v1.pdf'}
diff --git a/LNE1T4oBgHgl3EQfGwM1/content/2301.02917v1.pdf b/LNE1T4oBgHgl3EQfGwM1/content/2301.02917v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..47d92ea48d297759d95337b698f442c7d381fb60
--- /dev/null
+++ b/LNE1T4oBgHgl3EQfGwM1/content/2301.02917v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2864d77368821e07a7368011dc77b62a63582aeb6a400d0735edd9164efcc111
+size 319752
diff --git a/LNE1T4oBgHgl3EQfGwM1/vector_store/index.faiss b/LNE1T4oBgHgl3EQfGwM1/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..8baea7ab970e83368a8b1b74577360dc3866d1c8
--- /dev/null
+++ b/LNE1T4oBgHgl3EQfGwM1/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41093589e354a3b15b671e5bebcec98900fee73c8bded16364762364d48fad78
+size 1507373
diff --git a/LNE1T4oBgHgl3EQfGwM1/vector_store/index.pkl b/LNE1T4oBgHgl3EQfGwM1/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..5d92a846108dab8d3ab865415d44738dfb987d9a
--- /dev/null
+++ b/LNE1T4oBgHgl3EQfGwM1/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05b343d815fe54d689570d2d7ef03f9153d3ce5f34077e2322cc9a82093efc23
+size 66060
diff --git a/LdFAT4oBgHgl3EQfwh69/content/tmp_files/2301.08682v1.pdf.txt b/LdFAT4oBgHgl3EQfwh69/content/tmp_files/2301.08682v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6c9464f92fdd3d95c8bea4f2620b019b2f6a1696
--- /dev/null
+++ b/LdFAT4oBgHgl3EQfwh69/content/tmp_files/2301.08682v1.pdf.txt
@@ -0,0 +1,1609 @@
+
+1
+Using Ensemble Monte Carlo Methods to
+Evaluate Non-Equilibrium Green's Functions
+
+David K. Ferry
+School of Electrical, Computer, and Energy Engineering, Arizona State University, Tempe, AZ
+25287-6206; ferry@asu.edu
+
+
+Abstract
+The use of ensemble Monte Carlo (EMC) methods for the simulation of transport in semiconductor
+devices has become extensive over the past few decades. This method allows for simulation
+utilizing particles while addressing the full physics within the device, leaving the computational
+difficulties to the computer. More recently, the study of quantum mechanical effects within the
+devices, effects which also strongly affect the carrier transport itself, have become important. While
+particles have continued to be useful in quantum simulations using Wigner functions, interest in
+analytical solutions based upon the non-equilibrium Green's functions (NEGF) have become of
+greater interest in device simulation. While NEGF has been adopted by many commercial
+semiconductor, there remains considerable computational difficulty in this approach. Here, a
+particle approach to NEGF is discussed, and preliminary results presented illustrating the
+computational efficiency that remains with the use of particles. This approach adopts the natural
+basis functions for use in a high electric field and the preliminary results are obtained for quantum
+transport in Si at 300 K. This approach appears to offer significant advantages for the use of NEGF.
+
+Keywords: quantum transport, Non-equilibrium Green's functions, ensemble Monte Carlo, Airy transform
+
+
+1. Introduction
+
+In the classical world, transport in applied electric and
+magnetic fields is usually computed with the
+Boltzmann equation. In low fields, one has a nearly
+equilibrium situation in which the distribution
+function is a Maxwellian or a Fermi-Dirac at a
+temperature that likely increases above that of the
+lattice. In most devices, the distribution is well out of
+equilibrium, which means that it is unknown. Finding
+this distribution is typically the single most difficult
+problem. At least classically, an alternative approach
+is to use the computer to completely solve the transport
+problem with a stochastic methodology, and several
+methods have arisen to do this, two of which are an
+integral iteration technique [1,2] and the Monte Carlo
+method [3]. Today, the most widely used approach is
+the latter. The ensemble Monte Carlo (EMC)
+technique uses an ensemble of particles whose
+propagation is determined in parallel, so that ensemble
+averages can be computed as a function of time for the
+various observables of interest. It has been the subject
+of many reviews [4,5,6,7]. The methodology of the
+EMC approach is to spend time and effort on
+characterizing the correct physics, perhaps even to the
+use of atomistic full-band energy behavior [8], and the
+scattering properties, and let the computational burden
+be handled by a modern parallel processing computer.
+Hence, no attempt is made to arrive at an ab initio
+analytical form. These approaches have become
+extremely sophisticated, and many factors such as
+degeneracy [9], discrete impurities in real space [10],
+and many-body effects [11] can be incorporated.
+
+From the beginning, there has been a rich history
+for the use of particles in quantum mechanics, as the
+suggestion of particles and waves dates to de Broglie
+[12], and the use of particles has been discussed over
+the years within the quantum world from Kennard [13]
+to Feynman [14]. The immediate problem with
+adapting the EMC approach for quantum transport lies
+with the use of the particle paths in phase space in the
+classical approach. Generally, the uncertainty
+principle restricts being able to simultaneously define
+both a position and momentum for a particle path.
+Nevertheless, this strict interpretation has been
+finessed by a variety of approaches, many of which
+clearly endorse some form of a real path for a particle
+that exists with the wave. Not the least of these
+approaches is due to Feynman himself. One more
+successful approach is the phase-space representation
+
+
+2
+of the Wigner function [15], as this suggests a
+comparison between quantum dynamics and the
+corresponding classical motion [16,17]. However, the
+Wigner function can have a non-unitary evolution, and
+off-diagonal terms in the density matrix (from which
+the Wigner function is found) lead to non-classical
+propagation. Indeed, quantum coherence is reflected
+in oscillatory behavior of the Wigner function and
+non-positive definite regions in phase space, and this
+leads to complications in the use of EMC.
+Nevertheless, particle methods and the EMC have
+been adapted to treating quantum transport via the
+Wigner function [18,19,20].
+
+Quite generally, the operation and performance of
+semiconductor devices has evolved to incorporate
+many quantum effects [21]. Any quantum mechanical
+simulation of such a device has to meet all the
+requirements of a self-contained and consistent
+mathematical theory, just as in the classical case. But
+it generally reflects a much deeper physical behavior.
+Certainly, a variety of quantum approaches to the
+simulation of transport have arisen, including the
+Wigner function mentioned above. Indeed, there have
+been many attempts to try to put quantum mechanics
+into the normal kinetic equations stemming from the
+Boltzmann approach, although most limit the method
+to approximating the spectral density that replaces the
+energy-conserving delta function for scattering
+[22,23,24,25]. Such an approach has also been taken
+with Bohmian trajectories and a quantum potential
+[26]. However, today one of the most popular
+approaches to quantum transport and quantum
+distributions is thought to be non-equilibrium Green’s
+functions (NEGF). As in the classical case where the
+system is quite the same, the NEGF include new
+functions that must be found to describe both the non-
+equilibrium distribution function, but also how
+quantum correlations are built into the device [27,28].
+These Green’s functions also determine the transport
+of an "excitation" over a certain distance to where it is
+annihilated (thus these functions contain two times and
+two spatial variables). As a result, these functions
+bring considerable computational difficulty to any
+transport problem, with consequent long computation
+times. Nevertheless, they are being used extensively in
+modern device simulation and design.
+
+It is known in quantum mechanics that any
+convenient coordinate system, or set of basis
+functions, may be used for any given problem, since
+the Schrödinger equation is a linear equation. In this
+paper, the use of a particular set of basis set is adopted
+that is extremely useful in a high electric field. These
+constitute the Airy transform, after which one can
+obtain an integral equation for the important "less-
+than" Green's function that is amenable to solution via
+EMC techniques. While the approach is currently
+restricted to uniform fields and materials such as Si, it
+can be extended easily to the inhomogeneous world of
+semiconductor devices. In the following section, a
+brief introduction to NEGF, and the limitations of
+these functions, will be provided. Then, in Secs. 3 and
+4, the Airy transform approach will be described and
+the various NEGF developed with this transform and
+the case of Si. The restrictions to Si will also be
+discussed. Section 5 will describe the EMC
+simulations and the results. Finally, in Sec. 6, these
+results will be discussed and prospects for extending
+the approach to inhomogeneous devices and to other
+materials will appear.
+
+2. NEGF
+
+In general, it is a well-known relationship in quantum
+mechanics that the wave function can be connected to
+initial conditions through the existence of a propagator
+as
+
+
+𝜓(𝒙, 𝑡) = ∫ 𝑑𝒙′𝐾(𝒙, 𝒙′; 𝑡, 0)𝜓(𝒙′, 0) ,
+(1)
+where one can write the propagator as
+
+ 𝐾(𝒙, 𝒙′; 𝑡, 0) = ∑ 𝜑!
+"(𝒙′)𝜑!(𝒙)
+!
+𝑒#$!%/ℏ ,
+(2)
+and the 𝜑!(𝒙) are basis functions in a spatial expansion
+of the total wave function. It was Schwinger who
+formally connected this propagator to Green’s
+functions, which were already well-known in the
+mathematics of differential equations [29]. Initially,
+these were independent of time, as he was concerned
+with the equilibrium state, and he assumed time
+reversibility, hence sticking to an equilibrium system.
+This equilibrium was maintained, even in the presence
+of perturbations, by the equality of 𝑡 → ±∞. These
+perturbations were handled formally by the existence
+of a unitary operator that involved the perturbing
+Hamiltonian exponentially [30], although this operator
+was developed much earlier [31]. The important
+Green's functions in this equilibrium theory were the
+advanced and retarded functions given as
+
+
+𝐺((𝑟, 𝑟); 𝑡, 𝑡′) = 𝑖Θ(𝑡* − 𝑡)
+ × 〈{𝜓"(𝑟), 𝑡′), 𝜓(𝑟, 𝑡)}〉
+𝐺+(𝑟, 𝑟); 𝑡, 𝑡′) = −𝑖Θ(𝑡 − 𝑡′)
+ × 〈{𝜓(𝑟, 𝑡), 𝜓"(𝑟), 𝑡′)}〉
+.
+(3)
+Here, the curly brackets denote the normal anti-
+commutator relationships for fermions, while the
+angular brackets define normalization conditions,
+obtained with the equilibrium distribution.
+
+It is a well-recognized fact that the application of
+an electric field leads to current and a phase-transition
+that breaks the time-reversal symmetry of these
+functions [32]. This then leads to the non-equilibrium
+state. Normalization cannot be done with the
+
+
+3
+equilibrium distribution and one needs two new
+Green's functions, the correlation functions, in order to
+gain
+the
+non-equilibrium
+distribution.
+These
+correlation functions are given by
+
+
+𝐺,(𝒙, 𝒙′; 𝑡, 𝑡′) = 𝑖〈𝜓?"(𝒙), 𝑡′)𝜓?(𝒙, 𝑡)〉
+𝐺-(𝒙, 𝒙′; 𝑡, 𝑡′) = −𝑖〈𝜓?(𝒙, 𝑡)𝜓?"(𝒙), 𝑡′)〉 ,
+(4)
+and are known as the "less than" and "greater than"
+Green's functions, respectively. These are direct wave
+function products and do not have the anti-
+commutators that appear in (3). The important one for
+our purposes is the less-than function, which has an
+important relation to the distribution function via
+
+
+𝐺,(𝑘, 𝜔) = 𝑓(𝜔)𝐴(𝑘, 𝜔) ,
+(5)
+which is known as the Kadanoff-Baym ansatz [28].
+There are some views that this does not satisfy
+causality, and another form is required [33]. However,
+precisely this form will be found with the Airy
+transforms and will allow the determination of an
+equation for the distribution function 𝑓(𝜔). The
+spectral density 𝐴(𝑘, 𝜔) retains exactly the same form
+as in equilibrium and is connected to the retarded and
+advanced functions through
+
+ 𝐴(𝑘, 𝜔) = −𝐼𝑚{𝐺+(𝑘, 𝜔) − 𝐺((𝑘, 𝜔)} .
+(6)
+
+Each of these functions requires a pair of equations of
+motion. With the added complications of having four func-
+tions, these equations will be far more complicated than the
+simple Boltzmann equation. Hence, solving for these
+functions, and obtaining the distribution function in the
+quantum case, is a difficult problem, both analytically and
+computationally [21]. Indeed, it would be much easier if an
+approach based, for example, on EMC could be found. But,
+the computational problems are not the only problems with
+NEGF, for the equations needed are not limited to just these
+functions. In the quantum transport situation, one must deal
+with the existence of long-range correlations among the
+carriers, correlations that are not always alleviated with
+scattering. An example is given by impurity scattering, which
+is very long range due to the Coulomb interaction potential.
+Even classically, it has been observed that an electron can be
+interacting with several impurities at the same time [34]. In
+addition, with the wave accompanying the particle, a single
+Coulomb scattering center can create interference, much like
+a two-slit experiment, a result that has been found with both
+Green's functions [35] and Wigner functions [36]. These
+correlations affect the quantum transport and can lead to
+observable fluctuations in devices [37]. Such long range
+correlations are not limited to impurity scattering. They are
+expected for any long range potential; polar optical scattering
+is also Coulombic in nature and can be expected to lead to the
+same effects. Consequently, such correlations lead to
+difficulties
+when
+these
+scatterers
+are
+introduced
+perturbatively, as they require minimally evaluation of the
+Bethe-Salpeter equation for determination of mobility and
+transport [21,38]. This raises the difficulty level by a
+significant amount [39].
+
+3.
+Airy Transforms
+
+One of the first requirements in quantum transport is
+the choice of wave functions and the corresponding
+basis set. In the presence of an electric field that is
+taken along the z-direction for convenience, the one-
+dimensional eigenfunctions along this direction are
+determined from (time variation is assumed to enter in
+the normal manner, and only the z-variation is
+considered at the moment)
+
+
+!−
+ℏ!
+"#∗
+$!
+$%! + 𝑒𝐹𝑧'𝜑(𝑧) = 𝐸𝜑(𝑧) ,
+(7)
+where the symbol F is used to denote the field in order
+to differentiate it from the energy E. The wave
+functions in the two orthogonal directions (x,y) are
+taken to be normal plane waves that exist in the
+transverse band structure. For this latter, the normal k
+and r will denote the wave vector and position in the
+transverse plane of (x,y)-coordinates. As may be found
+in many quantum mechanics textbooks, the solutions
+to (7) are the Airy functions [40]
+
+
+𝜑(𝒓, 𝑧) =
+.
+/01 𝑒!𝒌∙𝒓𝐴𝑖 H−
+5#5"
+1 I ,
+(8)
+where 𝑧* is an appropriate zero of the Airy function
+𝐴𝑖(∙). The factor L is a normalizing length and is given
+by
+
+
+𝐿 = H
+ℏ#
+/6∗78I
+./9
+ .
+(9)
+The basis set of Airy functions has been used to find
+the analytical eigen-values for the triangular potential
+wells in a MOSFET for many years [40]. In the
+MOSFET case, this provides the set of discrete bound
+state eigen-values and eigenfunctions determined by
+the discrete zeroes of the Airy function. This is not the
+case of interest here.
+
+In the present situation, a continuous Airy
+transform is to be used. The difference is exactly the
+same as the difference between a Fourier series and a
+Fourier transform. The former is a discrete set while
+the latter is a continuous function. In the present case,
+s well be the continuous transform variable and the
+discrete set of zeroes will not appear explicitly. This
+transform is defined just as a Fourier transform would
+be defined, the Airy transform of the function 𝑓(𝒓, 𝑧)
+is given as
+
+ 𝐹(𝒌, 𝑠) = ∫ 𝑑/𝒓 ∫
+:5
+/01 𝑒!𝒌∙𝒓𝐴𝑖 H
+5#;
+1 I 𝑓(𝒓, 𝑧) . (10)
+For the Green’s function, there are two spatial
+variables, and hence one needs a double Airy
+
+
+4
+transform, although it is fair to assume homogeneity in
+the transverse dimensions. This transform may then be
+expressed as
+
+
+𝐹(𝒌, 𝑠) = ∫ 𝑑/𝒓 ∫
+:5
+/01 ∫
+:5%
+/01 𝑒!𝒌∙𝒓
+ × 𝐴𝑖 H
+5#;
+1 I 𝐴𝑖 H
+5%#;%
+1 I 𝑓(𝒓, 𝑧, 𝑧′)
+. (11)
+Here, the Green’s function response is only being
+developed for the one dimension along the electric
+field. One could, of course, also extend the equations
+to two variables in the transverse directions, but here
+only the one will be used to ease the complexity of the
+resulting equations. A function that is diagonal in both
+k and s is translationally invariant in the transverse
+plane, but not necessarily in the longitudinal plane
+along the electric field.
+
+4. Airy Form of NEGF
+
+The development of Green's function transport using
+the Airy functions is relatively old [41,42,43], and the
+present treatment is focused upon the development of
+the EMC method of its evaluation. Consequently, the
+details of various derivations will be omitted here, but
+can also be found in [39].
+
+4.1 The Retarded and Advanced Functions
+
+In the transverse dimensions, the Green’s function is
+translationally invariant and so is a function only of
+𝒓 − 𝒓′. Solving the differential equation for the
+retarded and advanced functions leads to an
+integration over the intermediate variables which
+appears as a convolution integral. After the transforms
+are taken, the convolution becomes a simple product,
+so that the retarded Green’s function satisfies the
+inhomogeneous integral equation
+
+𝐺+(𝒌, 𝑢, 𝑢′) = 𝐺*
+8(𝒌, 𝑢, 𝑢′)
+ + ∫𝑑𝑢. ∫ 𝑑𝑢/𝐺*
+8(𝒌, 𝑢, 𝑢.)
+ × Σ+(𝒌, 𝑢., 𝑢/)𝐺+(𝒌, 𝑢/, 𝑢′)
+. (12)
+Here, the short-hand notation 𝑢 = (𝑧, 𝑡) has been
+introduced. In fact, in the case of significant
+correlation among the carriers, the argument of the
+integrals could properly be considered a three-particle
+Green's function, which would be incredibly difficult
+to evaluate [14,44,45]. In most cases, it is assumed that
+it can be rewritten as the product of three single-
+particle functions as shown. These can also lead to
+disconnected diagrams which cannot be discarded in
+the NEGF cases [31,46], as these generate important
+phase factors that affect any correlation and/or
+interference terms [47]. These higher-order effects
+must be considered and will be discussed in the Si
+model. The unperturbed Green’s function, the first
+term on the right-hand side of (12), includes the role
+of the electric field and may be taken to be [48]
+
+𝐺*
+8(𝒌, 𝑠, 𝑠), 𝑡, 𝑡′) = 𝐺+,*
+8 (𝒌, 𝑠, 𝑠), 𝑡, 𝑡′)
+ = −
+!
+ℏ 𝜃*(𝑡 − 𝑡′)
+ × 𝑒#!$(𝒌,;)?%#%%@/ℏ𝛿(𝑠 − 𝑠))
+.(13)
+The function 𝜃*(𝜉) is the normal Heavyside step
+function which is 1 for 𝜉 ≥ 0, and 0 otherwise. Here,
+the energy is composed of the transverse energy plus
+that induced in the z-direction by the electric field, and
+this energy may be written as
+
+
+𝐸𝒌,; = 𝐸(𝒌, 𝑠) =
+ℏ#A#
+/6∗ + 𝑒𝐹𝑠 .
+(14)
+Both forms for the energy will be used below. The
+temporal Fourier transform of (13) is easily taken to be
+
+
+𝐺*
+8(𝒌, 𝑠, 𝑠), 𝑡, 𝑡′) =
+B?;#;%@
+ℏC#$(𝒌,;)D!E .
+(15)
+where the limit 𝜂 → 0 is usually taken, just as in the
+equilibrium Green’s function case. Here, the
+convergence factor is confusing, since it implies the
+field was turned on in the infinite past. But, this is not
+the case. As may be seen in (13), the field is turned on
+at the time 𝑡′, so this convergence factor is useful, but
+not the correct interpretation. Care has to be taken over
+this point, but it will not affect the final results.
+
+The full transformation with the Airy functions
+can now be taken to yield the equation for the retarded
+Green’s function to be
+
+𝐺+Y𝒌, 𝑠, 𝑠), 𝜔 − 𝐸𝒌,;Z = 𝐺*
+8(𝒌, 𝑠, 𝜔)𝛿(𝑠 − 𝑠))
+ × ∫
+:&𝒒
+G0# ∑
+[𝑀H[
+/ H𝑁H +
+.DI
+/ I
+IJ±.
+𝛿(𝜅)
+× 𝐺+(𝒘, 𝒘′, 𝜅′)
+ (16)
+with the substitutions
+
+
+𝜅 = 𝜔 − 𝜈𝜔* − 𝐸𝒌,; + 𝐸𝒌%,;
+𝜅′ = 𝜔 − 𝜈𝜔* − 𝐸𝒌,;%
+𝒘 = 𝒌 + 𝑘𝒛𝒂5 + 𝜈𝒒
+𝒘′ = 𝒌′ + 𝑘𝒛
+)𝒂5 + 𝜈𝒒
+ .
+(17)
+
+The retarded self-energy is the product of the
+retarded Green's function and the retarded phonon
+Green's function. In most semiconductors, the
+scattering is weak, and the self-energy can be
+calculated to lowest order in time-dependent
+perturbation theory. This means that the Green's
+function propagator is approximated by the free
+propagator (in the presence of the field) (15). While
+this may suggest that physical effects of the scattering
+are being minimized, this isn't the real case. The
+retarded Green's function (16) is still being solved self-
+consistently, and the self-energy corrections will
+certainly be built into the result from Dyson's equation.
+Here, a non-degenerate electron gas will be considered
+
+
+5
+as discussed below. This is equivalent to neglecting
+any screening of the optical phonons. Moreover, the
+phonons are taken to be in equilibrium, so that the
+normal Bose-Einstein distribution, at the lattice
+temperature, can be used to describe the statistics of
+the phonons. The retarded phonon propagator is then
+
+
+𝐷*,+(𝒒) = −𝑖𝜋 ∑
+[𝑀H[
+/
+IJ±.
+ × H𝑁H +
+.DI
+/ I 𝛿(𝜅)
+.
+(18)
+The summation runs over emission and absorption of
+the optical phonon, with the phonon energy considered
+to be constant, 𝑀H is the electron-phonon matrix
+element, and 𝑁H is the Bose-Einstein distribution. This
+leads to the self-energy term as
+
+Σ+(𝒌, 𝒌′, 𝑧, 𝑧′, 𝜔) = ∫
+:&𝒒
+G0# ∑
+[𝑀H[
+/
+IJ±.
+
+ × H𝑁H +
+.DI
+/ I 𝛿(𝜅)𝐺+(𝒘, 𝒘′, 𝜅′)
+. (19)
+While this still remains relatively simple, it is now
+necessary to take the Airy transform of this function,
+and this provides adequate complication. The Airy
+transform (11) is applied to both longitudinal
+variables, and may be expressed as
+Σ+(𝒌, 𝑠, 𝑠′, 𝜔) = ∫
+:&𝒒
+G0#1' ∑
+[𝑀H[
+/ H𝑁H +
+.DI
+/ I
+IJ±.
+× ∫ 𝑑𝑠. ∫ 𝑑𝑠/ 𝐺+(𝒌 + 𝜈𝒒, 𝑠., 𝑠/, 𝜅)
+× ∫ 𝑑𝑠. ∫ 𝑑𝑠/ ∫ 𝑑𝑧 ∫ 𝑑𝑧′𝐴𝑖(𝑧 − 𝑠)
+ × 𝐴𝑖(𝑧 − 𝑠.)𝐴𝑖(𝑧′ − 𝑠′)𝐴𝑖(𝑧′ − 𝑠/)
+.(20)
+Here, the Airy functions have reduced arguments with
+the values
+
+
+𝐴𝑖(𝑦) = 𝐴𝑖 H
+78M#ℏC
+N
+I
+Θ = f3(𝑒ℏ𝐹)2
+2𝑚∗ g
+1/3
+= 𝑒𝐹𝐿
+ .
+(21)
+Because 𝐺*
+8 is diagonal in s, it may be expected that
+the resultant self-energy is also going to be diagonal in
+s once the transform has been taken. Then, the
+integrals can be performed using the known properties
+of the Airy functions and their integrals [49]. The
+resulting self-energy is then found to be [41]
+
+
+Σ+(𝒌, 𝑠, 𝜔) =
+(6∗)&/#QR)Q
+#√N
+√/ℏ#
+
+ × ∑
+H𝑁H +
+.DI
+/ I 𝐹(𝑠, 𝜅)
+IJ±.
+ , (22)
+with
+
+
+𝐼𝑚{𝐹(𝑠, 𝜅)} = 𝐴𝑖′(−𝑦)𝐵𝑖′(−𝑦).
+ −𝑦𝐴𝑖(−𝑦)𝐵𝑖(−𝑦)
+𝑅𝑒{𝐹(𝑠, 𝜅)} = 𝐴𝑖′/(−𝑦) − 𝑦𝐴𝑖/(−𝑦)
+,
+(23)
+where Bi is the Airy function of the second kind, and
+the primes denote derivatives with respect to the
+argument. These functions will be plotted below, once
+the material system has been described.
+
+4.2 The Correlation Functions
+
+The less-than Green's function will originate from the
+above retarded functions, and will satisfy its own set
+of differential equations. From this less-than function,
+a distribution function will evolve. This is the goal, to
+find the quantum non-equilibrium distribution
+function. This function will satisfy an integral
+equation that will be amenable to EMC. The
+differential equations to be solved are Fourier and Airy
+transformed to become
+
+
+jℏ𝜔 − 𝐸𝒌,;l𝐺,(𝒌, 𝑠, 𝑠′, 𝜔) =
+ = [Σ+(𝒌, 𝑠, 𝜔)𝐺,(𝒌, 𝑠′, 𝜔)
+ +Σ,(𝒌, 𝑠, 𝜔)𝐺((𝒌, 𝑠′, 𝜔)]
+ , (24)
+and
+
+
+jℏ𝜔 − 𝐸𝒌,;%l𝐺,(𝒌, 𝑠, 𝑠′, 𝜔) =.
+ = [𝐺+(𝒌, 𝑠, 𝜔)Σ,(𝒌, 𝑠′, 𝜔)
+ +𝐺,(𝒌, 𝑠, 𝜔)Σ((𝒌, 𝑠′, 𝜔)]
+ .
+(25)
+Normally, at this point, the sum and difference of these
+two equations are taken. Here, however, this leads to
+the same equation [42], which may be directly solved
+for the less-than function. Taking the sum of the two
+equations,
+and
+performing
+a
+little
+algebraic
+reorganization, one finds that the less-than function
+reduces simply to [42]
+
+ 𝐺,(𝒌, 𝑠, 𝑠′, 𝜔) = 𝐺+(𝒌, 𝑠, 𝜔)
+ × 𝐺((𝒌, 𝑠′, 𝜔)Σ,(𝒌, 𝑠, 𝑠′, 𝜔) .
+(26)
+It is easily shown that the difference of the two
+equations leads to the same result. One can easily also
+show that 𝐺+ − 𝐺( = 2𝐼𝑚{Σ+(𝒌, 𝑠, 𝜔)}𝐺+𝐺(. These
+constraints allow us to rewrite the less-than function in
+the generalized Kadanoff-Baym manner as
+
+
+𝐺,(𝒌, 𝑠, 𝜔) = 𝐴(𝒌, 𝑠, 𝜔)𝑓(𝑠, 𝜔) ,
+𝑓(𝑠, 𝜔) =
+T*(𝒔,,C)
+/V6{T+(;,C)} .
+
+(27)
+In this form, a quantum distribution function has been
+defined and is clearly not one of equilibrium.
+Normally, in the Kadanoff-Baym form (5), the
+distribution function is only a function of 𝜔, but the
+parameter s has been retained here to indicate the
+dependence of the various paths for the particles on
+this quantity. While the form of this distribution
+function appears to be simple enough, care must be
+taken because the details of the less-than self-energy
+will certainly make this equation difficult.
+
+Just as the retarded self-energy was developed,
+the less-than self-energy can be easily developed for
+the case of the non-polar intervalley phonons in Si.
+
+
+6
+The self-energy can be written in the Airy transformed
+variables as
+
+
+Σ,(𝒌, 𝑠, 𝜔) =
+QR)Q#
+9,/-1# ∑
+H𝑁H +
+ID.
+/ I
+IJ±.
+× ∫𝑑/𝒒% × ∫ 𝑑𝑠′𝐴𝑖/ H
+;#;%
+9,/&1I
+× 𝐺,(𝒌 − 𝜈𝒒%, 𝑠′, 𝜔 − 𝜈𝜔*)
+ .
+(28)
+As this latter equation depends upon the less-than
+Green's function, it is clear that it will lead to an
+integral equation for the distribution function, that
+appears in (27). The integration over 𝒒% removes all
+dependence on the transverse momentum as it leads to
+a conservation of this momentum (all uncertainty now
+exists in the longitudinal change in momentum and
+this is in the spatial Airy variables. As a result, the
+distribution function can be written as the integral
+equation [42Error! Bookmark not defined.]
+ 𝑓(𝑠, 𝜔) =
+.
+V6{T+(;,C)} ∑
+H𝑁H +
+ID.
+/ I
+IJ±.
+ × ∫ 𝑑𝑠′𝐾(𝑠, 𝑠′, 𝜔 − 𝜈𝜔*)(𝑠′, 𝜔 − 𝜈𝜔*)
+ (29)
+with
+
+
+𝐾(𝑠, 𝑠′, 𝜔) =
+√9QR)Q
+#6∗
+ℏ1#
+𝐴𝑖/ H
+;#;%
+9,/&1I p
+0
+/
+ + 𝑎𝑡𝑎𝑛ℎ t
+ℏC#$𝒌,0%#Z7[T+?,;%,C@\
+V6{T+(,;%,C)}
+ug
+. (30)
+While it may not seem evident, this integral equation
+for the distribution function may be readily solved
+with EMC techniques. In the classical approach, the
+ballistic drift of the carriers is for a period of time that
+is determined from the probability that a carrier has not
+been scattering over that period of time. This is
+represented by the probability being a negative
+exponential of the time relative to the inverse of the
+scattering rate. The same principle applies. However,
+instead of the time, it is the ballistic distance 𝑠 − 𝑠),
+relative to the distance L. From this distance, one can
+determine the ballistic time and other dynamic
+variables, as discussed below. Thus, one has the same
+iterated procedure as in classical EMC: drift for a
+period of time/space, then scattering according to the
+various processes in the less-than self-energy. The
+leading term involving the imaginary part of the
+retarded self-energy serves merely to renormalize the
+distribution.
+
+4.3 The Si Model
+
+To illustrate this new EMC approach to solving for
+NEGF, the case of Si will be considered. While the
+situation for electrons in the conduction band is
+complicated by the multi-valley nature of this band,
+the scattering processes themselves are very local and
+do not require some to the more complicated higher-
+order corrections to the simpler Fermi golden rule,
+although the latter is of course different for the
+quantum case. The conduction band of Si has six
+equivalent ellipsoids located along the D lines [these
+are the set of (100) axes] about 85% of the way to the
+zone edge at X. Since scattering between these
+equivalent valleys is possible, the electric field will be
+taken along the (111) direction so that the same angle
+is made with each of the six valleys. Scattering within
+each ellipsoid is limited to acoustic phonons, as intra-
+valley optical processes are forbidden. Acoustic mode
+scattering, by way of the deformation potential, is
+characterized by two constants Xu and Xd, which are
+thought to have values of 9 eV and -6 eV, respectively
+[50]. The effective deformation potential is then the
+sum of these, or about 3 eV. Non-polar “optical”
+phonon scattering occurs for scattering between the
+equivalent ellipsoids. There are two possible types of
+phonons that are involved in this process. One is the
+g-phonon that couples the two valleys along opposite
+ends of the same (100) axis. This is an umklapp
+process so that, after reduction by a reciprocal lattice
+vector, the scattering has a net phonon wave vector of
+0.3p/a. The symmetry allows only the LO mode to
+contribute to this scattering. At the same time, f-
+phonons couple the (100) valley to the (010) and (001)
+valleys, and so on. The latter wave vector lies in the
+(110) direction has a magnitude of 21/2(0.85)p/a =
+1.2p/a, which lies in the square face of the Brillouin
+zone along the extension of the (110) line into the
+
+Figure 1. The scattering rates that enter the retarded
+self-energy are shown for Si at room temperature. These
+include the zero- and first-order intervalley optical
+phonons and the acoustic phonons. The rates are for a
+field of 10 kV/cm.
+
+IV
+0,em
+IV
+IV
+1,em
+1,em
+13
+IV
+0,abs
+Acoustic
+1x10
+1x1010
+-0.2
+0
+0.2
+0.4
+0.6
+0.8
+1
+Energy E(k,s) (eV)
+7
+second Brillouin zone [51]. The phonons here are near
+the X-point phonons in value but have a different
+symmetry.
+
+The energies of the LO g-phonon and the LA and
+TO f-phonons have nearly the same value, while the
+low-energy inter-valley phonons are forbidden. Long
+[52], however, determined from an analysis of the
+experimental mobility versus temperature that a weak
+low-energy inter-valley phonon is required to fit the
+experimental data. He treated the high-energy
+phonons by a single equivalent inter-valley phonon of
+64.3 meV, but introduced a low-energy inter-valley
+phonon with an energy of 16.4 meV. The presence of
+the low-energy phonons is also confirmed by studies
+of magneto-phonon resonance (where the phonon
+frequency is equal to a multiple of the cyclotron
+frequency) in Si inversion layers, which indicates that
+scattering by the low energy phonons is a weak
+contributor to the transport [53]. The low-energy
+phonon is certainly forbidden although Long treats it
+with a very weak coupling constant. It is more likely
+that this forbidden low-energy inter-valley phonon
+must be treated by a first-order interaction [54,55].
+This fits the data with a coupling constant of X0 = 5.6
+eV, while the allowed, higher-energy transition is
+treated with a coupling constant D = 9 ´ 108 eV/cm.
+This two-phonon model is adopted for the present
+treatment of Si.
+
+Non-parabolicity of the conduction bands away
+from the D minima arises more from repulsion from
+the upper (second) conduction along these directions
+than from a valence band interaction. Non-parabolicity
+is included in the simulations by the normal hyperbolic
+bands with an effective "gap" of 2.1 eV [51]. This non-
+parabolicity will be important in computing velocities
+from the energies and momentum determined during
+the Monte Carlo process described in the next section.
+
+The various scattering rates that go into the
+retarded self-energy, under the approximations
+described above, and for a field of 10 kV/cm, are
+shown in figure 1. The real and imaginary parts of the
+retarded self-energy are shown in figure 2 for three
+values of the electric field (10, 40, 70 kV/cm). The
+oscillations in the real part lead to the quasi-
+quantization of the density of states into two-
+dimensional sub-bands in the high electric field. The
+field tends to create a triangular potential well, which
+is expected from the quantum mechanics. When the
+real part is positive, the energy is lowered, while when
+it is negative, the energy is raised. This tends to form
+a set of sub-band energy levels [56]. This tendency
+toward formation of sub-bands is also evident in the
+imaginary part of the self-energy. It is clear, however,
+that the general shape of the imaginary part is only
+weakly dependent on the value of the electric field.
+The zero of energy in these plots is taken as the
+conduction band minimum for the one-electron bands.
+Broadening arises from both the scattering and the
+field, and this leads to values for negative energy.
+
+The retarded self-energy leads to the spectral
+density which describes how the various energy levels
+are broadened by the scattering and quantization
+
+
+(a)
+(b)
+Figure 2. (a) The imaginary part of the retarded self-energy as a function of the energy itself. (b) The real part of the retarded
+self-energy as a function of the energy itself. These are determined using the scattering mechanisms described in the text for Si
+at 300 K and for three values of the electric field (10, 40, and 70 kV/cm).
+
+0.06
+10 kV/cm
+0.05
+40 kVlcm
+70 kV/cm
+0.04
+Im(Zr(E)) (eV)
+Si bulk
+300 K
+FIl(111)
+0.03
+0.02
+0.01
+0
+-0.1 -0.05
+0
+0.05 0.1 0.15 0.2 0.25
+0.3
+Energy E(k,s) (eV)4x10-3
+10 kV/cm
+3x10-3
+ 40 kV/cm
+70 kV/cm
+2x10-3
+Re{'(E)) (eV)
+1x10-3
+0
+-1x10-3
+-2x10-3
+-3x10-3
+-0.1
+-0.05
+0
+0.05
+0.1
+0.15
+0.2
+Energy E(k,s) (eV)
+8
+processes. This spectral density is shown in figure 3
+for the three field values used in illustrating the self-
+energy. Here, it appears that, as with the imaginary
+part of the self-energy, the field has only a small effect
+on the spectral density. Thus, it may be concluded that
+the retarded (and advanced) functions are not
+particularly dependent upon the actual value of the
+electric field applied to the device, other than the
+changes due to the tendency to form two-dimensional
+sub-bands. These functions are more related to the
+intrinsic properties of the material used in the
+simulations, and not so much upon the actual value of
+the external field.
+
+5. The EMC Results
+
+Following from the above discussion, and the results
+of (29) and (30), the sequence of the Monte Carlo steps
+follow a similar path to that for the Boltzmann
+equation. Here, however, the exponential in time is
+replaced by the Airy function in space. That is, drift is
+defined by an effective path length first, and then
+scattering occurs. The difference between this
+approach and the Monte Carlo for the Boltzmann
+equation is that drift length replaces drift time,
+although these are intimately related, as will be
+described below. It may be noted that the imaginary
+part of the retarded self-energy sits outside the integral
+and is a final adjustment on the distribution function.
+To begin, the spatial motion of the particle is
+determined with a random number r according to
+
+
+𝑟 = ∫
+𝐴𝑖/(𝑢′)𝑑𝑢′
+]
+#]"
+∫
+𝐴𝑖/(𝑢′)𝑑𝑢′
+^
+#]"
+v
+ ,
+(31)
+so that a value for the drift length is determined from
+u as
+
+
+∆𝑠 = 3./9𝐿𝑢 .
+(32)
+There is a question about the use of the Airy function
+because it has values over a semi-infinite range of
+values. However, one may note that the scattering
+function in (30) seems to be over a single effective
+sub-band (the values of the bracketed term), so that the
+quantity −𝑢* is taken to be the first zero of the Airy
+function, so that there is a single maximum in the
+quantity integrated in (31). This gives a range of values
+for ∆𝑠 for the ensemble of particles. This may be
+averaged over the ensemble and over the number of
+iterations to determine an "average" value of ∆𝑠, and
+this is plotted as a function of the electric field in figure
+4. The error is extremely small, so that error bars do
+not show up; the average is computed over the 105
+particles and for 200 iterations of the EMC algorithm.
+While the velocity appears to converge rapidly, as
+discussed below, the number of iterations is used to
+assure that the distribution function is converged.
+
+In the EMC procedure, each particle has its own
+value for ∆𝑠. This leads to an energy gain in the field
+of 𝑒𝐹∆𝑠, which is directed along the electric field
+(taken to be the z axis for convenience). As a result,
+the momentum wave vector is changed by an amount
+∆𝑘 = (𝑒𝐹/ℏ)∆𝑡 (but see a comment below). It is
+therefore necessary to find this value for ∆𝑡. The drift
+time and the drift distance are trivially related in
+classical mechanics where the bands are parabolic.
+With the non-parabolic bands that exist in Si, this is no
+longer the case, and some work is necessary to connect
+these two quantities. The problem arises from the
+energy dependence of the effective mass in the non-
+
+Figure 3. The spectral density obtained from the retarded
+(and advanced) self-energies for Si at 300 K.
+
+Figure 4. The average drift length Ds as a function of the
+electric field. The average is computed over 105 particles
+and 200 iterations of the Monte Carlo algorithm.
+
+1.2x103
+ 10 kV/cm
+1x103
+40 kV/cm
+Spectral Density (rel. units)
+70 kV/cm
+8x102
+6x10
+4x102
+2x10
+0
+-0.02
+-0.01
+0
+0.01
+0.02
+hw - E(k,s)Si bulk
+300 K
+6
+Average △S (nm)
+5
+4
+3
+2
+20
+40
+60
+80
+100
+120
+0
+Electric Field (kV/cm)
+9
+parabolic bands, which breaks the simple connection
+between momentum wave number and velocity. In the
+hyperbolic bands used, the mass increases with energy
+according to [51]
+
+
+𝑚∗ = 𝑚_(1 + 𝐸/𝐸`) ,
+(33)
+where 𝑚_ is the mass at the conduction band minimum
+and 𝐸` is the effective energy gap mentioned in the
+previous section. This is determined for the one-
+electron bands, as the broadening is added with the
+quantization in the Green's functions. The change in
+velocity can be written as
+
+
+∆𝑣(𝑡) =
+78%
+61 H1 +
+$(*)
+$2 +
+78a%
+$2 I
+#.
+,
+(34)
+which leads to a difficult integral equation. Here, 𝐸(0)
+is the energy at the start of the drift. We may take the
+distance as 𝑧 = 𝑣𝑡. The latter suggests that the velocity
+be written as 𝑑𝑧/𝑑𝑡, so that (34) may be rewritten as
+
+
+:5
+:% H1 +
+$(*)
+$2 +
+78a%
+$2 I =
+78
+61 𝑡 ,
+(35)
+with z running from 0 to ∆𝑠 and t running from 0 to
+∆𝑡. This integral is now trivially evaluated to lead to
+the value for the increment in time as
+
+
+∆𝑡 = |
+/61
+78 |∆𝑠| H1 +
+$(*)
+$2 +
+78∆;
+/$2 I
+= |
+/61
+78 |∆𝑠| H1 +
+$
+$2 −
+78∆;
+/$2 I
+ ,
+(34)
+where E is now the energy at the end of the drift. Once
+this time increment is determined, the momentum
+wave vector can be updated, and the velocity
+determined at the end of the drift period. It also allows
+one to determine the average drift time for the
+ensemble of carriers, in order to establish an average
+time scale for the process.
+
+It has to be noted that ∆𝑠 can be either positive or
+negative. A negative value merely means that the
+particle is actually being accelerated or decelerated in
+the direction opposite to the field (acceleration will
+occur for a negative momentum wave vector). The
+value of ∆𝑡 should not depend upon this sign, but the
+change in momentum certainly does depend upon the
+sign. Thus, the momentum change given above should
+actually account for this sign as
+
+
+∆𝑘 =
+78∆%
+ℏ 𝑠𝑖𝑔𝑛(∆𝑠) .
+(35)
+
+One of the first steps to solving the Green's
+functions of section 4.2 is determining the less-than
+self-energy, especially the imaginary part that is
+necessary in (30). This is done within the EMC
+procedure and this scattering function is shown in
+figure 5 for the same three values of electric field used
+earlier. In contrast to the retarded self-energy, in this
+case the less-than self-energy clearly is a function of
+the applied electric field, with more scattering at
+higher electric fields. This presumably is a result of
+more carriers at higher energies, where the scattering
+rates are higher, as shown in figure 1. This is also
+reflected in the shorter drift lengths apparent at higher
+values of the field in figure 4, although the higher field
+itself affects these lengths through the parameter L.
+
+The drift velocity is the average velocity of the
+ensemble of carriers, and is always computed by such
+an ensemble average. This ensemble average is then
+averaged over the 200 iterations of the simulation. The
+resulting average drift velocity is plotted as. a function
+
+Figure 5. The imaginary part of the less-than self-energy is
+shown for three values of the electric field.
+
+Figure 6. The drift velocity as a function of the electric field
+in Si at 300 K. This is determined from the ensemble
+averages and an average over the iterations.
+
+0.3
+10 kV/cm
+40 kV/cm
+0.25
+70 kV/cm
+Im(Z(E)) (eV)
+0.2
+0.15
+0.1
+0.05
+0
+-0.1 -0.05
+0
+0.05
+0.1 0.15 0.2 0.25
+0.3
+Energy E(k,s) (eV)1
+Drift Velocity (10′ cm/s)
+Si bulk
+300 K
+EI (111)
+0.1
+20
+0
+40
+60
+80
+100
+Electric Field (kV/cm)
+10
+of the electric field in figure 6. While they are not
+evident, error bars are actually shown in the figure, for
+the average over the iterations. These errors turn out to
+be less than 0.8% of the actual velocity, and so are
+buried under the symbol used in the plot. It might seem
+natural to determine the average velocity from the
+ration ∆𝑠/∆𝑡. But, this would be a mistake as the latter
+quantity is the average over the drift length, and not
+the velocity at the end of the drift which is the quantity
+that leads to the drift velocity by virtue of Hamilton's
+equations of motion. The average of the velocity over
+the iterations is used to reduce the noise, but in many
+cases that can be misleading when the actual velocity
+is strongly time dependent. This is not the case here.
+The variation in the velocity during the simulation can
+be illustrated in another manner by plotting the
+velocity versus the average time that is determined
+from the average ∆𝑡 values, and this is done in figure
+7. Each value of the field has its own apparent time
+scale, which is the average steps per iteration. As may
+be seen from figure 4 and (34), these time scales differ
+for each value of the electric field. In essence, it is
+perhaps better to consider these "time" scales as
+merely a parameter that measures the evolution of the
+ensemble over time. This differs from classical Monte
+Carlo, where the time is determined more or less
+accurately in the simulation. In this quantum version,
+it is the distance that is determined more or less
+accurately (we return to this point in the next section).
+However, after an initial value (after the first iteration)
+that is slightly higher than the subsequent values, the
+velocity is relatively constant. The noise may be
+estimated by the fluctuations in figure 7. The slightly
+higher value after the first iteration may be indicative
+of some velocity overshoot [57], but this cannot be
+determined properly, especially as the normal
+overshoot in Si is relatively small.
+
+While the common value expected for velocity
+saturation is around 107 cm/s, a higher value is found
+here (figure 6). It is felt that the lower value found in
+semi-classical Monte Carlo arises from the rapid
+energy rise in scattering by the first-order coupled low
+energy intervalley phonon. However, in the quantum
+case, it is found that this first-order scattering is
+especially affected by the intra-collisional field effect
+that arises in quantum transport [58]. Hence, it is not
+surprising that a larger "saturated" velocity is found in
+quantum transport.
+
+In figure 8, the carrier population is plotted as a
+function of carrier energy for three different values of
+the electric field (same values as used earlier). As the
+field increases, carriers at low energy are lost to higher
+energy states, so that the distribution streams to higher
+energies. The various bumps and plateaus that tend to
+form arise from the tendency to have two-dimensional
+quantization and sub-bands in the high fields, as well
+as the role of onset of various phonons. The presence
+of particles at negative energies is the result of the
+broadening of the energy states, as represented by the
+spectral density of figure 3. While the plot uses relative
+units (the actual values do not relate to the number of
+particles), the differences between the three curves are
+accurate, and reflect the field induced differences.
+
+Figure 7. The variation in the drift velocity for three values
+of the electric field, as a function of the simulated time.
+This time scale differs for each value of the field, as
+discussed in the text.
+
+Figure 8. The relative carrier population as a function of
+energy. The bumps and plateaus are reflective of the
+tendency to two-dimensional quantization and some
+phonon effects.
+
+10 kVlcm
+1.4
+40 kVlcm
+70 kV/cm
+Drift Velocity (10′ cm/s)
+1.2
+1
+0.8
+0.6
+10-12
+10-11
+10-14
+10-13
+10-10
+Average Drift Time (s)1x10
+10 kV/cm
+40 kV/cm
+70 kv/cm
+1x10
+Population (arb. units)
+1x10
+1x102
+-0.05
+0
+0.05
+0.1
+0.15
+0.2
+Energy (eV)
+11
+6. Discussion
+
+In this paper, the use of the ensemble Monte Carlo
+process has been shown to be effective in evaluating
+non-equilibrium Green's functions in a fast and
+effective approach. While this procedure is limited, at
+the present time, to materials in which the scattering is
+very local, such as with nonpolar optical and
+acoustical phonons, this application is of importance
+in most semiconductor devices used for integrated
+circuits. Typically, the computation for a single value
+of the electric field takes only on the order of 1.6-2.1
+seconds on a modern laptop computer (here a
+MacBook Pro with the M1 Pro chip), with parallel
+processing available.
+
+The results obtained for the drfit velocity in figure
+6 are intriguing, in the sense that above ~40 kV/cm,
+there does not appear to be any hard saturation of the
+velocity, which is commonly seen in semi-classical
+Monte Carlo approaches [59,60], with a value of
+1 × 10c cm/s at room temperature, although some
+versions give values well above this [61]. However,
+this can be misleading, as there are a wide variety of
+such approaches. However, measurements are not
+clear on this point. Most measurements do not go
+beyond 40-50 kV/cm. For example, the oldest
+measurements using time-of-flight techniques extend
+to approximately 50 kV/cm [62], 13 kV/cm [63], or 20
+kV/cm [64], so that the appearance of a hard saturation
+is not seen in the data. It has also been found that
+quantum simulations generally give higher values of
+velocity than the semi-classical ones [65,66]. The
+value of 1.3 × 10c cm/s found here is close to that
+found by Reggiani et al. [61].
+
+If one accepts that the quantum drift velocity is
+larger than that found semi-classically, there must be
+a reason. First, it has been known for some time that
+the collision broadening inherent in the spectral
+density contributes to this [61]. In addition, normally
+the Monte Carlo determines the ballistic drift length
+from a random time step derived from the total
+scattering rate. In the quantum method presented here,
+the ballistic drift length is determined solely by the
+distance step that depends only upon the electric field.
+Moreover, the important scattering arises from the
+imaginary part of the less-than self-energy (figure 5)
+and is this directly a function of the electric field. Thus,
+there is a direct inter-twining of the field and the
+scattering that contributes to the effect referred to as
+the intra-collisional field effect [2,67], a point
+emphasized earlier [61]. But, there is a further
+contribution, apparent in (30). This is the fact that the
+final states for scattering are an almost two-
+dimensional density of states, represented by the
+inverse hyperbolic tangent function. Quite generally,
+the density of states in a two-dimensional system is
+less than that in a three-dimensional system. This
+lower density of states gives less scattering and
+therefore a higher velocity. So there are several
+reasons for the drift velocity to be higher in quantum
+simulations than in classical ones.
+
+Another important consideration for the future is
+that the present approach has been applied for
+homogeneous material, while devices tend to be quite
+inhomogeneous.
+Extending
+this
+approach
+to
+inhomogeneous devices should be only a minor
+problem, as the Monte Carlo technique is quite usual
+in such device simulation. Most of the computational
+time is taken with Poisson's equation, and from the
+results of this equation, the spatially dependent
+quantum distribution can be developed rather quickly.
+
+In devices, there is usually a grid upon which the
+density is projected in order to solve Poisson's
+equation. Moreover, when Monte Carlo techniques are
+used for the transport in cases of real-space treatment
+of the impurities [68], or treating the electron-electron
+interaction by molecular dynamics [69], there are
+usually two time scales, one attached to each particle
+and the laboratory time scale for updating the total
+potential. Techniques have been adopted for
+synchronizing these two times and distributing the
+acceleration between multiple bins of the laboratory
+time scale, since the potentials have to be updated on
+a much faster time scale than the scattering events
+[51]. In principle, dividing the ∆𝑠 or ∆𝑡 values can
+easily be done by this same approach for either
+inhomogeneous
+material
+and
+devices
+or
+for
+incorporation of carrier-carrier interactions.
+
+Extending this approach to the polar optical
+phonons involves the problem of the long-range of
+Coulomb interactions. The polar interaction is a
+dipolar one, but still is quite inhomogeneous in the
+scattering dynamics just as is impurity scattering. The
+latter has been made more amenable through treating
+the impurities in real space and developing the total
+potential from this real space approach [68]. This has
+not been done for the polar modes as yet. For solutions
+with the Green's functions, this should require direct
+solution of the Bethe-Salpeter equation for the very
+inhomogeneous scattering by the polar phonons.
+However, this latter equation is itself an iterative
+procedure for high accuracy, and it is not beyond
+expectations that a second Monte Carlo process may
+be used for its evaluation, beyond the Monte Carlo
+procedure for determining the resultant Green's
+function and distribution function. It is also feasible to
+consider that the polar mode interaction may be
+expressed in a real space approach, such as used for
+the impurities. While these suggestions remain to be
+speculation, it is certainly worth further research in
+this area to extend the present approach to devices and
+polar materials.
+
+
+12
+
+
+1.
+Budd H F 1967 Phys. Rev. 158 798
+2.
+Thornber K K and Feynman R F 1970 Phys. Rev. B 1
+4099
+3.
+Kurosawa T 1966 J. Phys. Soc. Jpn. Suppl. 21 424
+4.
+Jacoboni C and Reggiani L 1983 Rev. Mod. Phys. 65
+645
+5.
+Jacoboni C and Lugli P 1989 The Monte Carlo Method
+for Semiconductor Device Simulation (Vienna:
+Springer-Verlag)
+6.
+Hess K 1991 Monte Carlo Device Simulation: Full Band
+and Beyond (Boston: Kluwer Academic)
+7.
+Kosina H, Nedjalkov M and Selberherr S 2000 IEEE
+Trans. Electron Dev. 47 1898
+8.
+Shichijo H and Hess K 1981 Phys. Rev. B 23 4197
+9.
+Lugli P and Ferry D K 1985 IEEE Trans. Electron Dev.
+32 2431
+10.
+Joshi R P and Ferry D K 1991 Phys. Rev. B 43 9734
+11.
+Jacoboni C 1976 Proc. 13th Int. Conf. Phys.
+Semiconductors (Rome: Tipo. Marves) 1195
+12. De Broglie L 1924 Phil. Mag., Ser. 6, 47 446
+13.
+Kennard E H 1928 Phys. Rev. 31 876
+14.
+Feynman R P 1948 Rev. Mod. Phys. 20 367
+15.
+Wigner E 1932 Phys. Rev. 40 749
+16.
+E. J. Heller, J. Chem. Phys. 65, 1544 (1976).
+17.
+T. Dittrich, C. Viviescas, and L. Sandoval, Phys. Rev.
+Lett. 96, 070403 (2006).
+18.
+Vitanov P et al. 1994 Advances in Parallel
+Algorithms, Ed. by B. I. Sendov and I. Dimov
+(Amsterdam: IOS Press) pp. 117-128
+19.
+Shifren L and Ferry D K 2002 J. Comp. Electron. 1 55
+20.
+Nedjalkov M et al. 2002 Proc. 7th Intern. Conf. on
+Simulation of Semiconductor Processes (New York:
+SISPAD) pp. 187-190
+21.
+Ferry D K et al. 2021 Semicond. Sci. Technol. 36
+025024
+22.
+Fischetti M V 1984 Phys. Rev. Lett. 53 1755
+23. Porod W and Ferry D K 1985 Phys. Rev. Lett. 54 1189
+24.
+Jauho A P 1985 Phys. Rev. B 32 2248
+25.
+Reggiani L, Lugli P and Jauho A P 1987 Phys. Rev. B
+36 6602
+26.
+Benseney A et al. 1988 Euro. Phys. J. 68 286
+27.
+Baym G and Kadanoff L P 1961 Phys. Rev. 124 287
+28.
+Kadanoff L P and Baym G 1962 Quantum Statistical
+Mechanics (New York: W. A. Benjamin)
+29.
+Schwinger J 1951 Proc. Nat. Acad. Sci. 37 452
+30.
+Schwinger J 1960 Proc. Nat. Acad. Sci. 46 883
+31.
+Dyson F J 1949 Phys. Rev. 75 1736
+32.
+Prigogine I 2002 Phil. Proc. Royal Soc. A Math. Engr.
+Phys. Sci. 360 299
+33.
+Lipavský P, Ŝpiĉka V and Velický B 1986 Phys. Rev. B
+34 6933
+34.
+Joshi R P and Ferry D K 1991 Phys. Rev. B 43 9734
+35.
+Barker J R 2003 J. Comp. Electron. 2 153
+References
+36.
+Weinbub J, Ballacchia M and Nedjalkov M 2018
+Phys. Stat. Sol. RRL 12 18000111
+37.
+Gilbert M J and Ferry D K 2005 IEEE Trans.
+Nanotechnol. 4 355
+38.
+Vasileska D et al. 1995 J. Vac. Sci. Technol. B 13 1841
+39.
+Ferry D K 2017 An Introduction to Quantum
+Transport in Semiconductors (Singapore: Jenny
+Stanford Publ.)
+40.
+Ferry D K 2021 Quantum Mechanics, 3rd Ed. (Boca
+Raton: CRC Press) 51
+41.
+Bertoncini R, Kriman A M and Ferry D K 1989 Phys.
+Rev. B 40 3371
+42.
+Bertoncini R, Kriman A M and Ferry D K 1990 Phys.
+Rev. B 41 1390
+43.
+Bertoncini R, Kriman A M and Ferry D K 1990 J.
+Phys.: Cond. Matter 2 5991
+44.
+Dyson F 1951 Phys. Rev. 82 428
+45.
+Dyson F 1951 Phys. Rev. 83 608
+46.
+Castro E R and Roditi I 2019 J. Phys. A 52 335401
+47.
+Bernevig B A and Regnault N 2009 Phys. Rev. Lett.
+103 206801
+48.
+Jauho A P and Wilkins J W 1984 Phys. Rev. B 29 1919
+49.
+Aspnes D E 1966 Phys. Rev. 147 554
+50.
+Ridley B K 1982 Quantum Processes in
+Semiconductors (Oxford: The Clarendon Press)
+51. Ferry D K 2020 Semiconductors: Bonds and Bands
+2nd Ed. (Bristol: IOP Publishing)
+52.
+Long D 1960 Phys. Rev. 120 2024
+53.
+Eaves L et al. 1968 J. Phys. C 8 1975
+54.
+Ferry D K 1976 Phys. Rev. B14 1605
+55.
+Siegel W, Heinrich A, and Ziegler E 1976 Phys. Stat.
+Sol. (a) 35 269
+56.
+Reggiani L et al. 1991 Phys. Stat. Sol. (b) 168 K69
+57.
+Ruch J G 1972 IEEE Trans. Electron Dev. 19 652
+58.
+Ferry D K, Barker J R and Grubin H L 1981 IEEE Trans.
+Electron Dev. 28 905
+59.
+Osada Y 2012 IEEJ Trans. Electron. Info. Sys. 132
+1880
+60.
+Gada M L et al. 2012 Proc. NSTI Nanotechnol. Conf.
+712
+61.
+Reggiani L, Lugli P and Jauho A-P 1988 J. Appl. Phys.
+64 3072
+62. Canali C et al. 1973 Phys. Rev. B 12 2265
+63.
+Tove P A et al. 1970 IEEE Trans. Electron Dev. 17 407
+64.
+Haas G A, Pankey Jr. T and Harris F H 1973 J. Appl.
+Phys. 44 2433
+65.
+Bertoncini R and Jauho A-P 1998 Phys. Rev. Lett. 68
+2826
+66.
+Wacker A et al. 1999 Physica B 272 157
+67.
+Barker J R 1973 J. Phys. C 6 2663
+68.
+Vasileska D, Gross W J and Ferry D K 2000 Superlatt.
+Microstruc. 27 147
+69.
+Lugli P and Ferry D K 1986 Phys. Rev. Lett. 56 1295
+
diff --git a/LdFAT4oBgHgl3EQfwh69/content/tmp_files/load_file.txt b/LdFAT4oBgHgl3EQfwh69/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5714b411f77770b5949465a862c08d9a6aaa2018
--- /dev/null
+++ b/LdFAT4oBgHgl3EQfwh69/content/tmp_files/load_file.txt
@@ -0,0 +1,758 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf,len=757
+page_content="1 Using Ensemble Monte Carlo Methods to Evaluate Non-Equilibrium Green's Functions David K." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry School of Electrical, Computer, and Energy Engineering, Arizona State University, Tempe, AZ 25287-6206;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' ferry@asu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='edu Abstract The use of ensemble Monte Carlo (EMC) methods for the simulation of transport in semiconductor devices has become extensive over the past few decades.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This method allows for simulation utilizing particles while addressing the full physics within the device, leaving the computational difficulties to the computer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' More recently, the study of quantum mechanical effects within the devices, effects which also strongly affect the carrier transport itself, have become important.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" While particles have continued to be useful in quantum simulations using Wigner functions, interest in analytical solutions based upon the non-equilibrium Green's functions (NEGF) have become of greater interest in device simulation." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While NEGF has been adopted by many commercial semiconductor, there remains considerable computational difficulty in this approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, a particle approach to NEGF is discussed, and preliminary results presented illustrating the computational efficiency that remains with the use of particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This approach adopts the natural basis functions for use in a high electric field and the preliminary results are obtained for quantum transport in Si at 300 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This approach appears to offer significant advantages for the use of NEGF.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" Keywords: quantum transport, Non-equilibrium Green's functions, ensemble Monte Carlo, Airy transform 1." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Introduction In the classical world, transport in applied electric and magnetic fields is usually computed with the Boltzmann equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In low fields, one has a nearly equilibrium situation in which the distribution function is a Maxwellian or a Fermi-Dirac at a temperature that likely increases above that of the lattice.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In most devices, the distribution is well out of equilibrium, which means that it is unknown.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Finding this distribution is typically the single most difficult problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' At least classically, an alternative approach is to use the computer to completely solve the transport problem with a stochastic methodology, and several methods have arisen to do this, two of which are an integral iteration technique [1,2] and the Monte Carlo method [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Today, the most widely used approach is the latter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The ensemble Monte Carlo (EMC) technique uses an ensemble of particles whose propagation is determined in parallel, so that ensemble averages can be computed as a function of time for the various observables of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It has been the subject of many reviews [4,5,6,7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The methodology of the EMC approach is to spend time and effort on characterizing the correct physics, perhaps even to the use of atomistic full-band energy behavior [8], and the scattering properties, and let the computational burden be handled by a modern parallel processing computer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Hence, no attempt is made to arrive at an ab initio analytical form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These approaches have become extremely sophisticated, and many factors such as degeneracy [9], discrete impurities in real space [10], and many-body effects [11] can be incorporated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' From the beginning, there has been a rich history for the use of particles in quantum mechanics, as the suggestion of particles and waves dates to de Broglie [12], and the use of particles has been discussed over the years within the quantum world from Kennard [13] to Feynman [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The immediate problem with adapting the EMC approach for quantum transport lies with the use of the particle paths in phase space in the classical approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Generally, the uncertainty principle restricts being able to simultaneously define both a position and momentum for a particle path.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nevertheless, this strict interpretation has been finessed by a variety of approaches, many of which clearly endorse some form of a real path for a particle that exists with the wave.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Not the least of these approaches is due to Feynman himself.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' One more successful approach is the phase-space representation 2 of the Wigner function [15], as this suggests a comparison between quantum dynamics and the corresponding classical motion [16,17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, the Wigner function can have a non-unitary evolution, and off-diagonal terms in the density matrix (from which the Wigner function is found) lead to non-classical propagation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Indeed, quantum coherence is reflected in oscillatory behavior of the Wigner function and non-positive definite regions in phase space, and this leads to complications in the use of EMC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nevertheless, particle methods and the EMC have been adapted to treating quantum transport via the Wigner function [18,19,20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Quite generally, the operation and performance of semiconductor devices has evolved to incorporate many quantum effects [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Any quantum mechanical simulation of such a device has to meet all the requirements of a self-contained and consistent mathematical theory, just as in the classical case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' But it generally reflects a much deeper physical behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Certainly, a variety of quantum approaches to the simulation of transport have arisen, including the Wigner function mentioned above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Indeed, there have been many attempts to try to put quantum mechanics into the normal kinetic equations stemming from the Boltzmann approach, although most limit the method to approximating the spectral density that replaces the energy-conserving delta function for scattering [22,23,24,25].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Such an approach has also been taken with Bohmian trajectories and a quantum potential [26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, today one of the most popular approaches to quantum transport and quantum distributions is thought to be non-equilibrium Green’s functions (NEGF).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As in the classical case where the system is quite the same, the NEGF include new functions that must be found to describe both the non- equilibrium distribution function, but also how quantum correlations are built into the device [27,28].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These Green’s functions also determine the transport of an "excitation" over a certain distance to where it is annihilated (thus these functions contain two times and two spatial variables).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As a result, these functions bring considerable computational difficulty to any transport problem, with consequent long computation times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nevertheless, they are being used extensively in modern device simulation and design.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is known in quantum mechanics that any convenient coordinate system, or set of basis functions, may be used for any given problem, since the Schrödinger equation is a linear equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In this paper, the use of a particular set of basis set is adopted that is extremely useful in a high electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These constitute the Airy transform, after which one can obtain an integral equation for the important "less- than" Green\'s function that is amenable to solution via EMC techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While the approach is currently restricted to uniform fields and materials such as Si, it can be extended easily to the inhomogeneous world of semiconductor devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the following section, a brief introduction to NEGF, and the limitations of these functions, will be provided.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Then, in Secs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 3 and 4, the Airy transform approach will be described and the various NEGF developed with this transform and the case of Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The restrictions to Si will also be discussed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Section 5 will describe the EMC simulations and the results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Finally, in Sec.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 6, these results will be discussed and prospects for extending the approach to inhomogeneous devices and to other materials will appear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' NEGF In general, it is a well-known relationship in quantum mechanics that the wave function can be connected to initial conditions through the existence of a propagator as 𝜓(𝒙, 𝑡) = ∫ 𝑑𝒙′𝐾(𝒙, 𝒙′;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑡, 0)𝜓(𝒙′, 0) , (1) where one can write the propagator as 𝐾(𝒙, 𝒙′;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑡, 0) = ∑ 𝜑!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' "(𝒙′)𝜑!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (𝒙) !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑒#$!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='%/ℏ , (2) and the 𝜑!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (𝒙) are basis functions in a spatial expansion of the total wave function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It was Schwinger who formally connected this propagator to Green’s functions, which were already well-known in the mathematics of differential equations [29].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Initially, these were independent of time, as he was concerned with the equilibrium state, and he assumed time reversibility, hence sticking to an equilibrium system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This equilibrium was maintained, even in the presence of perturbations, by the equality of 𝑡 → ±∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These perturbations were handled formally by the existence of a unitary operator that involved the perturbing Hamiltonian exponentially [30], although this operator was developed much earlier [31].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" The important Green's functions in this equilibrium theory were the advanced and retarded functions given as 𝐺((𝑟, 𝑟);" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑡, 𝑡′) = 𝑖Θ(𝑡* − 𝑡) × 〈{𝜓"(𝑟), 𝑡′), 𝜓(𝑟, 𝑡)}〉 𝐺+(𝑟, 𝑟);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑡, 𝑡′) = −𝑖Θ(𝑡 − 𝑡′) × 〈{𝜓(𝑟, 𝑡), 𝜓"(𝑟), 𝑡′)}〉 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (3) Here, the curly brackets denote the normal anti- commutator relationships for fermions, while the angular brackets define normalization conditions, obtained with the equilibrium distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is a well-recognized fact that the application of an electric field leads to current and a phase-transition that breaks the time-reversal symmetry of these functions [32].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This then leads to the non-equilibrium state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" Normalization cannot be done with the 3 equilibrium distribution and one needs two new Green's functions, the correlation functions, in order to gain the non-equilibrium distribution." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These correlation functions are given by 𝐺,(𝒙, 𝒙′;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑡, 𝑡′) = 𝑖〈𝜓?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' "(𝒙), 𝑡′)𝜓?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (𝒙, 𝑡)〉 𝐺-(𝒙, 𝒙′;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝑡, 𝑡′) = −𝑖〈𝜓?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (𝒙, 𝑡)𝜓?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' "(𝒙), 𝑡′)〉 , (4) and are known as the "less than" and "greater than" Green\'s functions, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These are direct wave function products and do not have the anti- commutators that appear in (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The important one for our purposes is the less-than function, which has an important relation to the distribution function via 𝐺,(𝑘, 𝜔) = 𝑓(𝜔)𝐴(𝑘, 𝜔) , (5) which is known as the Kadanoff-Baym ansatz [28].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' There are some views that this does not satisfy causality, and another form is required [33].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, precisely this form will be found with the Airy transforms and will allow the determination of an equation for the distribution function 𝑓(𝜔).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The spectral density 𝐴(𝑘, 𝜔) retains exactly the same form as in equilibrium and is connected to the retarded and advanced functions through 𝐴(𝑘, 𝜔) = −𝐼𝑚{𝐺+(𝑘, 𝜔) − 𝐺((𝑘, 𝜔)} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (6) Each of these functions requires a pair of equations of motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' With the added complications of having four func- tions, these equations will be far more complicated than the simple Boltzmann equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Hence, solving for these functions, and obtaining the distribution function in the quantum case, is a difficult problem, both analytically and computationally [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Indeed, it would be much easier if an approach based, for example, on EMC could be found.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' But, the computational problems are not the only problems with NEGF, for the equations needed are not limited to just these functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the quantum transport situation, one must deal with the existence of long-range correlations among the carriers, correlations that are not always alleviated with scattering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' An example is given by impurity scattering, which is very long range due to the Coulomb interaction potential.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Even classically, it has been observed that an electron can be interacting with several impurities at the same time [34].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" In addition, with the wave accompanying the particle, a single Coulomb scattering center can create interference, much like a two-slit experiment, a result that has been found with both Green's functions [35] and Wigner functions [36]." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These correlations affect the quantum transport and can lead to observable fluctuations in devices [37].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Such long range correlations are not limited to impurity scattering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' They are expected for any long range potential;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' polar optical scattering is also Coulombic in nature and can be expected to lead to the same effects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Consequently, such correlations lead to difficulties when these scatterers are introduced perturbatively, as they require minimally evaluation of the Bethe-Salpeter equation for determination of mobility and transport [21,38].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This raises the difficulty level by a significant amount [39].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Airy Transforms One of the first requirements in quantum transport is the choice of wave functions and the corresponding basis set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the presence of an electric field that is taken along the z-direction for convenience, the one- dimensional eigenfunctions along this direction are determined from (time variation is assumed to enter in the normal manner, and only the z-variation is considered at the moment) !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='− ℏ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' "#∗ $!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' $%!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" + 𝑒𝐹𝑧'𝜑(𝑧) = 𝐸𝜑(𝑧) , (7) where the symbol F is used to denote the field in order to differentiate it from the energy E." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The wave functions in the two orthogonal directions (x,y) are taken to be normal plane waves that exist in the transverse band structure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' For this latter, the normal k and r will denote the wave vector and position in the transverse plane of (x,y)-coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As may be found in many quantum mechanics textbooks, the solutions to (7) are the Airy functions [40] 𝜑(𝒓, 𝑧) = .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' /01 𝑒!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='𝒌∙𝒓𝐴𝑖 H− 5#5" 1 I , (8) where 𝑧* is an appropriate zero of the Airy function 𝐴𝑖(∙).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The factor L is a normalizing length and is given by 𝐿 = H ℏ# /6∗78I .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='/9 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (9) The basis set of Airy functions has been used to find the analytical eigen-values for the triangular potential wells in a MOSFET for many years [40].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the MOSFET case, this provides the set of discrete bound state eigen-values and eigenfunctions determined by the discrete zeroes of the Airy function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is not the case of interest here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the present situation, a continuous Airy transform is to be used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The difference is exactly the same as the difference between a Fourier series and a Fourier transform.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The former is a discrete set while the latter is a continuous function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the present case, s well be the continuous transform variable and the discrete set of zeroes will not appear explicitly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This transform is defined just as a Fourier transform would be defined, the Airy transform of the function 𝑓(𝒓, 𝑧) is given as 𝐹(𝒌, 𝑠) = ∫ 𝑑/𝒓 ∫ :5 /01 𝑒!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='𝒌∙𝒓𝐴𝑖 H 5#;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1 I 𝑓(𝒓, 𝑧) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (10) For the Green’s function, there are two spatial variables, and hence one needs a double Airy 4 transform, although it is fair to assume homogeneity in the transverse dimensions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This transform may then be expressed as 𝐹(𝒌, 𝑠) = ∫ 𝑑/𝒓 ∫ :5 /01 ∫ :5% /01 𝑒!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='𝒌∙𝒓 × 𝐴𝑖 H 5#;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1 I 𝐴𝑖 H 5%#;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='% 1 I 𝑓(𝒓, 𝑧, 𝑧′) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (11) Here, the Green’s function response is only being developed for the one dimension along the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' One could, of course, also extend the equations to two variables in the transverse directions, but here only the one will be used to ease the complexity of the resulting equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' A function that is diagonal in both k and s is translationally invariant in the transverse plane, but not necessarily in the longitudinal plane along the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" Airy Form of NEGF The development of Green's function transport using the Airy functions is relatively old [41,42,43], and the present treatment is focused upon the development of the EMC method of its evaluation." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Consequently, the details of various derivations will be omitted here, but can also be found in [39].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 The Retarded and Advanced Functions In the transverse dimensions, the Green’s function is translationally invariant and so is a function only of 𝒓 − 𝒓′.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Solving the differential equation for the retarded and advanced functions leads to an integration over the intermediate variables which appears as a convolution integral.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' After the transforms are taken, the convolution becomes a simple product, so that the retarded Green’s function satisfies the inhomogeneous integral equation 𝐺+(𝒌, 𝑢, 𝑢′) = 𝐺* 8(𝒌, 𝑢, 𝑢′) + ∫𝑑𝑢.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' ∫ 𝑑𝑢/𝐺* 8(𝒌, 𝑢, 𝑢.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=') × Σ+(𝒌, 𝑢.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=', 𝑢/)𝐺+(𝒌, 𝑢/, 𝑢′) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (12) Here, the short-hand notation 𝑢 = (𝑧, 𝑡) has been introduced.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" In fact, in the case of significant correlation among the carriers, the argument of the integrals could properly be considered a three-particle Green's function, which would be incredibly difficult to evaluate [14,44,45]." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In most cases, it is assumed that it can be rewritten as the product of three single- particle functions as shown.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These can also lead to disconnected diagrams which cannot be discarded in the NEGF cases [31,46], as these generate important phase factors that affect any correlation and/or interference terms [47].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These higher-order effects must be considered and will be discussed in the Si model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The unperturbed Green’s function, the first term on the right-hand side of (12), includes the role of the electric field and may be taken to be [48] 𝐺* 8(𝒌, 𝑠, 𝑠), 𝑡, 𝑡′) = 𝐺+,* 8 (𝒌, 𝑠, 𝑠), 𝑡, 𝑡′) = − !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' ℏ 𝜃*(𝑡 − 𝑡′) × 𝑒#!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='$(𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=')?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='%#%%@/ℏ𝛿(𝑠 − 𝑠)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (13) The function 𝜃*(𝜉) is the normal Heavyside step function which is 1 for 𝜉 ≥ 0, and 0 otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, the energy is composed of the transverse energy plus that induced in the z-direction by the electric field, and this energy may be written as 𝐸𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' = 𝐸(𝒌, 𝑠) = ℏ#A# /6∗ + 𝑒𝐹𝑠 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (14) Both forms for the energy will be used below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The temporal Fourier transform of (13) is easily taken to be 𝐺* 8(𝒌, 𝑠, 𝑠), 𝑡, 𝑡′) = B?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='#;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='%@ ℏC#$(𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=')D!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='E .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (15) where the limit 𝜂 → 0 is usually taken, just as in the equilibrium Green’s function case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, the convergence factor is confusing, since it implies the field was turned on in the infinite past.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' But, this is not the case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As may be seen in (13), the field is turned on at the time 𝑡′, so this convergence factor is useful, but not the correct interpretation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Care has to be taken over this point, but it will not affect the final results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The full transformation with the Airy functions can now be taken to yield the equation for the retarded Green’s function to be 𝐺+Y𝒌, 𝑠, 𝑠), 𝜔 − 𝐸𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='Z = 𝐺* 8(𝒌, 𝑠, 𝜔)𝛿(𝑠 − 𝑠)) × ∫ :&𝒒 G0# ∑ [𝑀H[ / H𝑁H + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='DI / I IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝛿(𝜅) × 𝐺+(𝒘, 𝒘′, 𝜅′) (16) with the substitutions 𝜅 = 𝜔 − 𝜈𝜔* − 𝐸𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' + 𝐸𝒌%,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 𝜅′ = 𝜔 − 𝜈𝜔* − 𝐸𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='% 𝒘 = 𝒌 + 𝑘𝒛𝒂5 + 𝜈𝒒 𝒘′ = 𝒌′ + 𝑘𝒛 )𝒂5 + 𝜈𝒒 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" (17) The retarded self-energy is the product of the retarded Green's function and the retarded phonon Green's function." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In most semiconductors, the scattering is weak, and the self-energy can be calculated to lowest order in time-dependent perturbation theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" This means that the Green's function propagator is approximated by the free propagator (in the presence of the field) (15)." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" While this may suggest that physical effects of the scattering are being minimized, this isn't the real case." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" The retarded Green's function (16) is still being solved self- consistently, and the self-energy corrections will certainly be built into the result from Dyson's equation." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, a non-degenerate electron gas will be considered 5 as discussed below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is equivalent to neglecting any screening of the optical phonons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Moreover, the phonons are taken to be in equilibrium, so that the normal Bose-Einstein distribution, at the lattice temperature, can be used to describe the statistics of the phonons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The retarded phonon propagator is then 𝐷*,+(𝒒) = −𝑖𝜋 ∑ [𝑀H[ / IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' × H𝑁H + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='DI / I 𝛿(𝜅) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (18) The summation runs over emission and absorption of the optical phonon, with the phonon energy considered to be constant, 𝑀H is the electron-phonon matrix element, and 𝑁H is the Bose-Einstein distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This leads to the self-energy term as Σ+(𝒌, 𝒌′, 𝑧, 𝑧′, 𝜔) = ∫ :&𝒒 G0# ∑ [𝑀H[ / IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' × H𝑁H + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='DI / I 𝛿(𝜅)𝐺+(𝒘, 𝒘′, 𝜅′) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (19) While this still remains relatively simple, it is now necessary to take the Airy transform of this function, and this provides adequate complication.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" The Airy transform (11) is applied to both longitudinal variables, and may be expressed as Σ+(𝒌, 𝑠, 𝑠′, 𝜔) = ∫ :&𝒒 G0#1' ∑ [𝑀H[ / H𝑁H + ." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='DI / I IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' × ∫ 𝑑𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' ∫ 𝑑𝑠/ 𝐺+(𝒌 + 𝜈𝒒, 𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=', 𝑠/, 𝜅) × ∫ 𝑑𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' ∫ 𝑑𝑠/ ∫ 𝑑𝑧 ∫ 𝑑𝑧′𝐴𝑖(𝑧 − 𝑠) × 𝐴𝑖(𝑧 − 𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' )𝐴𝑖(𝑧′ − 𝑠′)𝐴𝑖(𝑧′ − 𝑠/) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (20) Here, the Airy functions have reduced arguments with the values 𝐴𝑖(𝑦) = 𝐴𝑖 H 78M#ℏC N I Θ = f3(𝑒ℏ𝐹)2 2𝑚∗ g 1/3 = 𝑒𝐹𝐿 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (21) Because 𝐺* 8 is diagonal in s, it may be expected that the resultant self-energy is also going to be diagonal in s once the transform has been taken.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Then, the integrals can be performed using the known properties of the Airy functions and their integrals [49].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The resulting self-energy is then found to be [41] Σ+(𝒌, 𝑠, 𝜔) = (6∗)&/#QR)Q #√N √/ℏ# × ∑ H𝑁H + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='DI / I 𝐹(𝑠, 𝜅) IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' , (22) with 𝐼𝑚{𝐹(𝑠, 𝜅)} = 𝐴𝑖′(−𝑦)𝐵𝑖′(−𝑦).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' −𝑦𝐴𝑖(−𝑦)𝐵𝑖(−𝑦) 𝑅𝑒{𝐹(𝑠, 𝜅)} = 𝐴𝑖′/(−𝑦) − 𝑦𝐴𝑖/(−𝑦) , (23) where Bi is the Airy function of the second kind, and the primes denote derivatives with respect to the argument.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These functions will be plotted below, once the material system has been described.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content="2 The Correlation Functions The less-than Green's function will originate from the above retarded functions, and will satisfy its own set of differential equations." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' From this less-than function, a distribution function will evolve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is the goal, to find the quantum non-equilibrium distribution function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This function will satisfy an integral equation that will be amenable to EMC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The differential equations to be solved are Fourier and Airy transformed to become jℏ𝜔 − 𝐸𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='l𝐺,(𝒌, 𝑠, 𝑠′, 𝜔) = = [Σ+(𝒌, 𝑠, 𝜔)𝐺,(𝒌, 𝑠′, 𝜔) +Σ,(𝒌, 𝑠, 𝜔)𝐺((𝒌, 𝑠′, 𝜔)] , (24) and jℏ𝜔 − 𝐸𝒌,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='%l𝐺,(𝒌, 𝑠, 𝑠′, 𝜔) =.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' = [𝐺+(𝒌, 𝑠, 𝜔)Σ,(𝒌, 𝑠′, 𝜔) +𝐺,(𝒌, 𝑠, 𝜔)Σ((𝒌, 𝑠′, 𝜔)] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (25) Normally, at this point, the sum and difference of these two equations are taken.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, however, this leads to the same equation [42], which may be directly solved for the less-than function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Taking the sum of the two equations, and performing a little algebraic reorganization, one finds that the less-than function reduces simply to [42] 𝐺,(𝒌, 𝑠, 𝑠′, 𝜔) = 𝐺+(𝒌, 𝑠, 𝜔) × 𝐺((𝒌, 𝑠′, 𝜔)Σ,(𝒌, 𝑠, 𝑠′, 𝜔) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (26) It is easily shown that the difference of the two equations leads to the same result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' One can easily also show that 𝐺+ − 𝐺( = 2𝐼𝑚{Σ+(𝒌, 𝑠, 𝜔)}𝐺+𝐺(.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These constraints allow us to rewrite the less-than function in the generalized Kadanoff-Baym manner as 𝐺,(𝒌, 𝑠, 𝜔) = 𝐴(𝒌, 𝑠, 𝜔)𝑓(𝑠, 𝜔) , 𝑓(𝑠, 𝜔) = T*(𝒔,,C) /V6{T+(;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=',C)} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (27) In this form, a quantum distribution function has been defined and is clearly not one of equilibrium.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Normally, in the Kadanoff-Baym form (5), the distribution function is only a function of 𝜔, but the parameter s has been retained here to indicate the dependence of the various paths for the particles on this quantity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While the form of this distribution function appears to be simple enough, care must be taken because the details of the less-than self-energy will certainly make this equation difficult.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Just as the retarded self-energy was developed, the less-than self-energy can be easily developed for the case of the non-polar intervalley phonons in Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 6 The self-energy can be written in the Airy transformed variables as Σ,(𝒌, 𝑠, 𝜔) = QR)Q# 9,/-1# ∑ H𝑁H + ID.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' / I IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' × ∫𝑑/𝒒% × ∫ 𝑑𝑠′𝐴𝑖/ H ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='#;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='% 9,/&1I × 𝐺,(𝒌 − 𝜈𝒒%, 𝑠′, 𝜔 − 𝜈𝜔*) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" (28) As this latter equation depends upon the less-than Green's function, it is clear that it will lead to an integral equation for the distribution function, that appears in (27)." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The integration over 𝒒% removes all dependence on the transverse momentum as it leads to a conservation of this momentum (all uncertainty now exists in the longitudinal change in momentum and this is in the spatial Airy variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As a result, the distribution function can be written as the integral equation [42Error!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Bookmark not defined.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='] 𝑓(𝑠, 𝜔) = .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' V6{T+(;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=',C)} ∑ H𝑁H + ID.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' / I IJ±.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' × ∫ 𝑑𝑠′𝐾(𝑠, 𝑠′, 𝜔 − 𝜈𝜔*)(𝑠′, 𝜔 − 𝜈𝜔*) (29) with 𝐾(𝑠, 𝑠′, 𝜔) = √9QR)Q #6∗ ℏ1# 𝐴𝑖/ H ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='#;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='% 9,/&1I p 0 / + 𝑎𝑡𝑎𝑛ℎ t ℏC#$𝒌,0%#Z7[T+?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=',;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='%,C@\\ V6{T+(,;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='%,C)} ug .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (30) While it may not seem evident, this integral equation for the distribution function may be readily solved with EMC techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the classical approach, the ballistic drift of the carriers is for a period of time that is determined from the probability that a carrier has not been scattering over that period of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is represented by the probability being a negative exponential of the time relative to the inverse of the scattering rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The same principle applies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, instead of the time, it is the ballistic distance 𝑠 − 𝑠), relative to the distance L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' From this distance, one can determine the ballistic time and other dynamic variables, as discussed below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Thus, one has the same iterated procedure as in classical EMC: drift for a period of time/space, then scattering according to the various processes in the less-than self-energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The leading term involving the imaginary part of the retarded self-energy serves merely to renormalize the distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='3 The Si Model To illustrate this new EMC approach to solving for NEGF, the case of Si will be considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While the situation for electrons in the conduction band is complicated by the multi-valley nature of this band, the scattering processes themselves are very local and do not require some to the more complicated higher- order corrections to the simpler Fermi golden rule, although the latter is of course different for the quantum case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The conduction band of Si has six equivalent ellipsoids located along the D lines [these are the set of (100) axes] about 85% of the way to the zone edge at X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Since scattering between these equivalent valleys is possible, the electric field will be taken along the (111) direction so that the same angle is made with each of the six valleys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Scattering within each ellipsoid is limited to acoustic phonons, as intra- valley optical processes are forbidden.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Acoustic mode scattering, by way of the deformation potential, is characterized by two constants Xu and Xd, which are thought to have values of 9 eV and -6 eV, respectively [50].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The effective deformation potential is then the sum of these, or about 3 eV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Non-polar “optical” phonon scattering occurs for scattering between the equivalent ellipsoids.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' There are two possible types of phonons that are involved in this process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' One is the g-phonon that couples the two valleys along opposite ends of the same (100) axis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is an umklapp process so that, after reduction by a reciprocal lattice vector, the scattering has a net phonon wave vector of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='3p/a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The symmetry allows only the LO mode to contribute to this scattering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' At the same time, f- phonons couple the (100) valley to the (010) and (001) valleys, and so on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The latter wave vector lies in the (110) direction has a magnitude of 21/2(0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='85)p/a = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2p/a, which lies in the square face of the Brillouin zone along the extension of the (110) line into the Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The scattering rates that enter the retarded self-energy are shown for Si at room temperature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These include the zero- and first-order intervalley optical phonons and the acoustic phonons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The rates are for a field of 10 kV/cm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' IV 0,em IV IV 1,em 1,em 13 IV 0,abs Acoustic 1x10 1x1010 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='8 1 Energy E(k,s) (eV) 7 second Brillouin zone [51].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The phonons here are near the X-point phonons in value but have a different symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The energies of the LO g-phonon and the LA and TO f-phonons have nearly the same value, while the low-energy inter-valley phonons are forbidden.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Long [52], however, determined from an analysis of the experimental mobility versus temperature that a weak low-energy inter-valley phonon is required to fit the experimental data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' He treated the high-energy phonons by a single equivalent inter-valley phonon of 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='3 meV, but introduced a low-energy inter-valley phonon with an energy of 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='4 meV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The presence of the low-energy phonons is also confirmed by studies of magneto-phonon resonance (where the phonon frequency is equal to a multiple of the cyclotron frequency) in Si inversion layers, which indicates that scattering by the low energy phonons is a weak contributor to the transport [53].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The low-energy phonon is certainly forbidden although Long treats it with a very weak coupling constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is more likely that this forbidden low-energy inter-valley phonon must be treated by a first-order interaction [54,55].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This fits the data with a coupling constant of X0 = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='6 eV, while the allowed, higher-energy transition is treated with a coupling constant D = 9 ´ 108 eV/cm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This two-phonon model is adopted for the present treatment of Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Non-parabolicity of the conduction bands away from the D minima arises more from repulsion from the upper (second) conduction along these directions than from a valence band interaction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Non-parabolicity is included in the simulations by the normal hyperbolic bands with an effective "gap" of 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 eV [51].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This non- parabolicity will be important in computing velocities from the energies and momentum determined during the Monte Carlo process described in the next section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The various scattering rates that go into the retarded self-energy, under the approximations described above, and for a field of 10 kV/cm, are shown in figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The real and imaginary parts of the retarded self-energy are shown in figure 2 for three values of the electric field (10, 40, 70 kV/cm).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The oscillations in the real part lead to the quasi- quantization of the density of states into two- dimensional sub-bands in the high electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The field tends to create a triangular potential well, which is expected from the quantum mechanics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' When the real part is positive, the energy is lowered, while when it is negative, the energy is raised.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This tends to form a set of sub-band energy levels [56].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This tendency toward formation of sub-bands is also evident in the imaginary part of the self-energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is clear, however, that the general shape of the imaginary part is only weakly dependent on the value of the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The zero of energy in these plots is taken as the conduction band minimum for the one-electron bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Broadening arises from both the scattering and the field, and this leads to values for negative energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The retarded self-energy leads to the spectral density which describes how the various energy levels are broadened by the scattering and quantization (a) (b) Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (a) The imaginary part of the retarded self-energy as a function of the energy itself.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (b) The real part of the retarded self-energy as a function of the energy itself.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These are determined using the scattering mechanisms described in the text for Si at 300 K and for three values of the electric field (10, 40, and 70 kV/cm).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='06 10 kV/cm 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 40 kVlcm 70 kV/cm 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='04 Im(Zr(E)) (eV) Si bulk 300 K FIl(111) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='01 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content="3 Energy E(k,s) (eV)4x10-3 10 kV/cm 3x10-3 40 kV/cm 70 kV/cm 2x10-3 Re{'(E)) (eV) 1x10-3 0 1x10-3 2x10-3 3x10-3 0." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 Energy E(k,s) (eV) 8 processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This spectral density is shown in figure 3 for the three field values used in illustrating the self- energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, it appears that, as with the imaginary part of the self-energy, the field has only a small effect on the spectral density.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Thus, it may be concluded that the retarded (and advanced) functions are not particularly dependent upon the actual value of the electric field applied to the device, other than the changes due to the tendency to form two-dimensional sub-bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These functions are more related to the intrinsic properties of the material used in the simulations, and not so much upon the actual value of the external field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The EMC Results Following from the above discussion, and the results of (29) and (30), the sequence of the Monte Carlo steps follow a similar path to that for the Boltzmann equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, however, the exponential in time is replaced by the Airy function in space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' That is, drift is defined by an effective path length first, and then scattering occurs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The difference between this approach and the Monte Carlo for the Boltzmann equation is that drift length replaces drift time, although these are intimately related, as will be described below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It may be noted that the imaginary part of the retarded self-energy sits outside the integral and is a final adjustment on the distribution function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' To begin, the spatial motion of the particle is determined with a random number r according to 𝑟 = ∫ 𝐴𝑖/(𝑢′)𝑑𝑢′ ] #]" ∫ 𝐴𝑖/(𝑢′)𝑑𝑢′ ^ #]" v , (31) so that a value for the drift length is determined from u as ∆𝑠 = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='/9𝐿𝑢 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (32) There is a question about the use of the Airy function because it has values over a semi-infinite range of values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, one may note that the scattering function in (30) seems to be over a single effective sub-band (the values of the bracketed term), so that the quantity −𝑢* is taken to be the first zero of the Airy function, so that there is a single maximum in the quantity integrated in (31).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This gives a range of values for ∆𝑠 for the ensemble of particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This may be averaged over the ensemble and over the number of iterations to determine an "average" value of ∆𝑠, and this is plotted as a function of the electric field in figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The error is extremely small, so that error bars do not show up;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' the average is computed over the 105 particles and for 200 iterations of the EMC algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While the velocity appears to converge rapidly, as discussed below, the number of iterations is used to assure that the distribution function is converged.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the EMC procedure, each particle has its own value for ∆𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This leads to an energy gain in the field of 𝑒𝐹∆𝑠, which is directed along the electric field (taken to be the z axis for convenience).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As a result, the momentum wave vector is changed by an amount ∆𝑘 = (𝑒𝐹/ℏ)∆𝑡 (but see a comment below).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is therefore necessary to find this value for ∆𝑡.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The drift time and the drift distance are trivially related in classical mechanics where the bands are parabolic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' With the non-parabolic bands that exist in Si, this is no longer the case, and some work is necessary to connect these two quantities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The problem arises from the energy dependence of the effective mass in the non- Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The spectral density obtained from the retarded (and advanced) self-energies for Si at 300 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The average drift length Ds as a function of the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The average is computed over 105 particles and 200 iterations of the Monte Carlo algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2x103 10 kV/cm 1x103 40 kV/cm Spectral Density (rel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' units) 70 kV/cm 8x102 6x10 4x102 2x10 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='01 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='02 hw - E(k,s)Si bulk 300 K 6 Average △S (nm) 5 4 3 2 20 40 60 80 100 120 0 Electric Field (kV/cm) 9 parabolic bands, which breaks the simple connection between momentum wave number and velocity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the hyperbolic bands used, the mass increases with energy according to [51] 𝑚∗ = 𝑚_(1 + 𝐸/𝐸`) , (33) where 𝑚_ is the mass at the conduction band minimum and 𝐸` is the effective energy gap mentioned in the previous section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" This is determined for the one- electron bands, as the broadening is added with the quantization in the Green's functions." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The change in velocity can be written as ∆𝑣(𝑡) = 78% 61 H1 + $(*) $2 + 78a% $2 I #.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' , (34) which leads to a difficult integral equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Here, 𝐸(0) is the energy at the start of the drift.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' We may take the distance as 𝑧 = 𝑣𝑡.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The latter suggests that the velocity be written as 𝑑𝑧/𝑑𝑡, so that (34) may be rewritten as :5 :% H1 + $(*) $2 + 78a% $2 I = 78 61 𝑡 , (35) with z running from 0 to ∆𝑠 and t running from 0 to ∆𝑡.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This integral is now trivially evaluated to lead to the value for the increment in time as ∆𝑡 = | /61 78 |∆𝑠| H1 + $(*) $2 + 78∆;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' /$2 I = | /61 78 |∆𝑠| H1 + $ $2 − 78∆;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' /$2 I , (34) where E is now the energy at the end of the drift.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Once this time increment is determined, the momentum wave vector can be updated, and the velocity determined at the end of the drift period.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It also allows one to determine the average drift time for the ensemble of carriers, in order to establish an average time scale for the process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It has to be noted that ∆𝑠 can be either positive or negative.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' A negative value merely means that the particle is actually being accelerated or decelerated in the direction opposite to the field (acceleration will occur for a negative momentum wave vector).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The value of ∆𝑡 should not depend upon this sign, but the change in momentum certainly does depend upon the sign.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Thus, the momentum change given above should actually account for this sign as ∆𝑘 = 78∆% ℏ 𝑠𝑖𝑔𝑛(∆𝑠) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" (35) One of the first steps to solving the Green's functions of section 4." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 is determining the less-than self-energy, especially the imaginary part that is necessary in (30).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is done within the EMC procedure and this scattering function is shown in figure 5 for the same three values of electric field used earlier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In contrast to the retarded self-energy, in this case the less-than self-energy clearly is a function of the applied electric field, with more scattering at higher electric fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This presumably is a result of more carriers at higher energies, where the scattering rates are higher, as shown in figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is also reflected in the shorter drift lengths apparent at higher values of the field in figure 4, although the higher field itself affects these lengths through the parameter L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The drift velocity is the average velocity of the ensemble of carriers, and is always computed by such an ensemble average.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This ensemble average is then averaged over the 200 iterations of the simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The resulting average drift velocity is plotted as.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' a function Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The imaginary part of the less-than self-energy is shown for three values of the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The drift velocity as a function of the electric field in Si at 300 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is determined from the ensemble averages and an average over the iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='3 10 kV/cm 40 kV/cm 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='25 70 kV/cm Im(Z(E)) (eV) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='3 Energy E(k,s) (eV)1 Drift Velocity (10′ cm/s) Si bulk 300 K EI (111) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 20 0 40 60 80 100 Electric Field (kV/cm) 10 of the electric field in figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While they are not evident, error bars are actually shown in the figure, for the average over the iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' These errors turn out to be less than 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='8% of the actual velocity, and so are buried under the symbol used in the plot.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It might seem natural to determine the average velocity from the ration ∆𝑠/∆𝑡.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" But, this would be a mistake as the latter quantity is the average over the drift length, and not the velocity at the end of the drift which is the quantity that leads to the drift velocity by virtue of Hamilton's equations of motion." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The average of the velocity over the iterations is used to reduce the noise, but in many cases that can be misleading when the actual velocity is strongly time dependent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is not the case here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The variation in the velocity during the simulation can be illustrated in another manner by plotting the velocity versus the average time that is determined from the average ∆𝑡 values, and this is done in figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Each value of the field has its own apparent time scale, which is the average steps per iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As may be seen from figure 4 and (34), these time scales differ for each value of the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In essence, it is perhaps better to consider these "time" scales as merely a parameter that measures the evolution of the ensemble over time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This differs from classical Monte Carlo, where the time is determined more or less accurately in the simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In this quantum version, it is the distance that is determined more or less accurately (we return to this point in the next section).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, after an initial value (after the first iteration) that is slightly higher than the subsequent values, the velocity is relatively constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The noise may be estimated by the fluctuations in figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The slightly higher value after the first iteration may be indicative of some velocity overshoot [57], but this cannot be determined properly, especially as the normal overshoot in Si is relatively small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While the common value expected for velocity saturation is around 107 cm/s, a higher value is found here (figure 6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is felt that the lower value found in semi-classical Monte Carlo arises from the rapid energy rise in scattering by the first-order coupled low energy intervalley phonon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, in the quantum case, it is found that this first-order scattering is especially affected by the intra-collisional field effect that arises in quantum transport [58].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Hence, it is not surprising that a larger "saturated" velocity is found in quantum transport.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In figure 8, the carrier population is plotted as a function of carrier energy for three different values of the electric field (same values as used earlier).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' As the field increases, carriers at low energy are lost to higher energy states, so that the distribution streams to higher energies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The various bumps and plateaus that tend to form arise from the tendency to have two-dimensional quantization and sub-bands in the high fields, as well as the role of onset of various phonons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The presence of particles at negative energies is the result of the broadening of the energy states, as represented by the spectral density of figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While the plot uses relative units (the actual values do not relate to the number of particles), the differences between the three curves are accurate, and reflect the field induced differences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The variation in the drift velocity for three values of the electric field, as a function of the simulated time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This time scale differs for each value of the field, as discussed in the text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The relative carrier population as a function of energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The bumps and plateaus are reflective of the tendency to two-dimensional quantization and some phonon effects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 10 kVlcm 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='4 40 kVlcm 70 kV/cm Drift Velocity (10′ cm/s) 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='6 10-12 10-11 10-14 10-13 10-10 Average Drift Time (s)1x10 10 kV/cm 40 kV/cm 70 kv/cm 1x10 Population (arb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' units) 1x10 1x102 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='2 Energy (eV) 11 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" Discussion In this paper, the use of the ensemble Monte Carlo process has been shown to be effective in evaluating non-equilibrium Green's functions in a fast and effective approach." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While this procedure is limited, at the present time, to materials in which the scattering is very local, such as with nonpolar optical and acoustical phonons, this application is of importance in most semiconductor devices used for integrated circuits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Typically, the computation for a single value of the electric field takes only on the order of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='6-2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='1 seconds on a modern laptop computer (here a MacBook Pro with the M1 Pro chip), with parallel processing available.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The results obtained for the drfit velocity in figure 6 are intriguing, in the sense that above ~40 kV/cm, there does not appear to be any hard saturation of the velocity, which is commonly seen in semi-classical Monte Carlo approaches [59,60], with a value of 1 × 10c cm/s at room temperature, although some versions give values well above this [61].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, this can be misleading, as there are a wide variety of such approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' However, measurements are not clear on this point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Most measurements do not go beyond 40-50 kV/cm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' For example, the oldest measurements using time-of-flight techniques extend to approximately 50 kV/cm [62], 13 kV/cm [63], or 20 kV/cm [64], so that the appearance of a hard saturation is not seen in the data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It has also been found that quantum simulations generally give higher values of velocity than the semi-classical ones [65,66].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The value of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content='3 × 10c cm/s found here is close to that found by Reggiani et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' [61].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' If one accepts that the quantum drift velocity is larger than that found semi-classically, there must be a reason.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' First, it has been known for some time that the collision broadening inherent in the spectral density contributes to this [61].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In addition, normally the Monte Carlo determines the ballistic drift length from a random time step derived from the total scattering rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In the quantum method presented here, the ballistic drift length is determined solely by the distance step that depends only upon the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Moreover, the important scattering arises from the imaginary part of the less-than self-energy (figure 5) and is this directly a function of the electric field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Thus, there is a direct inter-twining of the field and the scattering that contributes to the effect referred to as the intra-collisional field effect [2,67], a point emphasized earlier [61].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' But, there is a further contribution, apparent in (30).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This is the fact that the final states for scattering are an almost two- dimensional density of states, represented by the inverse hyperbolic tangent function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Quite generally, the density of states in a two-dimensional system is less than that in a three-dimensional system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This lower density of states gives less scattering and therefore a higher velocity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' So there are several reasons for the drift velocity to be higher in quantum simulations than in classical ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Another important consideration for the future is that the present approach has been applied for homogeneous material, while devices tend to be quite inhomogeneous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Extending this approach to inhomogeneous devices should be only a minor problem, as the Monte Carlo technique is quite usual in such device simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" Most of the computational time is taken with Poisson's equation, and from the results of this equation, the spatially dependent quantum distribution can be developed rather quickly." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" In devices, there is usually a grid upon which the density is projected in order to solve Poisson's equation." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Moreover, when Monte Carlo techniques are used for the transport in cases of real-space treatment of the impurities [68], or treating the electron-electron interaction by molecular dynamics [69], there are usually two time scales, one attached to each particle and the laboratory time scale for updating the total potential.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Techniques have been adopted for synchronizing these two times and distributing the acceleration between multiple bins of the laboratory time scale, since the potentials have to be updated on a much faster time scale than the scattering events [51].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' In principle, dividing the ∆𝑠 or ∆𝑡 values can easily be done by this same approach for either inhomogeneous material and devices or for incorporation of carrier-carrier interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Extending this approach to the polar optical phonons involves the problem of the long-range of Coulomb interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The polar interaction is a dipolar one, but still is quite inhomogeneous in the scattering dynamics just as is impurity scattering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' The latter has been made more amenable through treating the impurities in real space and developing the total potential from this real space approach [68].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' This has not been done for the polar modes as yet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" For solutions with the Green's functions, this should require direct solution of the Bethe-Salpeter equation for the very inhomogeneous scattering by the polar phonons." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=" However, this latter equation is itself an iterative procedure for high accuracy, and it is not beyond expectations that a second Monte Carlo process may be used for its evaluation, beyond the Monte Carlo procedure for determining the resultant Green's function and distribution function." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' It is also feasible to consider that the polar mode interaction may be expressed in a real space approach, such as used for the impurities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' While these suggestions remain to be speculation, it is certainly worth further research in this area to extend the present approach to devices and polar materials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 12 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Budd H F 1967 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 158 798 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Thornber K K and Feynman R F 1970 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 1 4099 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Kurosawa T 1966 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Jpn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Suppl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 21 424 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Jacoboni C and Reggiani L 1983 Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Mod.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 65 645 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Jacoboni C and Lugli P 1989 The Monte Carlo Method for Semiconductor Device Simulation (Vienna: Springer-Verlag) 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Hess K 1991 Monte Carlo Device Simulation: Full Band and Beyond (Boston: Kluwer Academic) 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Kosina H, Nedjalkov M and Selberherr S 2000 IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron Dev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 47 1898 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Shichijo H and Hess K 1981 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 23 4197 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lugli P and Ferry D K 1985 IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron Dev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 32 2431 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Joshi R P and Ferry D K 1991 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 43 9734 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Jacoboni C 1976 Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 13th Int.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Conf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Semiconductors (Rome: Tipo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Marves) 1195 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' De Broglie L 1924 Phil.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Mag.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=', Ser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 6, 47 446 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Kennard E H 1928 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 31 876 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Feynman R P 1948 Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Mod.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 20 367 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Wigner E 1932 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 40 749 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Heller, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Chem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 65, 1544 (1976).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Dittrich, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Viviescas, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sandoval, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 96, 070403 (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Vitanov P et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1994 Advances in Parallel Algorithms, Ed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' by B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sendov and I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Dimov (Amsterdam: IOS Press) pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 117-128 19.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Shifren L and Ferry D K 2002 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Comp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1 55 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nedjalkov M et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 2002 Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 7th Intern.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Conf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' on Simulation of Semiconductor Processes (New York: SISPAD) pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 187-190 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry D K et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 2021 Semicond.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Technol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 36 025024 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Fischetti M V 1984 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 53 1755 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Porod W and Ferry D K 1985 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 54 1189 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Jauho A P 1985 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 32 2248 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Reggiani L, Lugli P and Jauho A P 1987 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 36 6602 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Benseney A et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1988 Euro.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 68 286 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Baym G and Kadanoff L P 1961 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 124 287 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Kadanoff L P and Baym G 1962 Quantum Statistical Mechanics (New York: W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Benjamin) 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Schwinger J 1951 Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Acad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 37 452 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Schwinger J 1960 Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Acad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 46 883 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Dyson F J 1949 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 75 1736 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Prigogine I 2002 Phil.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Royal Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' A Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Engr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 360 299 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lipavský P, Ŝpiĉka V and Velický B 1986 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 34 6933 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Joshi R P and Ferry D K 1991 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 43 9734 35.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Barker J R 2003 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Comp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 2 153 References 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Weinbub J, Ballacchia M and Nedjalkov M 2018 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Stat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' RRL 12 18000111 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Gilbert M J and Ferry D K 2005 IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Nanotechnol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 4 355 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Vasileska D et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1995 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Vac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Technol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 13 1841 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry D K 2017 An Introduction to Quantum Transport in Semiconductors (Singapore: Jenny Stanford Publ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=') 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry D K 2021 Quantum Mechanics, 3rd Ed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (Boca Raton: CRC Press) 51 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Bertoncini R, Kriman A M and Ferry D K 1989 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 40 3371 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Bertoncini R, Kriman A M and Ferry D K 1990 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 41 1390 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Bertoncini R, Kriman A M and Ferry D K 1990 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' : Cond.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Matter 2 5991 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Dyson F 1951 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 82 428 45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Dyson F 1951 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 83 608 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Castro E R and Roditi I 2019 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' A 52 335401 47.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Bernevig B A and Regnault N 2009 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 103 206801 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Jauho A P and Wilkins J W 1984 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 29 1919 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Aspnes D E 1966 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 147 554 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ridley B K 1982 Quantum Processes in Semiconductors (Oxford: The Clarendon Press) 51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry D K 2020 Semiconductors: Bonds and Bands 2nd Ed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (Bristol: IOP Publishing) 52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Long D 1960 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 120 2024 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Eaves L et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1968 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' C 8 1975 54.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry D K 1976 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B14 1605 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Siegel W, Heinrich A, and Ziegler E 1976 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Stat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (a) 35 269 56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Reggiani L et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1991 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Stat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' (b) 168 K69 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ruch J G 1972 IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron Dev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 19 652 58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Ferry D K, Barker J R and Grubin H L 1981 IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron Dev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 28 905 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Osada Y 2012 IEEJ Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Info.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Sys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 132 1880 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Gada M L et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 2012 Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' NSTI Nanotechnol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Conf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 712 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Reggiani L, Lugli P and Jauho A-P 1988 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 64 3072 62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Canali C et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1973 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' B 12 2265 63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Tove P A et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1970 IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Electron Dev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 17 407 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Haas G A, Pankey Jr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' T and Harris F H 1973 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 44 2433 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Bertoncini R and Jauho A-P 1998 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 68 2826 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Wacker A et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 1999 Physica B 272 157 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Barker J R 1973 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' C 6 2663 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Vasileska D, Gross W J and Ferry D K 2000 Superlatt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Microstruc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 27 147 69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lugli P and Ferry D K 1986 Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
+page_content=' 56 1295' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/LdFAT4oBgHgl3EQfwh69/content/2301.08682v1.pdf'}
diff --git a/MdAyT4oBgHgl3EQfT_fX/content/tmp_files/2301.00118v1.pdf.txt b/MdAyT4oBgHgl3EQfT_fX/content/tmp_files/2301.00118v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..aff56d21a71b2c4b48f0eb8bdba90ccac342f48c
--- /dev/null
+++ b/MdAyT4oBgHgl3EQfT_fX/content/tmp_files/2301.00118v1.pdf.txt
@@ -0,0 +1,1013 @@
+arXiv:2301.00118v1 [physics.app-ph] 31 Dec 2022
+Airborne Ultrasound Focusing with Amplitude Mask
+Airborne Ultrasound Focusing Aperture with Binary Amplitude Mask
+Over Planar Ultrasound Emissions
+Masatake Kitano1 and Keisuke Hasegawa1, a)
+The Department of Mathematical Engineering and Information Physics, Faculty of Engineering, the University of Tokyo, Tokyo,
+113-8656, Japan.
+(*Electronic mail: keisuke_hasegawa@ipc.i.u-tokyo.ac.jp)
+(Dated: 3 January 2023)
+Phased arrays of airborne ultrasound transducers are widely utilized as a key technology to achieve mid-air conver-
+gence of intense ultrasound, which is applied to a variety of systems, such as contactless tactile presentation, acoustic-
+levitation and its application, mid-air-flow acceleration, etc. However, it requires considerably precise phase control
+with temporally severe synchronization between elements, which leads to difficulty in scaling up the entire system
+beyond the tabletop size as most of the current application systems. Here, we propose a much simpler and easier
+scaling-up method of airborne ultrasound convergence, where a binary amplitude mask that serves as a Fresnel Zone
+Plate (FZP) is placed on the planar in-phase ultrasound sources.
+We experimentally demonstrate that the FZP-based ultrasound focusing achieved a spatial resolution that is com-
+parable to conventional methods, based on the use of phase-controlled transducers. The ultrasound foci created using
+FZPs are sufficiently intense for most application scenarios that are currently in practical use. We also determine favor-
+able side effects of our method suppressing grating lobes, which is inevitable with the conventional phase-controlling
+method.
+The FZPs and planar ultrasound sources are both readily implemented with inexpensive ingredients and components.
+The result of our study contributes to upsizing dimensions in which a mid-air convergent ultrasound field is successfully
+generated. Accordingly, unprecedented application scenarios that target the entire room as the workspace will be
+possible.
+I.
+INTRODUCTION
+A.
+Prevalent use of phase-controlled airborne ultrasound
+transducer arrays in nonlinear mid-air ultrasound applications
+The nonlinear acoustic effect of strong mid-air ultrasound
+has been known1–4 for more than a century. However, its prac-
+tical applications had mostly been limited within underwater
+cases. Recently, however, the situation has been altered by the
+advent of wave emission control devices employed for gen-
+erating spatially localized intense ultrasound fields in the air,
+which can be electronically steered. Such strong and local-
+ized airborne ultrasonic power fields enable the generation of
+nonlinear acoustic phenomena at desired locations in space.
+This has led to the development of various applications of
+mid-air convergent ultrasound in several fields over the past
+decade.
+Examples of such real-world applications include
+mid-air ultrasound tactile presentation5–8, acoustic levitation
+systems9–11, mid-air three-dimensional displays12,13, utiliza-
+tion of mid-air ultrasound-driven acoustic flows14–16, etc. To
+date, new applications have been continuously devised by
+many researchers.
+As aforementioned, most of these applications rely on the
+technique of creating ultrasound fields with controlled spa-
+tial distribution, which is based on the principle of ultrasound
+a)Keisuke Hasegawa is the author to whom correspondence should be ad-
+dressed.
+source emission that is spatially controlled in a greater emis-
+sion area than the wavelength. Currently, the most prevalently
+employed method is the ultrasonic phased array technique,
+where the coherent ultrasound emission plane is constituted
+by a large number of ultrasonic transducers, with their indi-
+vidual emission amplitudes and phase delays electronically
+controlled. By appropriately controlling the driving phases
+of the transducers, a wide variety of spatial distributions of
+ultrasound fields are created. The fabrication of the first air-
+borne ultrasound phased array device5 triggered widespread
+research on its applications. Development of the airborne ul-
+trasound phased arrays have been continued to date by several
+research groups and most of the aforementioned application
+scenarios are based on this technique.
+B.
+Difficulty in upsizing current phased-array-based
+ultrasound manipulation scenarios
+However, the phased array technique suffers from difficul-
+ties in technical implementation17. Particularly, the need for
+synchronization and minute phase control of all individual
+transducers requiring a µs precision, prevents the workspace
+of mid-air ultrasound systems from being upscaled. There-
+fore, most of the current mid-air ultrasound applications are
+limited in tabletop systems, in terms of their spatial tract.
+As a potential solution that could address this challenge,
+there is a different approach to manipulate airborne ultra-
+sound. It is the development of wave manipulating elements
+that convert the incident ultrasound from a single fixed ultra-
+
+2
+sound source on its surface into desired spatial distribution of
+the ultrasound field. They are passive wave-manipulation el-
+ements and upsizing, and such devices are much easier and
+less expensive than upsizing the phased-array system for most
+cases. Among them, many phase controlling element arrays
+are studied, which serves as a phased array with fixed spa-
+tial phase delay profiles in conjunction with wave sources.
+One of the most intuitive ones are phase delay elements di-
+rectly placed above the plane wave source to convert the in-
+cident waves inside them into the desired field. They oper-
+ate in the water18, in vivo19, and in the air20,21. There are
+also a number of methods that handle incident waves at a dis-
+tance from the wave source that comes into the elements22,23.
+Several reflective elements that control the phase of incident
+wave have also been proposed both in water and air17,24,25 as
+well. Such devices do not require electronical synchroniza-
+tions among individuals once fabricated. At the same time,
+most of these elements do not allow their phase distribution
+to be changed once they are constructed, apart from some that
+can be adjusted manually20. Other examples of related tech-
+nologies for wave control are the use and fabrication of acous-
+tic metamaterials26–28.
+There is another approach for fabricating wave manipula-
+tion devices, which does not rely on phase control of inci-
+dent waves.
+Instead, such devices control only the ampli-
+tude distribution of the wave emissions. There is no need
+for the phase control of the vibrating surface, which sig-
+nificantly simplifies the system implementation; in addition,
+the required precision in element fabrication is substantially
+less than that of phase-controlling alternatives. In addition,
+it is advantageous in terms of the simplicity of the fabrica-
+tion in that amplitude-control-based devices do not require
+acoustic impedance matching to efficiently transmit incident
+waves, whereas this is indispensable for phase-controlling de-
+vices.
+The main drawback of this approach is that unlike
+the phase-controlling devices, a portion of radiated wave is
+weakened by amplitude control. This results in the require-
+ment of a larger ultrasound-emitting aperture in applications
+that require strong ultrasonic output when compared with the
+phase-control scheme.
+However, fabrication simpleness of
+such amplitude-controlling devices enables them to be made
+larger with less cost and effort than that required with upsiz-
+ing the phase-controlling devices. There are several examples
+of acoustic convergence in air and water achieved by concen-
+tric amplitude lenses29–34 and concentric transducer arrays35.
+With proper designing, the spatial resolution of the gener-
+ated sound fields by the amplitude-controlled emissions are
+not particularly degraded, compared with the phase-controlled
+cases.
+C.
+Proposed technique: Large-aperture amplitude mask on
+planar ultrasonic source for ultrasound manipulation
+The acoustic convergence using concentric amplitude emis-
+sions described above is a technique called Fresnel Zone Plate
+(FZP). Converged sound field is realized by placing the FZP
+at a certain distance from, or directly on radiating sound wave
+sources, to block off the ultrasound emissions that are not
+“in phase” at the desired region in the workspace in the air.
+Several airborne ultrasonic applications are implemented us-
+ing 40 kHz ultrasonic transducers. However, to the best of
+the authors’ knowledge, there have been no examples of the
+use of large-aperture FZPs for convergence of 40 kHz large-
+amplitude airborne ultrasonic waves in mid-air ultrasound ap-
+plications. FZPs can be easily scaled up in size, and the real-
+ization of a large-area ultrasonic radiation surface using FZP
+is expected to significantly expand the range of applications
+in mid-air ultrasound research, because of the ability to create
+well-concentrated ultrasound foci at a great distance from the
+aperture. For example, a mid-air tactile display will be real-
+ized that can present tactile stimuli all over the body of users
+at several locations in the room, whereas the current system
+can only stimulate a part of the user’s limb situated in front
+of the fixed small-apertured phased arrays. It is also expected
+that a wide range of aerial object manipulation using the entire
+room as a workspace will be achieved.
+As a fundamental technology to turn these potential appli-
+cations into reality, we propose a method for realizing a con-
+vergent sound field as an alternative to the phased array, by
+placing a thin, large-area FZP binary amplitude mask on a
+planar ultrasonic source. More specifically, we demonstrate
+the generation of an ultrasonic focus with this setup, which is
+often utilized in various applications including mid-air tactile
+presentation. The proposed FZP amplitude mask can be made
+from any material that has acoustic impedance sufficiently dif-
+ferent from that of the air and blocks off emissions from the
+plane wave source. In this study, we utilize an acrylic plate cut
+with a laser cutting machine to fabricate the FZP and demon-
+strate that a focus can be generated with it. In the proposed
+sound field control method, a machining accuracy of millime-
+ters is sufficient for 40 kHz ultrasonic waves (8.5 mm in wave-
+length in the air), which are commonly utilized in airborne
+ultrasonic research, and any of such complicated machining
+processes as required in fabricating metamaterials are unnec-
+essary. The proposed technique does not have a real-time fo-
+cus shifting function like the phased array. Nevertheless, the
+fabricated FZP mask can be larger than the area of the ultra-
+sound radiation surface, and the focus can be shifted by trans-
+lating it over the fixed radiation surface. Although not as easy
+as a phased array whose transducers’ phases can be electron-
+ically controlled, this focus shifting strategy can be achieved
+with appropriate actuators.
+In this study, it is assumed that a large number of ultrasonic
+transducers forming a large emitting aperture is utilized as the
+plane wave radiation surface. A reason for this assumption
+is the difficulty in fabricating a monolithic plane-wave radia-
+tion surface that utilizes a single flat plate to perform exclusive
+normal mode vibration at ultrasonic frequency. It is much eas-
+ier to construct a planar radiation surface using a large num-
+ber of separate ultrasonic transducers driving in phase instead.
+In this study, phased arrays that have already been developed
+were used as a plane wave source in the experiment by driving
+all their transducers with no phase differences among individ-
+uals. For actual applications, we envision the use of transducer
+arrays in which all elements are driven by a common driving
+
+3
+FIG. 1. Schematic illustration of how the ultrasonic focus is formed by phased array transducers (Left figure) and in-phase planar wave sources
+under an FZP (right figure).
+signal. This strategy does not require synchronization control
+of each element, unlike the case with phased arrays, and thus
+can be easily applied to large scale application systems.
+There is another finding in this paper that is concerned with
+the grating lobe issue, which is the strong and localized radia-
+tion of ultrasonic energy in an unintended direction, caused by
+phased arrays whose element spacing on their radiation plane
+is wider than half of the wavelength. In contrast, the spatial
+resolution of the amplitude mask fabricated in this study is
+finer than half of the wavelength; therefore, it is experimen-
+tally demonstrated that the afore-mentioned grating lobes do
+not occur when the in-phase driven transducer array is covered
+with the FZP mask. This feature in this study has great prac-
+tical significance, in that it suppresses people’s unintentional
+exposure to strong ultrasound in several application scenarios.
+II.
+PHYSICAL PRINCIPLES
+A.
+Ultrasound focusing by phased array systems
+Prior to the description of the formation principle of ultra-
+sound foci by FZPs, we start with brief introduction of focal
+formation by phased arrays: it has much in common with our
+method, and thus would bring better comprehension of this re-
+search. Figure 1 illustrates how the two strategies, the phase-
+controlled-transducer-based and FZP-based methods, form an
+ultrasound focus.
+As aforementioned, an airborne ultrasonic phased array
+has a radiation surface with many transducers, and the out-
+put signal of each transducer can be individually controlled.
+With a phased array, focus formation is achieved by electron-
+ically controlling the phase delay of each transducer so that
+the sound waves from all transducers are in-phase and yield
+strong acoustic energy spot at a desired point (the focus). Let
+rt be the position of an element in the array, rf be the focal
+point, and k be the ultrasonic wave number. Then, the driving
+phase θ(rt) of the ultrasonic wave emitted by the element at
+rt should be set to compensate for the phase delay owing to
+the distance between the element and focal point. Therefore,
+the driving phase θ(rt) is expressed as
+θ(rt) = k||rt − rf ||+ α,
+(1)
+where α is an arbitrary constant expressing the phase indefi-
+niteness and ||·|| denotes the Euclidean norm of a vector ·.
+B.
+Principles and designing procedures of amplitude FZP for
+ultrasound focusing
+The FZP converges ultrasound power around a desired po-
+sition by blocking off a part of the wave emission from an ul-
+trasonic source. Let a plane wave source be constructed by a
+set of in-phase driven ultrasonic elements and rt be an element
+position. Then at the focal position, the observed phase delay
+of the sound wave emitted from each element varies with the
+distance between the element and focus, as in the case with a
+phased array. Here, we consider driving the wave source with
+the rule that only those elements that are in phase or nearly in-
+phase at the focal point are activated, whereas the rest are de-
+activated. It is expected that nearly-in-phase addition of sound
+waves will be realized at the focus.
+The designing procedure of FZPs follows this principle.
+The amplitude distribution P(rt) on the FZP is generated from
+the distribution of the driving phase θ(rt) of the phased array
+element, calculated in Eq. (1) with arbitrary spatial distribu-
+tions. The most commonly used phase-to-amplitude conver-
+sion rules are as follows: 1) determine α so that the phase at
+
+UltrasoundFocus
+UltrasoundFocus
+FZP
+r
+UltrasoundSources
+UitrasoundSourceswithUniform
+withVariedDrivingPhases
+Driving Phases and FZP on them
+2元
+TT
+04
+FIG. 2. Calculated FZP patterns (left column) and normalized ultrasound amplitude fields by a continuous planar wave source under the FZPs
+in numerical simulations in the xy-plane (Middle column) and xz-plane (right column), for the focal depth of 150 mm and 400 mm, respectively.
+The coordinate system is as defined in the Fig. 6 illustrating the experiment setup.
+the point on the irradiation plane closest to the focus is zero,
+2) calculate the remainder of the driving phase divided by 2π,
+and 3) set the amplitude of the element to ON (P(rt) = 1)
+when the remainder is from zero to π and set the amplitude
+OFF (P(rt) = 0) otherwise:
+P(rt) =
+�
+1,
+2nπ ≤ θ(rt) < (2n + 1)π
+0, (2n + 1)π ≤ θ(rt) < 2(n + 1)π. ,n = 0,1,2,...
+(2)
+The above rule one is considered as effective in creating a
+strong sound field, because each element cannot be regarded
+as completely nondirectional in practice, and its ultrasound
+emission to the front is the strongest. In the following parts of
+the paper, we describe our investigations on the spatial proper-
+ties of the ultrasound field generated according to this method
+via numerical and real environment experiments.
+III.
+NUMERICAL EXPERIMENTS
+A.
+Calculation of acoustic wave convergence with amplitude
+FZP
+First, we evaluated the focusing performance of FZPs at-
+tached to an in-phase planer wave source. In real-environment
+experiments described in the following section, we utilized
+airborne ultrasonic phased arrays with each element driven in-
+phase as a plane wave source. Therefore, the size of the plane
+wave source in the numerical simulations was set to 370 mm
+× 290 mm, which is approximately equivalent to that of the
+actual phased arrays. We assumed that each point of the am-
+plitude on the FZP was a point source. Under these conditions,
+the ultrasound field was calculated using the wave superposi-
+tion principle. Two types of amplitude FZPs were designed
+in line with the above design criteria using Eq. (2) to con-
+verge sound around a focal point located apart from the plane
+wave source by 150 mm and 40 mm, respectively. The spa-
+tial resolution of the wave field calculation area was also set
+to 1 mm. The ultrasonic frequency was set to 40 kHz, which
+is widely utilized in current mid-air ultrasound applications.
+The FZP patterns were calculated with a spatial resolution of
+1 mm, less than 1/4 of the wavelength.
+Figure 2 illustrates a simulation result of the ultrasonic
+sound field including the focal point for each FZP. We adopted
+MATLAB for all numerical simulations in this study. The
+sound field generated by the designed FZPs indicates success-
+ful formation of an ultrasound focus for each case. Concentric
+acoustic emissions outside the focal region were observed in
+both FZPs in the xy-plane amplitude simulations.
+B.
+Focus formation by FZP using a plane wave source
+comprising multiple ultrasonic transducers
+As mentioned in the introduction, we consider forming the
+vibrating surface by arranging cylindrical aerial ultrasonic
+transducers in a two-dimensional grid, which are commer-
+cially available and have been utilized for several preceding
+studies.
+In this second numerical simulation, we assumed
+that each transducer could be modeled as a set of infinitesimal
+point sources uniformly distributed on its vibrating surface of
+10 mm diameter, which is equivalent to that of the transduc-
+ers utilized in the following experiments. The arrangement of
+the transducer in the simulation was set identical to the real
+devices employed in the real environment experiment. The
+
+O5
+FIG. 3. Simulation results for the case where FZPs are located on the transducer array driven with synchronized uniform phase delays.
+Calculated amplitude patterns (left column), normalized amplitude fields in numerical simulations in the xy-plane (middle column), and the
+xz-plane (right column), for the focal depths of 150 mm and 400 mm, respectively.
+FIG. 4. Simulation results for the case where the transducer array is driven with individual element phase delays for forming an ultrasound
+focus. Calculated amplitude patterns (left column), normalized amplitude fields in numerical simulations in the xy-plane (middle column) and
+the xz-plane (right column), for the focal depths of 150 mm and 400 mm.
+FZP pattern was superimposed onto the transducer array in
+which the driving signals of all transducers set were identi-
+cal. As with the previous simulations, the spatial distribution
+of the source FZP plane was set to 1 mm. For the fidelity
+of simulations compared with the real transducer arrangement
+on the phased arrays, periodic gaps were created between the
+cylinders and some regions were set where transducers were
+not mounted, that corresponded to the screws in the actual
+devices utilized in the experiments.
+Figure 3 illustrates the
+calculated source amplitude distributions and generated sound
+fields. The results demonstrate that adequate focusing was
+achieved with a periodically perforated emission plane com-
+prising a set of in-phase-driven transducers. Weak grid-like
+amplitude patterns that were superimposed on the focus in the
+xy-plane simulation are observed. These patterns were not
+observed in the amplitude distributions generated from the
+previous simulations, where no periodical source gaps were
+considered. Therefore, generation of these patterns can be as-
+
+016
+FIG. 5. Fabricated FZPs with focal depths of (a) 150 mm and (b) 400 mm.
+FIG. 6. Coordinate system in the experiments and experimental setup
+with four units of ultrasound phased arrays as planar wave sources,
+covered by the FZP.
+cribed to the periodical defects in the emission plane of the
+wave source under the FZPs. Apart from that, the generated
+patterns with these two simulations have great similarity to
+one another.
+Next, we evaluated the focal formation by the conventional
+phase controlling method of transducer arrays, for the relative
+assessment of the focusing performances of the FZP-based
+methods. The arrangement and amplitudes of the transduc-
+ers were set identical to those in the previous simulation. The
+output phase of each transducer was calculated in line with Eq.
+(1) with α = 0. Figure 4 illustrates the calculation results of
+the amplitude distribution by the phase-controlled transducer
+arrays. With the focal depth of 150 mm, prominent grating
+lobes are observed at a distance from the focal point, instead
+of the widespread granular amplitude patterns observed in the
+case with FZPs. This is owing to the fact that the spatial reso-
+lution of the phase distribution control of the radiating surface
+depends on the size of the transducer elements, and therefore
+cannot achieve a phase distribution finer than half a wave-
+length. The spatial distributions of FZPs we create in the ex-
+periment are finer than half the wavelength, and this mitigated
+the generation of the grating lobes. In the case where the focal
+depth is 400 mm, the grating lobes are located further from the
+focus, and cannot be seen in the region of simulation.
+The phased array is expected to be more efficient than FZPs
+in forming a focal point because the radiating surface is not
+shielded unlike the FZPs. As indicated in the next chapter,
+numerical simulation and real-environment experiments both
+indicate that the ultrasound intensity at the focus was reduced
+by using FZPs than performing phase control over all trans-
+ducers, when the output of each element was the same.
+IV.
+PHYSICAL EXPERIMENTS
+A.
+Ultrasound focusing by the binary hologram
+We fabricated two types of FZPs made of acrylic plate with
+2 mm of thickness, which have 150 mm and 500 mm of focal
+depth, respectively (Fig. 5). The acrylic sheets were cut out
+by a laser cutter (VD-60100, COMMAX, Co., Ltd., Japan)
+based on CAD data with the geometric positioning and size of
+each component defined with spatial quantization of 1 mm.
+In the experiment, we utilized custom-made 40 kHz phased
+arrays36 with their all transducers driven in-phase as a wave
+source, on which the fabricated FZP was placed to form an
+ultrasound focus. The custom-made phased array unit utilized
+in the experiment contained 249 40 kHz ultrasound transduc-
+ers (T4010A1, Nippon Ceramic, Co., Ltd., Japan) arranged
+in a two-dimensional lattice.
+We employed four units of
+the phased arrays forming an ultrasound emitting aperture
+
+FZPon
+Transducer
+Array
+X
+Z
+Microphone
+on
+Actuators(a) f = 0.15m
+(b) f = 0.4m
+0.42m7
+of 374.0 mm × 292.8 mm in the experiment, which corre-
+sponds to the spatial configuration of the numerical experi-
+ments. For ultrasound scanning measurement, a standard mi-
+crophone system (1/8-in. microphone, type 4138-A-015; pre-
+amplifier, type 2670; condition amplifier, type 2690-A; all
+products of Hottinger, Brüel and Kjær, Denmark) was utilized.
+The microphone was mounted on the tip of three-dimensional
+linear actuators (type ICSB3, product of IAI, Japan). The mi-
+crophone on the actuators scanned the sound field to capture
+the acoustic pressure distributions in designated regions. The
+entire experimental setup and coordinate system in the exper-
+iments are illustrated in Fig. 6.
+The scanning was completed with spatial interval of 5 mm
+for the x- and y- axes, and 10 mm for the z- axis illustrated in
+Fig. 6. The ultrasound output of the arrays was adequately
+weakened, compared with its possible maximum for avoid-
+ing measurement saturation of the microphone. Across all the
+measurements, the output intensity of the transducers were set
+identical. For each measurement, the center of the coordinate
+system in the xy-plane was slightly adjusted so that it yielded
+the maximum observed pressure within the range less than
+2 mm after manually setting the coordinate origin on the cen-
+ter of the arrays. In this manual calibration, the origin of the
+z-axis was set to the surface of the phased array transducers.
+While setting the transducer output power unchanged, we also
+sequentially measured the sound field created by the phase-
+controlled transducers at corresponding focal depths for the
+comparison of focusing performances among the two meth-
+ods.
+Figure 7 illustrates the results of the sound field measure-
+ments with the FZPs. As in the simulation experiment, it was
+confirmed that the generation of grating lobes was suppressed
+by the FZP with focal depth of 150 mm. Figure 8 illustrates
+spatially finer measurement results with 2 mm scanning in-
+tervals along the x- and z- axes with amplitude distributions
+obtained by the numerical simulation, normalized by the out-
+put obtained with the phase-controlled cases. The graphs in-
+dicate that the spatial profile of the amplitude distributions in-
+dicate good agreement with those predicted by numerical sim-
+ulations. The focal amplitudes with the FZPs clearly reduced
+when compared with that generated by the phase-controlled
+transducers. Simultaneously, the sizes of the focus with both
+cases with the FZP and phase-controlled transducers were
+comparable.
+Although the focal power efficiency was reduced, com-
+pared with the conventional phased array method, we consider
+it was still valid for most of the current mid-air ultrasound
+use. We confirmed that aerial vibrotactile presentation, one
+of such prevalent applications, can be achieved. By applying
+amplitude modulation at 150 Hz as is often done in airborne
+ultrasound presentation systems, distinct pinpoint tactile stim-
+uli could be felt by placing bare hands over the ultrasound
+focus generated by the FZPs for the possible maximum ultra-
+sound output from the transducers. From this result, it is ex-
+pected that FZP-based ultrasound manipulation can be applied
+in other scenarios that have been demonstrated by precedent
+studies.
+B.
+The mobility of the focus in binary holograms and
+multi-focusing
+Furthermore, we placed an FZP with focal depth of 150 mm
+on the in-phase transducer array. We shifted the center of the
+FZPs apart from the center of the arrays. We performed the
+sound field measurement as in the previous experiments and
+observed that the focus moved to the same position as the
+center of the FZP. Then, we placed another 150 mm-focal-
+depth FZP on that FZP and observed that two ultrasound foci
+were simultaneously generated at positions corresponding to
+the centers of both FZPs (Fig.9).
+The experiment demonstrates that the focal movement by
+only translational movement of FZP is realized, which is
+much easier than moving the entire sound sources toward de-
+sired focal positions. Simultaneously, the power of foci were
+approximately halved from that with a single focus case. Re-
+garding the case with two foci, the relative power of unwanted
+acoustic emission around the focal region was stronger than
+that observed by the single-layer-FZP case. For this case, vi-
+brotactile stimulation was faintly felt on the palm with 150 Hz
+amplitude modulation applied to the possible maximum ultra-
+sound output from the transducers.
+V.
+DISCUSSIONS
+A.
+Focusing performance for varied focal depths
+The sound pressure at the desired focus point with the
+same emission power was lower when using FZP than phase-
+controlled transducers, observed both in the numerical sim-
+ulations (59.1 % of the phase-controlled case for the focal
+depth of 150 mm and 38.2 % for 400 mm) and the measure-
+ment results (52.1 % of the phase-controlled case for the focal
+depth of 150 mm and 33.3 % for 400 mm). This tendency
+is more prominent with a longer focal depth, which can be
+attributed to the fact that the “rings” on the FZP get more dis-
+tant from one another for a longer focal depth, resulting in a
+smaller number of rings existing in a finite FZP aperture. For
+the most extreme case where the focus is fairly far from the
+FZP, there may be only one open circle in the FZP. In that
+case, the resulting ultrasound field is equivalent to that with a
+windowed planar emission, where no longer proper focusing
+is expected. In the same situation, a driving phase distribution
+on the emission plane is realized with the phase-controlled
+transducer case, which is still capable of forming a focus.
+The generation of grating lobes by the phase-controlled
+transducers was suppressed with the use of FZP, at the cost of
+unintended ultrasound emission around the focal region. This
+is because of the emission pattern with the FZP being finer
+than the wavelength, unlike that with phase-controlled trans-
+ducers inevitably becoming coarser owing to the transducer
+size. However, such strong grating lobes are not observed in
+the case with the focal depth of 400 mm in the measurement
+area. The grating lobes exist in a region more apart from the
+focus with a greater focal depth. Therefore, in the cases where
+the grating lobes are so far from the focus that they can be
+
+8
+FIG. 7. Measured ultrasound amplitude distribution when focus was formed by the FZPs (left column) or the phase controlling of transducer
+arrays (right column), for the focal depth of 150 mm ((a), (b), (e), and (f)) and 400 mm ((c), (d), (g), and (h)).
+neglected, the phase-controlled scheme outperforms the FZP-
+based method owing to less unintended acoustic emissions.
+The intervals of rings on the FZP also depend on the wave-
+length. As the wavelength decreases, the central circle be-
+comes smaller, and more rings exist in the FZP, which is ex-
+pected to improve sound collection performance. At the same
+time, it becomes much more difficult to decrease the dimen-
+sion of the phase-controlled transducers to realize proper fo-
+cusing. Therefore, FZP-based focusing scheme will be more
+suitable for ultrasound emission with a higher frequency.
+
+9
+FIG. 8. Measured and simulated acoustic amplitude distributions normalized by the maximum amplitude of phase-controlled cases (indicated
+as “PC” in the figure legends), for the focal depth of 150 mm (left column) and 400 mm (right column), along the x-axis (upper row) and the
+z-axis (lower row).
+B.
+Effect of FZP thickness
+In the numerical experiment, the thickness of the FZPs were
+not considered. In practice, direct waves to the focus were ex-
+pected to be partially blocked off by the thickness of FZPs.
+The smaller the elevation angle from the sound source to the
+focus is, the greater the part of sound waves is blocked, be-
+cause the sides of the FZP rings serve as walls. We utilized
+2 mm-thick acrylic plates for fabricating FZPs in our experi-
+ments, with which the blockage percentage by the FZP walls
+was estimated to be several percent compared to the case with
+FZP with no thickness.
+With an experiment where two FZPs were stacked to cre-
+ate two individual foci, the amplitudes of the foci were less
+than that of the single focus created with one FZP. This is pre-
+sumably because one FZP blocked the positive contribution to
+the focus from the other FZP, and the valid area of the planar
+sound source became considerably smaller. Another factor for
+this power reduction may be the total thickness of the two FZP
+layers being 4 mm, which might have caused more blocking-
+off effect of direct waves and complicated sound reflections
+between the layers.
+C.
+Variations of driving frequencies
+FZP-based focusing can achieve finer spatial resolution
+than the conventional phased-array technique. This resolution
+gap between the two methods is more prominent for a higher
+ultrasound frequency, because of the difficulty in downsizing
+transducers of the corresponding resonant frequency. How-
+ever, the spatial resolution of FZP patterning can be readily
+improved, as there have been a great bunch of minute ma-
+chining techniques including the laser cutting. In addition,
+a higher frequency source results in more rings created in
+FZPs, which will contribute to a more proper focusing. There-
+fore, our method can be effectively implemented as a form of
+miniature emission plane with a higher frequency source, as
+well as enlarged mid-air-ultrasound workspaces.
+VI.
+CONCLUSIONS
+In this paper, a method of controlling ultrasound fields us-
+ing an FZP amplitude binary mask on a plane wave source
+was developed. In the experiments, a 2 mm thick acrylic plate
+cut out by a laser cutting machine was utilized as a binary
+mask. We evaluated focusing performances with our proposed
+
+10
+FIG. 9. Measured acoustic amplitude distribution with shifted single
+FZP (Ueft figure) and two layers of FZPs (lower figure).
+method, where ultrasonic convergence occurs to the same de-
+gree of spatial resolution as in the case with a conventional
+method using phase-controlled transducers. As a favorable
+side effect, FZP suppressed grating lobes observed with the
+conventional method. We also determined that shifting the
+FZP over a fixed source can move the focus. Furthermore,
+multi-focusing was achieved by using layers of multiple FZPs
+with their centers corresponding to the focal positions.
+Configuration of our method is very simple.
+A planar
+source with its common driving voltage pulses and the FZPs
+can be made both inexpensive, thin, and readily implemented
+and upsized. Our method is suitable for mid-air ultrasound
+control system with large apertures or systems using ultra-
+sound sources with a higher frequency. Implementation of
+large ultrasound aperture will lead to new large-scale mid-
+air ultrasonic application systems that employ the entire walls
+and ceilings as ultrasonic emission planes.
+Subsequent challenges include electronical control of am-
+plitude distribution over the source. We consider that this sce-
+nario has the possibility to be realized, as it only needs on-off
+binary control of emission patterns, instead of inter-element
+phase control with the temporal preciseness less than one-
+tenth of the ultrasound period. We believe that such amplitude
+controlling mechanism requires less effort to be realized than
+that required to upsize the phased array by simply consolidat-
+ing a large number of array units with a tremendous number
+of transducers individually phase-controlled.
+ACKNOWLEDGMENTS
+This study was supported by JST, PRESTO Grant Number
+JPMJPR21R9, Japan.
+AUTHOR DECLARATIONS
+The authors have no conflicts of interest.
+AUTHOR CONTRIBUTIONS
+Masatake Kitano: Conceptualization (equal); Formal Anal-
+ysis (lead); Investigation (equal); Methodology (supportive);
+Writing – original draft (lead); Keisuke Hasegawa: Conceptu-
+alization (equal); Formal Analysis (supportive); Investigation
+(equal); Methodology (lead); Writing – review and editing
+(lead); Supervision (lead); Funding Acquisition (lead); Re-
+sources(lead); Project Administration(lead);
+DATA AVAILABILITY
+The data that support the findings of this study are available
+from the corresponding author upon reasonable request.
+1J.
+W.
+Strutt,
+“I.
+on
+the
+circulation
+of
+air
+observed
+in
+kundt’s
+tubes,
+and
+on
+some
+allied
+acoustical
+problems,”
+Philosophical Transactions of the Royal Society of London 175, 1–21 (1884),
+https://royalsocietypublishing.org/doi/pdf/10.1098/rstl.1884.0002.
+2L.
+V.
+King,
+“On
+the
+acoustic
+radiation
+pressure
+on
+spheres,”
+Proceedings of the Royal Society of London. Series A - Mathematical and Physical Sciences 147, 212–240 (1934),
+https://royalsocietypublishing.org/doi/pdf/10.1098/rspa.1934.0215.
+3P. J. Westervelt, “The theory of steady forces caused by sound waves,”
+The Journal of the Acoustical Society of America 23, 312–315 (1951),
+https://doi.org/10.1121/1.1906764.
+4P. J. Westervelt, “The theory of steady rotational flow generated by a sound
+field,” The Journal of the Acoustical Society of America 25, 60–67 (1953),
+https://doi.org/10.1121/1.1907009.
+5T. Hoshi, M. Takahashi, T. Iwamoto,
+and H. Shinoda, “Noncon-
+tact tactile display based on radiation pressure of airborne ultrasound,”
+IEEE Transactions on Haptics 3, 155–165 (2010).
+6R. Takahashi,
+K. Hasegawa,
+and H. Shinoda, “Tactile stimula-
+tion
+by
+repetitive
+lateral
+movement
+of
+midair
+ultrasound
+focus,”
+IEEE Transactions on Haptics 13, 334–342 (2020).
+7W. Frier, D. Ablart, J. Chilles, B. Long, M. Giordano, M. Obrist,
+and
+S. Subramanian, “Using spatiotemporal modulation to draw tactile patterns
+in mid-air,” in Haptics: Science, Technology, and Applications, edited by
+D. Prattichizzo, H. Shinoda, H. Z. Tan, E. Ruffaldi, and A. Frisoli (Springer
+International Publishing, Cham, 2018) pp. 270–281.
+8D. Hajas, D. Pittera, A. Nasce, O. Georgiou,
+and M. Obrist, “Mid-air
+haptic rendering of 2d geometric shapes with a dynamic tactile pointer,”
+IEEE Transactions on Haptics 13, 806–817 (2020).
+9R.
+Morales,
+A.
+Marzo,
+S.
+Subramanian,
+and
+D.
+Martínez,
+“Leviprops:
+Animating
+levitated
+optimized
+fab-
+ric
+structures
+using
+holographic
+acoustic
+tweezers,”
+in
+Proceedings of the 32nd Annual ACM Symposium on User Interface Software and Technology,
+UIST ’19 (Association for Computing Machinery, New York, NY, USA,
+2019) p. 651–661.
+
+11
+10A. Marzo,
+A. Ghobrial,
+L. Cox,
+M. Caleap,
+A. Croxford,
+and
+B.
+W.
+Drinkwater,
+“Realization
+of
+compact
+tractor
+beams
+using
+acoustic
+delay-lines,”
+Applied Physics Letters 110, 014102 (2017),
+https://doi.org/10.1063/1.4972407.
+11S. Inoue, S. Mogami, T. Ichiyama, A. Noda, Y. Makino, and H. Shinoda,
+“Acoustical boundary hologram for macroscopic rigid-body levitation,”
+The Journal of the Acoustical Society of America 145, 328–337 (2019),
+https://doi.org/10.1121/1.5087130.
+12R. Hirayama, D. Martinez Plasencia, N. Masuda, and S. Subramanian, “A
+volumetric display for visual, tactile and audio presentation using acoustic
+trapping,” Nature 575, 320–323 (2019).
+13R.
+Hirayama,
+G.
+Christopoulos,
+D.
+M.
+Plasencia,
+and
+S.
+Subramanian,
+“High-speed
+acoustic
+holography
+with
+arbi-
+trary
+scattering
+objects,”
+Science Advances 8, eabn7614 (2022),
+https://www.science.org/doi/pdf/10.1126/sciadv.abn7614.
+14K.
+Hasegawa,
+L.
+Qiu,
+A.
+Noda,
+S.
+Inoue,
+and
+H.
+Shin-
+oda,
+“Electronically
+steerable
+ultrasound-driven
+long
+nar-
+row
+air
+stream,”
+Applied Physics Letters 111, 064104 (2017),
+https://doi.org/10.1063/1.4985159.
+15K.
+Hasegawa,
+L.
+Qiu,
+and
+H.
+Shin-
+oda,
+“Midair
+ultrasound
+fragrance
+rendering,”
+IEEE Transactions on Visualization and Computer Graphics 24, 1477–1485 (2018).
+16K.
+Hasegawa,
+H.
+Yuki,
+and
+H.
+Shinoda,
+“Curved
+acceleration
+path
+of
+ultrasound-driven
+air
+flow,”
+Journal of Applied Physics 125, 054902 (2019),
+https://doi.org/10.1063/1.5052423.
+17J. Prat-Camps, G. Christopoulos, J. Hardwick,
+and S. Subramanian,
+“A manually reconfigurable reflective spatial sound modulator for ultra-
+sonic waves in air,” Advanced Materials Technologies 5, 2000041 (2020),
+https://onlinelibrary.wiley.com/doi/pdf/10.1002/admt.202000041.
+18K. Melde, A. G. Mark, T. Qiu, and P. Fischer, “Holograms for acoustics,”
+Nature 537, 518–522 (2016).
+19S.
+Jiménez-Gambín,
+N.
+Jiménez,
+and
+F.
+Camarena,
+“Tran-
+scranial
+focusing
+of
+ultrasonic
+vortices
+by
+acoustic
+holograms,”
+Phys. Rev. Applied 14, 054070 (2020).
+20G. Memoli, M. Caleap, M. Asakawa, D. R. Sahoo, B. W. Drinkwater, and
+S. Subramanian, “Metamaterial bricks and quantization of meta-surfaces,”
+Nature Communications 8, 14608 (2017).
+21S.-D. Zhao, A.-L. Chen, Y.-S. Wang,
+and C. Zhang, “Continuously
+tunable acoustic metasurface for transmitted wavefront modulation,”
+Phys. Rev. Applied 10, 054066 (2018).
+22M. Molerón, M. Serra-Garcia, and C. Daraio, “Acoustic fresnel lenses with
+extraordinary transmission,” Applied Physics Letters 105, 114109 (2014),
+https://aip.scitation.org/doi/pdf/10.1063/1.4896276.
+23Y. Li, G. Yu, B. Liang, X. Zou, G. Li, S. Cheng,
+and J. Cheng,
+“Three-dimensional ultrathin planar lenses by acoustic metamaterials,”
+Scientific Reports 4, 6830 (2014).
+24S. Polychronopoulos and G. Memoli, “Acoustic levitation with optimized
+reflective metamaterials,” Scientific Reports 10, 4254 (2020).
+25A. Sallam, V. C. Meesala,
+M. R. Hajj,
+and S. Shahab, “Holo-
+graphic mirrors for spatial ultrasound modulation in contactless acous-
+tic energy transfer systems,” Applied Physics Letters 119, 144101 (2021),
+https://doi.org/10.1063/5.0065489.
+26L. Zhao, E. Laredo, O. Ryan, A. Yazdkhasti, H.-T. Kim, R. Ganye, T. Ho-
+riuchi,
+and M. Yu, “Ultrasound beam steering with flattened acoustic
+metamaterial luneburg lens,” Applied Physics Letters 116, 071902 (2020),
+https://doi.org/10.1063/1.5140467.
+27T. Brunet, A. Merlin, B. Mascaro, K. Zimny, J. Leng, O. Poncelet,
+C. Aristégui, and O. Mondain-Monval, “Soft 3d acoustic metamaterial with
+negative index,” Nature Materials 14, 384–388 (2015).
+28S.
+Krödel
+and
+C.
+Daraio,
+“Microlattice
+metamaterials
+for
+tai-
+loring
+ultrasonic
+transmission
+with
+elastoacoustic
+hybridization,”
+Phys. Rev. Applied 6, 064005 (2016).
+29D. Schindel,
+A. Bashford,
+and D. Hutchins, “Focussing of ul-
+trasonic waves in air using a micromachined
+fresnel zone-plate,”
+Ultrasonics 35, 275–285 (1997).
+30Y.-X. Shen, Y.-G. Peng, F. Cai, K. Huang, D.-G. Zhao, C.-W. Qiu,
+H. Zheng, and X.-F. Zhu, “Ultrasonic super-oscillation wave-packets with
+an acoustic meta-lens,” Nature Communications 10, 3411 (2019).
+31J.
+T.
+Welter,
+S.
+Sathish,
+D.
+E.
+Christensen,
+P.
+G.
+Bro-
+drick,
+J.
+D.
+Heebl,
+and
+M.
+R.
+Cherry,
+“Focusing
+of
+longi-
+tudinal
+ultrasonic
+waves
+in
+air
+with
+an
+aperiodic
+flat
+lens,”
+The Journal of the Acoustical Society of America 130, 2789–2796 (2011),
+https://doi.org/10.1121/1.3640841.
+32M. D. Brown, J. Jaros, B. T. Cox, and B. E. Treeby, “Control of broadband
+optically generated ultrasound pulses using binary amplitude holograms,”
+The Journal of the Acoustical Society of America 139, 1637–1647 (2016),
+https://doi.org/10.1121/1.4944758.
+33N.
+Korozlu
+and
+A.
+Cicek,
+“Compact
+acoustic
+lens
+composed
+of
+annular
+cavities
+covered
+by
+a
+mem-
+brane,”
+Applied Physics Letters 113, 183504 (2018),
+https://doi.org/10.1063/1.5043600.
+34D. Tarrazó-Serrano, S. Pérez-López, P. Candelas, A. Uris,
+and C. Ru-
+bio, “Acoustic focusing enhancement in fresnel zone plate lenses,”
+Scientific Reports 9, 7067 (2019).
+35S. Hur, H. Choi, G. H. Yoon, N. W. Kim, D.-G. Lee,
+and Y. T.
+Kim, “Planar ultrasonic transducer based on a metasurface piezo-
+electric ring array for subwavelength acoustic focusing in water,”
+Scientific Reports 12, 1485 (2022).
+36S. Suzuki,
+S. Inoue,
+M. Fujiwara,
+Y. Makino,
+and H. Shin-
+oda,
+“Autd3:
+Scalable
+airborne
+ultrasound
+tactile
+display,”
+IEEE Transactions on Haptics 14, 740–749 (2021).
+
diff --git a/MdAyT4oBgHgl3EQfT_fX/content/tmp_files/load_file.txt b/MdAyT4oBgHgl3EQfT_fX/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0d1bc2711dc5a527b1da9aaeaf0f329ba099903b
--- /dev/null
+++ b/MdAyT4oBgHgl3EQfT_fX/content/tmp_files/load_file.txt
@@ -0,0 +1,585 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf,len=584
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='00118v1 [physics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='app-ph] 31 Dec 2022 Airborne Ultrasound Focusing with Amplitude Mask Airborne Ultrasound Focusing Aperture with Binary Amplitude Mask Over Planar Ultrasound Emissions Masatake Kitano1 and Keisuke Hasegawa1, a) The Department of Mathematical Engineering and Information Physics, Faculty of Engineering, the University of Tokyo, Tokyo, 113-8656, Japan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' (*Electronic mail: keisuke_hasegawa@ipc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='u-tokyo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='jp) (Dated: 3 January 2023) Phased arrays of airborne ultrasound transducers are widely utilized as a key technology to achieve mid-air conver- gence of intense ultrasound, which is applied to a variety of systems, such as contactless tactile presentation, acoustic- levitation and its application, mid-air-flow acceleration, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' However, it requires considerably precise phase control with temporally severe synchronization between elements, which leads to difficulty in scaling up the entire system beyond the tabletop size as most of the current application systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Here, we propose a much simpler and easier scaling-up method of airborne ultrasound convergence, where a binary amplitude mask that serves as a Fresnel Zone Plate (FZP) is placed on the planar in-phase ultrasound sources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We experimentally demonstrate that the FZP-based ultrasound focusing achieved a spatial resolution that is com- parable to conventional methods, based on the use of phase-controlled transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The ultrasound foci created using FZPs are sufficiently intense for most application scenarios that are currently in practical use.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We also determine favor- able side effects of our method suppressing grating lobes, which is inevitable with the conventional phase-controlling method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The FZPs and planar ultrasound sources are both readily implemented with inexpensive ingredients and components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The result of our study contributes to upsizing dimensions in which a mid-air convergent ultrasound field is successfully generated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Accordingly, unprecedented application scenarios that target the entire room as the workspace will be possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' INTRODUCTION A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Prevalent use of phase-controlled airborne ultrasound transducer arrays in nonlinear mid-air ultrasound applications The nonlinear acoustic effect of strong mid-air ultrasound has been known1–4 for more than a century.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' However, its prac- tical applications had mostly been limited within underwater cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Recently, however, the situation has been altered by the advent of wave emission control devices employed for gen- erating spatially localized intense ultrasound fields in the air, which can be electronically steered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Such strong and local- ized airborne ultrasonic power fields enable the generation of nonlinear acoustic phenomena at desired locations in space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This has led to the development of various applications of mid-air convergent ultrasound in several fields over the past decade.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Examples of such real-world applications include mid-air ultrasound tactile presentation5–8, acoustic levitation systems9–11, mid-air three-dimensional displays12,13, utiliza- tion of mid-air ultrasound-driven acoustic flows14–16, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' To date, new applications have been continuously devised by many researchers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As aforementioned, most of these applications rely on the technique of creating ultrasound fields with controlled spa- tial distribution, which is based on the principle of ultrasound a)Keisuke Hasegawa is the author to whom correspondence should be ad- dressed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' source emission that is spatially controlled in a greater emis- sion area than the wavelength.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Currently, the most prevalently employed method is the ultrasonic phased array technique, where the coherent ultrasound emission plane is constituted by a large number of ultrasonic transducers, with their indi- vidual emission amplitudes and phase delays electronically controlled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' By appropriately controlling the driving phases of the transducers, a wide variety of spatial distributions of ultrasound fields are created.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The fabrication of the first air- borne ultrasound phased array device5 triggered widespread research on its applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Development of the airborne ul- trasound phased arrays have been continued to date by several research groups and most of the aforementioned application scenarios are based on this technique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Difficulty in upsizing current phased-array-based ultrasound manipulation scenarios However, the phased array technique suffers from difficul- ties in technical implementation17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Particularly, the need for synchronization and minute phase control of all individual transducers requiring a µs precision, prevents the workspace of mid-air ultrasound systems from being upscaled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There- fore, most of the current mid-air ultrasound applications are limited in tabletop systems, in terms of their spatial tract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As a potential solution that could address this challenge, there is a different approach to manipulate airborne ultra- sound.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' It is the development of wave manipulating elements that convert the incident ultrasound from a single fixed ultra- 2 sound source on its surface into desired spatial distribution of the ultrasound field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' They are passive wave-manipulation el- ements and upsizing, and such devices are much easier and less expensive than upsizing the phased-array system for most cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Among them, many phase controlling element arrays are studied, which serves as a phased array with fixed spa- tial phase delay profiles in conjunction with wave sources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' One of the most intuitive ones are phase delay elements di- rectly placed above the plane wave source to convert the in- cident waves inside them into the desired field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' They oper- ate in the water18, in vivo19, and in the air20,21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There are also a number of methods that handle incident waves at a dis- tance from the wave source that comes into the elements22,23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Several reflective elements that control the phase of incident wave have also been proposed both in water and air17,24,25 as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Such devices do not require electronical synchroniza- tions among individuals once fabricated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' At the same time, most of these elements do not allow their phase distribution to be changed once they are constructed, apart from some that can be adjusted manually20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Other examples of related tech- nologies for wave control are the use and fabrication of acous- tic metamaterials26–28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There is another approach for fabricating wave manipula- tion devices, which does not rely on phase control of inci- dent waves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Instead, such devices control only the ampli- tude distribution of the wave emissions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There is no need for the phase control of the vibrating surface, which sig- nificantly simplifies the system implementation;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' in addition, the required precision in element fabrication is substantially less than that of phase-controlling alternatives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In addition, it is advantageous in terms of the simplicity of the fabrica- tion in that amplitude-control-based devices do not require acoustic impedance matching to efficiently transmit incident waves, whereas this is indispensable for phase-controlling de- vices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The main drawback of this approach is that unlike the phase-controlling devices, a portion of radiated wave is weakened by amplitude control.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This results in the require- ment of a larger ultrasound-emitting aperture in applications that require strong ultrasonic output when compared with the phase-control scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' However, fabrication simpleness of such amplitude-controlling devices enables them to be made larger with less cost and effort than that required with upsiz- ing the phase-controlling devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There are several examples of acoustic convergence in air and water achieved by concen- tric amplitude lenses29–34 and concentric transducer arrays35.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' With proper designing, the spatial resolution of the gener- ated sound fields by the amplitude-controlled emissions are not particularly degraded, compared with the phase-controlled cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Proposed technique: Large-aperture amplitude mask on planar ultrasonic source for ultrasound manipulation The acoustic convergence using concentric amplitude emis- sions described above is a technique called Fresnel Zone Plate (FZP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Converged sound field is realized by placing the FZP at a certain distance from, or directly on radiating sound wave sources, to block off the ultrasound emissions that are not “in phase” at the desired region in the workspace in the air.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Several airborne ultrasonic applications are implemented us- ing 40 kHz ultrasonic transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' However, to the best of the authors’ knowledge, there have been no examples of the use of large-aperture FZPs for convergence of 40 kHz large- amplitude airborne ultrasonic waves in mid-air ultrasound ap- plications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' FZPs can be easily scaled up in size, and the real- ization of a large-area ultrasonic radiation surface using FZP is expected to significantly expand the range of applications in mid-air ultrasound research, because of the ability to create well-concentrated ultrasound foci at a great distance from the aperture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For example, a mid-air tactile display will be real- ized that can present tactile stimuli all over the body of users at several locations in the room, whereas the current system can only stimulate a part of the user’s limb situated in front of the fixed small-apertured phased arrays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' It is also expected that a wide range of aerial object manipulation using the entire room as a workspace will be achieved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As a fundamental technology to turn these potential appli- cations into reality, we propose a method for realizing a con- vergent sound field as an alternative to the phased array, by placing a thin, large-area FZP binary amplitude mask on a planar ultrasonic source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' More specifically, we demonstrate the generation of an ultrasonic focus with this setup, which is often utilized in various applications including mid-air tactile presentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The proposed FZP amplitude mask can be made from any material that has acoustic impedance sufficiently dif- ferent from that of the air and blocks off emissions from the plane wave source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In this study, we utilize an acrylic plate cut with a laser cutting machine to fabricate the FZP and demon- strate that a focus can be generated with it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In the proposed sound field control method, a machining accuracy of millime- ters is sufficient for 40 kHz ultrasonic waves (8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='5 mm in wave- length in the air), which are commonly utilized in airborne ultrasonic research, and any of such complicated machining processes as required in fabricating metamaterials are unnec- essary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The proposed technique does not have a real-time fo- cus shifting function like the phased array.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Nevertheless, the fabricated FZP mask can be larger than the area of the ultra- sound radiation surface, and the focus can be shifted by trans- lating it over the fixed radiation surface.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Although not as easy as a phased array whose transducers’ phases can be electron- ically controlled, this focus shifting strategy can be achieved with appropriate actuators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In this study, it is assumed that a large number of ultrasonic transducers forming a large emitting aperture is utilized as the plane wave radiation surface.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' A reason for this assumption is the difficulty in fabricating a monolithic plane-wave radia- tion surface that utilizes a single flat plate to perform exclusive normal mode vibration at ultrasonic frequency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' It is much eas- ier to construct a planar radiation surface using a large num- ber of separate ultrasonic transducers driving in phase instead.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In this study, phased arrays that have already been developed were used as a plane wave source in the experiment by driving all their transducers with no phase differences among individ- uals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For actual applications, we envision the use of transducer arrays in which all elements are driven by a common driving 3 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Schematic illustration of how the ultrasonic focus is formed by phased array transducers (Left figure) and in-phase planar wave sources under an FZP (right figure).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' signal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This strategy does not require synchronization control of each element, unlike the case with phased arrays, and thus can be easily applied to large scale application systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There is another finding in this paper that is concerned with the grating lobe issue, which is the strong and localized radia- tion of ultrasonic energy in an unintended direction, caused by phased arrays whose element spacing on their radiation plane is wider than half of the wavelength.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In contrast, the spatial resolution of the amplitude mask fabricated in this study is finer than half of the wavelength;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' therefore, it is experimen- tally demonstrated that the afore-mentioned grating lobes do not occur when the in-phase driven transducer array is covered with the FZP mask.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This feature in this study has great prac- tical significance, in that it suppresses people’s unintentional exposure to strong ultrasound in several application scenarios.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' PHYSICAL PRINCIPLES A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ultrasound focusing by phased array systems Prior to the description of the formation principle of ultra- sound foci by FZPs, we start with brief introduction of focal formation by phased arrays: it has much in common with our method, and thus would bring better comprehension of this re- search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Figure 1 illustrates how the two strategies, the phase- controlled-transducer-based and FZP-based methods, form an ultrasound focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As aforementioned, an airborne ultrasonic phased array has a radiation surface with many transducers, and the out- put signal of each transducer can be individually controlled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' With a phased array, focus formation is achieved by electron- ically controlling the phase delay of each transducer so that the sound waves from all transducers are in-phase and yield strong acoustic energy spot at a desired point (the focus).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Let rt be the position of an element in the array, rf be the focal point, and k be the ultrasonic wave number.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Then, the driving phase θ(rt) of the ultrasonic wave emitted by the element at rt should be set to compensate for the phase delay owing to the distance between the element and focal point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Therefore, the driving phase θ(rt) is expressed as θ(rt) = k||rt − rf ||+ α, (1) where α is an arbitrary constant expressing the phase indefi- niteness and ||·|| denotes the Euclidean norm of a vector ·.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Principles and designing procedures of amplitude FZP for ultrasound focusing The FZP converges ultrasound power around a desired po- sition by blocking off a part of the wave emission from an ul- trasonic source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Let a plane wave source be constructed by a set of in-phase driven ultrasonic elements and rt be an element position.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Then at the focal position, the observed phase delay of the sound wave emitted from each element varies with the distance between the element and focus, as in the case with a phased array.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Here, we consider driving the wave source with the rule that only those elements that are in phase or nearly in- phase at the focal point are activated, whereas the rest are de- activated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' It is expected that nearly-in-phase addition of sound waves will be realized at the focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The designing procedure of FZPs follows this principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The amplitude distribution P(rt) on the FZP is generated from the distribution of the driving phase θ(rt) of the phased array element, calculated in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' (1) with arbitrary spatial distribu- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The most commonly used phase-to-amplitude conver- sion rules are as follows: 1) determine α so that the phase at UltrasoundFocus UltrasoundFocus FZP r UltrasoundSources UitrasoundSourceswithUniform withVariedDrivingPhases Driving Phases and FZP on them 2元 TT 04 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Calculated FZP patterns (left column) and normalized ultrasound amplitude fields by a continuous planar wave source under the FZPs in numerical simulations in the xy-plane (Middle column) and xz-plane (right column), for the focal depth of 150 mm and 400 mm, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The coordinate system is as defined in the Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 6 illustrating the experiment setup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' the point on the irradiation plane closest to the focus is zero, 2) calculate the remainder of the driving phase divided by 2π, and 3) set the amplitude of the element to ON (P(rt) = 1) when the remainder is from zero to π and set the amplitude OFF (P(rt) = 0) otherwise: P(rt) = � 1, 2nπ ≤ θ(rt) < (2n + 1)π 0, (2n + 1)π ≤ θ(rt) < 2(n + 1)π.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' ,n = 0,1,2,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' (2) The above rule one is considered as effective in creating a strong sound field, because each element cannot be regarded as completely nondirectional in practice, and its ultrasound emission to the front is the strongest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In the following parts of the paper, we describe our investigations on the spatial proper- ties of the ultrasound field generated according to this method via numerical and real environment experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' NUMERICAL EXPERIMENTS A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Calculation of acoustic wave convergence with amplitude FZP First, we evaluated the focusing performance of FZPs at- tached to an in-phase planer wave source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In real-environment experiments described in the following section, we utilized airborne ultrasonic phased arrays with each element driven in- phase as a plane wave source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Therefore, the size of the plane wave source in the numerical simulations was set to 370 mm × 290 mm, which is approximately equivalent to that of the actual phased arrays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We assumed that each point of the am- plitude on the FZP was a point source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Under these conditions, the ultrasound field was calculated using the wave superposi- tion principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Two types of amplitude FZPs were designed in line with the above design criteria using Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' (2) to con- verge sound around a focal point located apart from the plane wave source by 150 mm and 40 mm, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The spa- tial resolution of the wave field calculation area was also set to 1 mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The ultrasonic frequency was set to 40 kHz, which is widely utilized in current mid-air ultrasound applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The FZP patterns were calculated with a spatial resolution of 1 mm, less than 1/4 of the wavelength.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Figure 2 illustrates a simulation result of the ultrasonic sound field including the focal point for each FZP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We adopted MATLAB for all numerical simulations in this study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The sound field generated by the designed FZPs indicates success- ful formation of an ultrasound focus for each case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Concentric acoustic emissions outside the focal region were observed in both FZPs in the xy-plane amplitude simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Focus formation by FZP using a plane wave source comprising multiple ultrasonic transducers As mentioned in the introduction, we consider forming the vibrating surface by arranging cylindrical aerial ultrasonic transducers in a two-dimensional grid, which are commer- cially available and have been utilized for several preceding studies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In this second numerical simulation, we assumed that each transducer could be modeled as a set of infinitesimal point sources uniformly distributed on its vibrating surface of 10 mm diameter, which is equivalent to that of the transduc- ers utilized in the following experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The arrangement of the transducer in the simulation was set identical to the real devices employed in the real environment experiment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The O5 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Simulation results for the case where FZPs are located on the transducer array driven with synchronized uniform phase delays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Calculated amplitude patterns (left column), normalized amplitude fields in numerical simulations in the xy-plane (middle column), and the xz-plane (right column), for the focal depths of 150 mm and 400 mm, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Simulation results for the case where the transducer array is driven with individual element phase delays for forming an ultrasound focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Calculated amplitude patterns (left column), normalized amplitude fields in numerical simulations in the xy-plane (middle column) and the xz-plane (right column), for the focal depths of 150 mm and 400 mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' FZP pattern was superimposed onto the transducer array in which the driving signals of all transducers set were identi- cal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As with the previous simulations, the spatial distribution of the source FZP plane was set to 1 mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For the fidelity of simulations compared with the real transducer arrangement on the phased arrays, periodic gaps were created between the cylinders and some regions were set where transducers were not mounted, that corresponded to the screws in the actual devices utilized in the experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Figure 3 illustrates the calculated source amplitude distributions and generated sound fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The results demonstrate that adequate focusing was achieved with a periodically perforated emission plane com- prising a set of in-phase-driven transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Weak grid-like amplitude patterns that were superimposed on the focus in the xy-plane simulation are observed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' These patterns were not observed in the amplitude distributions generated from the previous simulations, where no periodical source gaps were considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Therefore, generation of these patterns can be as- 016 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Fabricated FZPs with focal depths of (a) 150 mm and (b) 400 mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Coordinate system in the experiments and experimental setup with four units of ultrasound phased arrays as planar wave sources, covered by the FZP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' cribed to the periodical defects in the emission plane of the wave source under the FZPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Apart from that, the generated patterns with these two simulations have great similarity to one another.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Next, we evaluated the focal formation by the conventional phase controlling method of transducer arrays, for the relative assessment of the focusing performances of the FZP-based methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The arrangement and amplitudes of the transduc- ers were set identical to those in the previous simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The output phase of each transducer was calculated in line with Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' (1) with α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Figure 4 illustrates the calculation results of the amplitude distribution by the phase-controlled transducer arrays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' With the focal depth of 150 mm, prominent grating lobes are observed at a distance from the focal point, instead of the widespread granular amplitude patterns observed in the case with FZPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This is owing to the fact that the spatial reso- lution of the phase distribution control of the radiating surface depends on the size of the transducer elements, and therefore cannot achieve a phase distribution finer than half a wave- length.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The spatial distributions of FZPs we create in the ex- periment are finer than half the wavelength, and this mitigated the generation of the grating lobes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In the case where the focal depth is 400 mm, the grating lobes are located further from the focus, and cannot be seen in the region of simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The phased array is expected to be more efficient than FZPs in forming a focal point because the radiating surface is not shielded unlike the FZPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As indicated in the next chapter, numerical simulation and real-environment experiments both indicate that the ultrasound intensity at the focus was reduced by using FZPs than performing phase control over all trans- ducers, when the output of each element was the same.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' PHYSICAL EXPERIMENTS A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ultrasound focusing by the binary hologram We fabricated two types of FZPs made of acrylic plate with 2 mm of thickness, which have 150 mm and 500 mm of focal depth, respectively (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The acrylic sheets were cut out by a laser cutter (VD-60100, COMMAX, Co.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=', Ltd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=', Japan) based on CAD data with the geometric positioning and size of each component defined with spatial quantization of 1 mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In the experiment, we utilized custom-made 40 kHz phased arrays36 with their all transducers driven in-phase as a wave source, on which the fabricated FZP was placed to form an ultrasound focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The custom-made phased array unit utilized in the experiment contained 249 40 kHz ultrasound transduc- ers (T4010A1, Nippon Ceramic, Co.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=', Ltd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=', Japan) arranged in a two-dimensional lattice.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We employed four units of the phased arrays forming an ultrasound emitting aperture FZPon Transducer Array X Z Microphone on Actuators(a) f = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='15m (b) f = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='4m 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='42m7 of 374.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='0 mm × 292.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='8 mm in the experiment, which corre- sponds to the spatial configuration of the numerical experi- ments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For ultrasound scanning measurement, a standard mi- crophone system (1/8-in.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' microphone, type 4138-A-015;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' pre- amplifier, type 2670;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' condition amplifier, type 2690-A;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' all products of Hottinger, Brüel and Kjær, Denmark) was utilized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The microphone was mounted on the tip of three-dimensional linear actuators (type ICSB3, product of IAI, Japan).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The mi- crophone on the actuators scanned the sound field to capture the acoustic pressure distributions in designated regions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The entire experimental setup and coordinate system in the exper- iments are illustrated in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The scanning was completed with spatial interval of 5 mm for the x- and y- axes, and 10 mm for the z- axis illustrated in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The ultrasound output of the arrays was adequately weakened, compared with its possible maximum for avoid- ing measurement saturation of the microphone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Across all the measurements, the output intensity of the transducers were set identical.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For each measurement, the center of the coordinate system in the xy-plane was slightly adjusted so that it yielded the maximum observed pressure within the range less than 2 mm after manually setting the coordinate origin on the cen- ter of the arrays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In this manual calibration, the origin of the z-axis was set to the surface of the phased array transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' While setting the transducer output power unchanged, we also sequentially measured the sound field created by the phase- controlled transducers at corresponding focal depths for the comparison of focusing performances among the two meth- ods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Figure 7 illustrates the results of the sound field measure- ments with the FZPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As in the simulation experiment, it was confirmed that the generation of grating lobes was suppressed by the FZP with focal depth of 150 mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Figure 8 illustrates spatially finer measurement results with 2 mm scanning in- tervals along the x- and z- axes with amplitude distributions obtained by the numerical simulation, normalized by the out- put obtained with the phase-controlled cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The graphs in- dicate that the spatial profile of the amplitude distributions in- dicate good agreement with those predicted by numerical sim- ulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The focal amplitudes with the FZPs clearly reduced when compared with that generated by the phase-controlled transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Simultaneously, the sizes of the focus with both cases with the FZP and phase-controlled transducers were comparable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Although the focal power efficiency was reduced, com- pared with the conventional phased array method, we consider it was still valid for most of the current mid-air ultrasound use.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We confirmed that aerial vibrotactile presentation, one of such prevalent applications, can be achieved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' By applying amplitude modulation at 150 Hz as is often done in airborne ultrasound presentation systems, distinct pinpoint tactile stim- uli could be felt by placing bare hands over the ultrasound focus generated by the FZPs for the possible maximum ultra- sound output from the transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' From this result, it is ex- pected that FZP-based ultrasound manipulation can be applied in other scenarios that have been demonstrated by precedent studies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The mobility of the focus in binary holograms and multi-focusing Furthermore, we placed an FZP with focal depth of 150 mm on the in-phase transducer array.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We shifted the center of the FZPs apart from the center of the arrays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We performed the sound field measurement as in the previous experiments and observed that the focus moved to the same position as the center of the FZP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Then, we placed another 150 mm-focal- depth FZP on that FZP and observed that two ultrasound foci were simultaneously generated at positions corresponding to the centers of both FZPs (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The experiment demonstrates that the focal movement by only translational movement of FZP is realized, which is much easier than moving the entire sound sources toward de- sired focal positions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Simultaneously, the power of foci were approximately halved from that with a single focus case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Re- garding the case with two foci, the relative power of unwanted acoustic emission around the focal region was stronger than that observed by the single-layer-FZP case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For this case, vi- brotactile stimulation was faintly felt on the palm with 150 Hz amplitude modulation applied to the possible maximum ultra- sound output from the transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' DISCUSSIONS A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Focusing performance for varied focal depths The sound pressure at the desired focus point with the same emission power was lower when using FZP than phase- controlled transducers, observed both in the numerical sim- ulations (59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1 % of the phase-controlled case for the focal depth of 150 mm and 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='2 % for 400 mm) and the measure- ment results (52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1 % of the phase-controlled case for the focal depth of 150 mm and 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='3 % for 400 mm).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This tendency is more prominent with a longer focal depth, which can be attributed to the fact that the “rings” on the FZP get more dis- tant from one another for a longer focal depth, resulting in a smaller number of rings existing in a finite FZP aperture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' For the most extreme case where the focus is fairly far from the FZP, there may be only one open circle in the FZP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In that case, the resulting ultrasound field is equivalent to that with a windowed planar emission, where no longer proper focusing is expected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In the same situation, a driving phase distribution on the emission plane is realized with the phase-controlled transducer case, which is still capable of forming a focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The generation of grating lobes by the phase-controlled transducers was suppressed with the use of FZP, at the cost of unintended ultrasound emission around the focal region.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This is because of the emission pattern with the FZP being finer than the wavelength, unlike that with phase-controlled trans- ducers inevitably becoming coarser owing to the transducer size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' However, such strong grating lobes are not observed in the case with the focal depth of 400 mm in the measurement area.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The grating lobes exist in a region more apart from the focus with a greater focal depth.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Therefore, in the cases where the grating lobes are so far from the focus that they can be 8 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Measured ultrasound amplitude distribution when focus was formed by the FZPs (left column) or the phase controlling of transducer arrays (right column), for the focal depth of 150 mm ((a), (b), (e), and (f)) and 400 mm ((c), (d), (g), and (h)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' neglected, the phase-controlled scheme outperforms the FZP- based method owing to less unintended acoustic emissions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The intervals of rings on the FZP also depend on the wave- length.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As the wavelength decreases, the central circle be- comes smaller, and more rings exist in the FZP, which is ex- pected to improve sound collection performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' At the same time, it becomes much more difficult to decrease the dimen- sion of the phase-controlled transducers to realize proper fo- cusing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Therefore, FZP-based focusing scheme will be more suitable for ultrasound emission with a higher frequency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 9 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Measured and simulated acoustic amplitude distributions normalized by the maximum amplitude of phase-controlled cases (indicated as “PC” in the figure legends), for the focal depth of 150 mm (left column) and 400 mm (right column), along the x-axis (upper row) and the z-axis (lower row).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Effect of FZP thickness In the numerical experiment, the thickness of the FZPs were not considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In practice, direct waves to the focus were ex- pected to be partially blocked off by the thickness of FZPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' The smaller the elevation angle from the sound source to the focus is, the greater the part of sound waves is blocked, be- cause the sides of the FZP rings serve as walls.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We utilized 2 mm-thick acrylic plates for fabricating FZPs in our experi- ments, with which the blockage percentage by the FZP walls was estimated to be several percent compared to the case with FZP with no thickness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' With an experiment where two FZPs were stacked to cre- ate two individual foci, the amplitudes of the foci were less than that of the single focus created with one FZP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This is pre- sumably because one FZP blocked the positive contribution to the focus from the other FZP, and the valid area of the planar sound source became considerably smaller.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Another factor for this power reduction may be the total thickness of the two FZP layers being 4 mm, which might have caused more blocking- off effect of direct waves and complicated sound reflections between the layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Variations of driving frequencies FZP-based focusing can achieve finer spatial resolution than the conventional phased-array technique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' This resolution gap between the two methods is more prominent for a higher ultrasound frequency, because of the difficulty in downsizing transducers of the corresponding resonant frequency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' How- ever, the spatial resolution of FZP patterning can be readily improved, as there have been a great bunch of minute ma- chining techniques including the laser cutting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In addition, a higher frequency source results in more rings created in FZPs, which will contribute to a more proper focusing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' There- fore, our method can be effectively implemented as a form of miniature emission plane with a higher frequency source, as well as enlarged mid-air-ultrasound workspaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' VI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' CONCLUSIONS In this paper, a method of controlling ultrasound fields us- ing an FZP amplitude binary mask on a plane wave source was developed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' In the experiments, a 2 mm thick acrylic plate cut out by a laser cutting machine was utilized as a binary mask.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We evaluated focusing performances with our proposed 10 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Measured acoustic amplitude distribution with shifted single FZP (Ueft figure) and two layers of FZPs (lower figure).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' method, where ultrasonic convergence occurs to the same de- gree of spatial resolution as in the case with a conventional method using phase-controlled transducers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' As a favorable side effect, FZP suppressed grating lobes observed with the conventional method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We also determined that shifting the FZP over a fixed source can move the focus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Furthermore, multi-focusing was achieved by using layers of multiple FZPs with their centers corresponding to the focal positions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Configuration of our method is very simple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' A planar source with its common driving voltage pulses and the FZPs can be made both inexpensive, thin, and readily implemented and upsized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Our method is suitable for mid-air ultrasound control system with large apertures or systems using ultra- sound sources with a higher frequency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Implementation of large ultrasound aperture will lead to new large-scale mid- air ultrasonic application systems that employ the entire walls and ceilings as ultrasonic emission planes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subsequent challenges include electronical control of am- plitude distribution over the source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We consider that this sce- nario has the possibility to be realized, as it only needs on-off binary control of emission patterns, instead of inter-element phase control with the temporal preciseness less than one- tenth of the ultrasound period.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' We believe that such amplitude controlling mechanism requires less effort to be realized than that required to upsize the phased array by simply consolidat- ing a large number of array units with a tremendous number of transducers individually phase-controlled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' ACKNOWLEDGMENTS This study was supported by JST, PRESTO Grant Number JPMJPR21R9, Japan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' AUTHOR DECLARATIONS The authors have no conflicts of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' AUTHOR CONTRIBUTIONS Masatake Kitano: Conceptualization (equal);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Formal Anal- ysis (lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Investigation (equal);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Methodology (supportive);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Writing – original draft (lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Keisuke Hasegawa: Conceptu- alization (equal);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Formal Analysis (supportive);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Investigation (equal);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Methodology (lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Writing – review and editing (lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Supervision (lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Funding Acquisition (lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Re- sources(lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Project Administration(lead);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' DATA AVAILABILITY The data that support the findings of this study are available from the corresponding author upon reasonable request.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 1J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Strutt, “I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' on the circulation of air observed in kundt’s tubes, and on some allied acoustical problems,” Philosophical Transactions of the Royal Society of London 175, 1–21 (1884), https://royalsocietypublishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/doi/pdf/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1098/rstl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1884.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='0002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 2L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' King, “On the acoustic radiation pressure on spheres,” Proceedings of the Royal Society of London.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Series A - Mathematical and Physical Sciences 147, 212–240 (1934), https://royalsocietypublishing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/doi/pdf/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1098/rspa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1934.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='0215.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 3P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Westervelt, “The theory of steady forces caused by sound waves,” The Journal of the Acoustical Society of America 23, 312–315 (1951), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1121/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1906764.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 4P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Westervelt, “The theory of steady rotational flow generated by a sound field,” The Journal of the Acoustical Society of America 25, 60–67 (1953), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1121/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1907009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 5T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hoshi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Takahashi, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Iwamoto, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shinoda, “Noncon- tact tactile display based on radiation pressure of airborne ultrasound,” IEEE Transactions on Haptics 3, 155–165 (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 6R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Takahashi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hasegawa, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shinoda, “Tactile stimula- tion by repetitive lateral movement of midair ultrasound focus,” IEEE Transactions on Haptics 13, 334–342 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 7W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Frier, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ablart, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Chilles, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Long, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Giordano, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Obrist, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subramanian, “Using spatiotemporal modulation to draw tactile patterns in mid-air,” in Haptics: Science, Technology, and Applications, edited by D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Prattichizzo, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shinoda, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Tan, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ruffaldi, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Frisoli (Springer International Publishing, Cham, 2018) pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 270–281.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 8D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hajas, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Pittera, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Nasce, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Georgiou, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Obrist, “Mid-air haptic rendering of 2d geometric shapes with a dynamic tactile pointer,” IEEE Transactions on Haptics 13, 806–817 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 9R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Morales, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Marzo, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subramanian, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Martínez, “Leviprops: Animating levitated optimized fab- ric structures using holographic acoustic tweezers,” in Proceedings of the 32nd Annual ACM Symposium on User Interface Software and Technology, UIST ’19 (Association for Computing Machinery, New York, NY, USA, 2019) p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 651–661.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 11 10A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Marzo, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ghobrial, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cox, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Caleap, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Croxford, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Drinkwater, “Realization of compact tractor beams using acoustic delay-lines,” Applied Physics Letters 110, 014102 (2017), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='4972407.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 11S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Inoue, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Mogami, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ichiyama, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Noda, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Makino, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shinoda, “Acoustical boundary hologram for macroscopic rigid-body levitation,” The Journal of the Acoustical Society of America 145, 328–337 (2019), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1121/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='5087130.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 12R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hirayama, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Martinez Plasencia, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Masuda, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subramanian, “A volumetric display for visual, tactile and audio presentation using acoustic trapping,” Nature 575, 320–323 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 13R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hirayama, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Christopoulos, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Plasencia, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subramanian, “High-speed acoustic holography with arbi- trary scattering objects,” Science Advances 8, eabn7614 (2022), https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='science.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/doi/pdf/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1126/sciadv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='abn7614.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 14K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hasegawa, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Qiu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Noda, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Inoue, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shin- oda, “Electronically steerable ultrasound-driven long nar- row air stream,” Applied Physics Letters 111, 064104 (2017), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='4985159.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 15K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hasegawa, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Qiu, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shin- oda, “Midair ultrasound fragrance rendering,” IEEE Transactions on Visualization and Computer Graphics 24, 1477–1485 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 16K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hasegawa, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Yuki, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shinoda, “Curved acceleration path of ultrasound-driven air flow,” Journal of Applied Physics 125, 054902 (2019), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='5052423.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 17J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Prat-Camps, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Christopoulos, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hardwick, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subramanian, “A manually reconfigurable reflective spatial sound modulator for ultra- sonic waves in air,” Advanced Materials Technologies 5, 2000041 (2020), https://onlinelibrary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='wiley.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='com/doi/pdf/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1002/admt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='202000041.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 18K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Melde, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Mark, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Qiu, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Fischer, “Holograms for acoustics,” Nature 537, 518–522 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 19S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Jiménez-Gambín, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Jiménez, and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Camarena, “Tran- scranial focusing of ultrasonic vortices by acoustic holograms,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Applied 14, 054070 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 20G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Memoli, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Caleap, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Asakawa, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Sahoo, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Drinkwater, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Subramanian, “Metamaterial bricks and quantization of meta-surfaces,” Nature Communications 8, 14608 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 21S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zhao, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Chen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Wang, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zhang, “Continuously tunable acoustic metasurface for transmitted wavefront modulation,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Applied 10, 054066 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 22M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Molerón, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Serra-Garcia, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Daraio, “Acoustic fresnel lenses with extraordinary transmission,” Applied Physics Letters 105, 114109 (2014), https://aip.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='scitation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/doi/pdf/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='4896276.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 23Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Li, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Yu, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Liang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zou, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Li, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cheng, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cheng, “Three-dimensional ultrathin planar lenses by acoustic metamaterials,” Scientific Reports 4, 6830 (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 24S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Polychronopoulos and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Memoli, “Acoustic levitation with optimized reflective metamaterials,” Scientific Reports 10, 4254 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 25A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Sallam, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Meesala, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hajj, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shahab, “Holo- graphic mirrors for spatial ultrasound modulation in contactless acous- tic energy transfer systems,” Applied Physics Letters 119, 144101 (2021), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='0065489.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 26L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zhao, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Laredo, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ryan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Yazdkhasti, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Kim, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ganye, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ho- riuchi, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Yu, “Ultrasound beam steering with flattened acoustic metamaterial luneburg lens,” Applied Physics Letters 116, 071902 (2020), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='5140467.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 27T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Brunet, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Merlin, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Mascaro, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zimny, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Leng, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Poncelet, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Aristégui, and O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Mondain-Monval, “Soft 3d acoustic metamaterial with negative index,” Nature Materials 14, 384–388 (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 28S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Krödel and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Daraio, “Microlattice metamaterials for tai- loring ultrasonic transmission with elastoacoustic hybridization,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Applied 6, 064005 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 29D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Schindel, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Bashford, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hutchins, “Focussing of ul- trasonic waves in air using a micromachined fresnel zone-plate,” Ultrasonics 35, 275–285 (1997).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 30Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Peng, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cai, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Huang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zhao, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Qiu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zheng, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Zhu, “Ultrasonic super-oscillation wave-packets with an acoustic meta-lens,” Nature Communications 10, 3411 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 31J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Welter, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Sathish, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Christensen, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Bro- drick, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Heebl, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cherry, “Focusing of longi- tudinal ultrasonic waves in air with an aperiodic flat lens,” The Journal of the Acoustical Society of America 130, 2789–2796 (2011), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1121/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='3640841.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 32M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Brown, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Jaros, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cox, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Treeby, “Control of broadband optically generated ultrasound pulses using binary amplitude holograms,” The Journal of the Acoustical Society of America 139, 1637–1647 (2016), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1121/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='4944758.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 33N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Korozlu and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Cicek, “Compact acoustic lens composed of annular cavities covered by a mem- brane,” Applied Physics Letters 113, 183504 (2018), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='5043600.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 34D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Tarrazó-Serrano, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Pérez-López, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Candelas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Uris, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Ru- bio, “Acoustic focusing enhancement in fresnel zone plate lenses,” Scientific Reports 9, 7067 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 35S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Hur, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Choi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Yoon, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Kim, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Lee, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Kim, “Planar ultrasonic transducer based on a metasurface piezo- electric ring array for subwavelength acoustic focusing in water,” Scientific Reports 12, 1485 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' 36S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Suzuki, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Inoue, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Fujiwara, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Makino, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
+page_content=' Shin- oda, “Autd3: Scalable airborne ultrasound tactile display,” IEEE Transactions on Haptics 14, 740–749 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/MdAyT4oBgHgl3EQfT_fX/content/2301.00118v1.pdf'}
diff --git a/MtFOT4oBgHgl3EQf1jQj/content/2301.12939v1.pdf b/MtFOT4oBgHgl3EQf1jQj/content/2301.12939v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..2ee8309669c4e80fb7b30422b33e19ef0d3e6b7d
--- /dev/null
+++ b/MtFOT4oBgHgl3EQf1jQj/content/2301.12939v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6a5288eadd39bd366178458c5e243022e1368f8af86b42e5e763bd239c8b7a7
+size 650246
diff --git a/MtFOT4oBgHgl3EQf1jQj/vector_store/index.faiss b/MtFOT4oBgHgl3EQf1jQj/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..08af4ccb2bc879e1e563a455aa657e29f27022d7
--- /dev/null
+++ b/MtFOT4oBgHgl3EQf1jQj/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db7c9dc167be03473dd9d6b33f9c8304b49e25c1ad6b7d26e66dd2a2bd21950c
+size 2162733
diff --git a/N9AyT4oBgHgl3EQftPlh/content/2301.00591v1.pdf b/N9AyT4oBgHgl3EQftPlh/content/2301.00591v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..95489560262f0933f3ac4e4591bbab83e7f7b0bc
--- /dev/null
+++ b/N9AyT4oBgHgl3EQftPlh/content/2301.00591v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58a020f1a553754a274a8e726b1505885b2d8ddb9488e5a5e7e074c3bda2bc23
+size 1431749
diff --git a/N9AyT4oBgHgl3EQftPlh/vector_store/index.faiss b/N9AyT4oBgHgl3EQftPlh/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..1fc84d5eba5258904f818638d01e1558f2bf8a91
--- /dev/null
+++ b/N9AyT4oBgHgl3EQftPlh/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:495cdde1d5bb90df00b449b8be1a302711cf9bd4d3f1e631299b690075440bf2
+size 1638445
diff --git a/N9AyT4oBgHgl3EQftPlh/vector_store/index.pkl b/N9AyT4oBgHgl3EQftPlh/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..4df8c9af06523df9cd88e9821b63545b58c27f16
--- /dev/null
+++ b/N9AyT4oBgHgl3EQftPlh/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:983025cfd73871aa8c8eb8ff82edf4cfddd0764dbf1445c4c63eab1129b366e9
+size 58764
diff --git a/PtAzT4oBgHgl3EQfIvs5/content/2301.01066v1.pdf b/PtAzT4oBgHgl3EQfIvs5/content/2301.01066v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..a957d181fe52665fca6e23b094052b9322293643
--- /dev/null
+++ b/PtAzT4oBgHgl3EQfIvs5/content/2301.01066v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:503eaea6451ffb88ca1e736c05cf98cdd87def5ae7f6d908677922c0b4ee07e9
+size 832437
diff --git a/PtAzT4oBgHgl3EQfIvs5/vector_store/index.pkl b/PtAzT4oBgHgl3EQfIvs5/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..e62f63c0598c9ae56b3e63126007454d3b3865f3
--- /dev/null
+++ b/PtAzT4oBgHgl3EQfIvs5/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6aabde1d5e69d60ffe7ad943b89f84d328baea913d68b9e2d28924a76d1be6c2
+size 144015
diff --git a/Q9E4T4oBgHgl3EQf_A6p/vector_store/index.faiss b/Q9E4T4oBgHgl3EQf_A6p/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..33e8e017f2cfdcc0f0a4957033bfea8acfc1e47f
--- /dev/null
+++ b/Q9E4T4oBgHgl3EQf_A6p/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e29951aa1cb89efd9691503ac4192f721f281016364cb73969b4bceb6d51c7a6
+size 2621485
diff --git a/QNE0T4oBgHgl3EQf1QJC/content/2301.02696v1.pdf b/QNE0T4oBgHgl3EQf1QJC/content/2301.02696v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..7e31f8068abf8bcde11243687ad273e70d6ee77f
--- /dev/null
+++ b/QNE0T4oBgHgl3EQf1QJC/content/2301.02696v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e663e4f1bf0b13d217a2c40d6b88b14a72d4abad237869de602da90dcfdb979
+size 10310926
diff --git a/QNE0T4oBgHgl3EQf1QJC/vector_store/index.faiss b/QNE0T4oBgHgl3EQf1QJC/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..8a3c6b775dd21751e0f7b43410f2fb0be943e99d
--- /dev/null
+++ b/QNE0T4oBgHgl3EQf1QJC/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0ad5acc98338d93de5b0897015944547f8f6911137fc4a6eec904d75c2ff0f8
+size 4915245
diff --git a/QdFJT4oBgHgl3EQf2y0y/vector_store/index.faiss b/QdFJT4oBgHgl3EQf2y0y/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..e8a891ac170f7a5d9fb45606dc4c1ac45010e5ff
--- /dev/null
+++ b/QdFJT4oBgHgl3EQf2y0y/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7c333cd7b2dd314b9d4aaf2ff21f62705ee1d9c290f4146a3c07b502d795b2a
+size 1966125
diff --git a/QdFJT4oBgHgl3EQf2y0y/vector_store/index.pkl b/QdFJT4oBgHgl3EQf2y0y/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..a2f43cae063b691027fa951cea35ca01a802e89f
--- /dev/null
+++ b/QdFJT4oBgHgl3EQf2y0y/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:818d40f627a8333807e4bb741df0af998056f6419721f5bcd4bd8795747f92dc
+size 82012
diff --git a/RdAyT4oBgHgl3EQft_l_/content/2301.00605v1.pdf b/RdAyT4oBgHgl3EQft_l_/content/2301.00605v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..7bff19d894cbecf9966ec0301a3b366a6ddbf256
--- /dev/null
+++ b/RdAyT4oBgHgl3EQft_l_/content/2301.00605v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbe35ff70a653b9055f5834e63a1798774f129e6e76a7b4a07d3a84ed342112b
+size 272718
diff --git a/RdAyT4oBgHgl3EQft_l_/vector_store/index.faiss b/RdAyT4oBgHgl3EQft_l_/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..cc146597423c1f51d56bc9e776bba7619c172175
--- /dev/null
+++ b/RdAyT4oBgHgl3EQft_l_/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69e7811c3aff5aabf81bea3f200c4df8e4f2240be177f26f0960a32ec6791579
+size 4259885
diff --git a/RtAzT4oBgHgl3EQf0P7C/content/2301.01782v1.pdf b/RtAzT4oBgHgl3EQf0P7C/content/2301.01782v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..d1ce1dfc52941724fe3ce13f4dc0836d841e759e
--- /dev/null
+++ b/RtAzT4oBgHgl3EQf0P7C/content/2301.01782v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63216e22ab03812c58a8772464a0001eb7a8c47961c4c79089a5c4b0298f5701
+size 1299109
diff --git a/RtAzT4oBgHgl3EQf0P7C/vector_store/index.pkl b/RtAzT4oBgHgl3EQf0P7C/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..6711d39c04d931b93fae33cee29de7bd447e6870
--- /dev/null
+++ b/RtAzT4oBgHgl3EQf0P7C/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25796bb49e9d21431c3ac3d520e7e403fc4f5bb8d8717836e33508eff5ce42c4
+size 180753
diff --git a/StE0T4oBgHgl3EQflAEY/content/tmp_files/2301.02479v1.pdf.txt b/StE0T4oBgHgl3EQflAEY/content/tmp_files/2301.02479v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d41103e1d2aa351dcd5faaccc95b275cf54f8231
--- /dev/null
+++ b/StE0T4oBgHgl3EQflAEY/content/tmp_files/2301.02479v1.pdf.txt
@@ -0,0 +1,2079 @@
+Quantum Multiple Access Wiretap Channel: On the
+One-Shot Achievable Secrecy Rate Regions
+Hadi Aghaee
+Faculty of Electrical Engineering
+K. N. Toosi University of Technology
+Tehran, Iran
+Email: Aghaee_Hadi@email.kntu.ac.ir
+
+Bahareh Akhbari
+Faculty of Electrical Engineering
+K. N. Toosi University of Technology
+Tehran, Iran
+Email: akhbari@eetd.kntu.ac.ir
+
+Abstract— In this paper, we want to investigate classical-
+quantum multiple access wiretap channels (CQ-MA-WTC) under
+one-shot setting. In this regard, we analyze the CQ-MA-WTC
+using simultaneous position-based decoder for reliable decoding
+and using a newly introduced technique in order to decode
+securely. Also, for the sake of comparison, we analyze the CQ-MA-
+WTC using Sen’s one-shot joint typicality lemma for reliable
+decoding. The simultaneous position-based decoder tends to a
+multiple hypothesis testing problem. Also, using convex splitting
+to analyze the privacy criteria in a simultaneous scenario becomes
+problematic. To overcome both problems, we first introduce a new
+channel that can be considered as a dual to the CQ-MA-WTC.
+This channel is called a point-to-point quantum wiretap channel
+with multiple messages (PP-QWTC). In the following, as a strategy
+to solve the problem, we also investigate and analyze quantum
+broadcast channels (QBCs) under the one-shot setting.
+Keywords—Quantum Channel; Mutual Information; Secrecy
+Capacity; Multiple Access Channel
+I. INTRODUCTION
+The quantum multiple access channel (QMAC) was first
+introduced by Winter [1]. A QMAC can accept two or more
+messages (classical or quantum) as inputs and one output.
+Similar to the classical world, decoding messages over a
+QMAC is based on two main techniques: successive
+cancelation decoding and simultaneous decoding. In [1], the
+author employs the successive cancelation decoding technique.
+A quantum broadcast channel (QBC) is a channel with a
+sender and two or more receivers. The sender wishes to transmit
+two or more messages (classical or quantum) over the channel
+to the receivers. The QBC was first introduced by Yard et al.
+[2]. In [2], the authors derived an inner bound for QBC for i.i.d.
+(independent and identical) case, and in [3], the authors derived
+the same inner bound using a more straightforward method and
+more in the spirit of its classical analogous [4] than the method
+in [2].
+In recent decades, with development of quantum data
+processing and its applications, the necessity to study the
+security of quantum channels has increased. In this regard, the
+quantum wiretap channel (QWTC) was first introduced in [5]
+and [6].
+Then, the secrecy constraints are extended to multi-user
+quantum channels such as quantum interference channel (QIC)
+[7,8], and quantum multiple access channel (QMAC) [9-13].
+There are two bottlenecks in studying the security of
+quantum channels. The first is decoding three or more messages
+simultaneously (reliability), and the second is about how we can
+securely decode two or more messages (confidentiality). The
+first bottleneck arises from the nonexistence of a general
+quantum joint typicality lemma. However, this problem has
+been solved in some cases, such as the min-entropy case and
+QMACs with commutative output [14]. Therefore, in the
+independent and identical distributed (i.i.d.) case, successive
+decoding combined with time-sharing techniques should be
+used. In this setting, transmitters are allowed to transmit their
+messages by only one use channel. Sen proved a joint typicality
+lemma which helps to decode any number of messages
+simultaneously in the one-shot case [14]. Obtaining secrecy
+against the eavesdropper by Wyner's technique [15] of
+randomizing over a block becomes problematic in the quantum
+setting. Wyner's technique has been shown to work for point-
+to-point quantum channels by Devetak [6] and explained
+further in [16]. However, there are no easy generalizations to
+multiple senders for a quantum channel. This issue is discussed
+in detail in [16].
+In this paper, we want to investigate the secrecy problem of
+quantum multiple access channel (QMAC) with classical inputs
+under one-shot setting. Also, we have investigated some
+bottlenecks connected to decoding process for CQ-MA-WTC.
+The achievement of this paper is about analyzing bottlenecks in
+decoding process and providing solutions to overcome them.
+Also, we present two techniques for quantum multiple
+access wiretap channel with classical inputs (CQMA-WTC).
+The first approach is based on the method presented in [14], and
+another technique is the simultaneous position-based decoder.
+From [17], we know that the simultaneous position-based
+decoder tends to a multiple quantum hypothesis testing problem
+which is solvable in a special case. Also, from [18], we know
+that the convex split lemma could not be used to analyze the
+privacy of multiple messages in simultaneous decoding.
+The paper is organized as follows: In Section II, some
+seminal definitions are presented. In Section III, the main
+channel and information processing tasks are presented. In
+Section IV, the results and main theorems are presented.
+Section V is dedicated to discussion.
+II. PRELIMINARIES
+Let A (Alice), B (Bob), and C (Charlie) be three quantum
+systems. These quantum systems can be denoted by their
+corresponding Hilbert spaces as ℋ�, ℋ�, and ℋ�. The states
+of the above quantum systems are presented as density
+operators ��, ��, and ��, respectively, while the shared state
+between Alice, Bob, and Charlie is denoted by ����. A density
+
+operator is a positive semidefinite operator with a unit trace.
+Alice, Bob, or Charlie’s state can be defined by a partial trace
+operator over the shared state. The partial trace is used to model
+the lack of access to a quantum system. Thus, Alice’s density
+operator using partial trace is �� = ����{����}. |�⟩� denotes
+the pure state of system A. The corresponding density operator
+is �� = |�⟩⟨�|�. The von Neumann entropy of the state �� is
+defined by �(�)� = −��{�� log ��}. For an arbitrarily state
+such as ���, the quantum conditional entropy is defined by
+�(�|�)� = �(�, �)� − �(�)�.
+The
+quantum
+mutual
+information is defined by �(�; �)� = �(�)� + �(�)� −
+�(�, �)�, and the conditional quantum mutual information is
+defined by:
+�(�; �|�)� = �(�|�)� + �(�|�)� − �(�, �|�)�
+Quantum operations can be denoted by completely positive
+trace-preserving (CPTP) maps ��→�. The CPTP maps accept
+input states in A and output states in B. The distance between
+two quantum states, such as A and B, is defined by trace
+distance. The trace distance between two arbitrary states, such
+as � and � is:
+‖� − �‖� = ��|� − �|
+(1)
+where |Ψ| = √Ψ�Ψ. This quantity is zero for two similar and
+perfectly distinguishable states.
+Fidelity is defined as �(�, �) = ���√���
+�, and purified
+distance is a metric on �(ℋ) and is defined as �(�, �) ≔
+�1 − �(�, �)�. Most of the above definitions are given in [19].
+Definition 1: (Hypothesis testing mutual information)
+Hypothesis testing mutual information is denoted by ��
+�(�; �)
+∶= ��
+� (���‖�� ⊗ ��), � ∈ (0,1) and is considered as quantum
+hypothesis testing divergence [17] where ��
+� (. ‖.) is hypothesis
+testing relative entropy [17]. � is the smoothing variable,
+�ℋ�ℋ� is the joint classical-quantum state of input and output
+over their Hilbert spaces (ℋ�, ℋ�), and it can be shown as ���:
+��� = � ��(�)|�⟩⟨�|� ⊗ ��
+�
+�
+
+where �� is the input distribution.
+Definition 2: (Quantum relative entropy [20]): Consider
+states ��, �� ∈ �(ℋ�). The Quantum relative entropy is
+defined as:
+�(��‖��)
+≔ ���{���log� �� − log� ���}
+����(��) ⊆ ����(��)
++∞
+��ℎ������
+
+where ����(��) refers to the set-theoretic support of �.
+����(�) is the subspace of ℋ spanned by all eigenvectors of �
+with non-zero eigenvalues.
+Fact 1: The following relation exists between the quantum
+relative entropy and hypothesis testing relative entropy for � ∈
+(0,1) [21]:
+��
+�(��‖��) ≤
+1
+1 − � ��(��‖��) + ℎ�(�)�
+where ℎ�(�) ≔ −� log� � − (1 − �) log�(1 − �) is a binary
+entropy function.
+Definition 3: (Max mutual information [21]) Consider a
+bipartite state ��� and a parameter � ∈ (0,1). The max mutual
+information can be defined as follows:
+����(�; �)� ≔ ����(��� ‖��⨂�� )�
+where � refers to the state ��� and ����(∣∣) is the max-relative
+entropy [22] for ��, �� ∈ ℋ�:
+����(�� ‖��) ≔ inf{� ∈ ℝ: �� ≤ 2���}
+Definition 4: (Quantum smooth max relative entropy [22])
+Consider states ��, �� ∈ �(ℋ�), and � ∈ (0,1). The quantum
+smooth max relative entropy is defined as:
+����
+�
+(��‖��) ∶=
+inf
+��
+� ∈ℬ�(��) ����(��
+� ‖�� )
+where ℬ�(��) ≔ {��
+� ∈ �(ℋ�): �(��
+� , ��) ≤ �} is �-ball for
+���.
+Definition 5: (Quantum smooth max mutual information
+[21]) Consider ��� ∶= ∑
+��(�)|�⟩⟨�|� ⊗
+�∈�
+��
+� as a classical-
+quantum state and a parameter � ∈ (0,1). The smooth max
+mutual information between the systems � and � can be defined
+as follows:
+����
+�
+(�; �) ∶=
+inf
+���
+�
+∈ℬ�(���) ����(���
+� ‖��⨂�� )
+=
+inf
+���
+�
+∈ℬ�(���)����(�; �)�� ,
+where ℬ�(���) ≔ {���
+�
+∈ �(ℋ� ⊗ ℋ�): �(���
+� , ���) ≤ �} is
+�-ball for ���.
+Definition 6: (Conditional smooth hypothesis testing
+mutual
+information
+[23])
+Consider
+����
+∶= ∑
+��(�)|�⟩⟨�|�
+�∈�
+⊗ ���
+� be a tripartite classical-quantum
+state and � ∈ (0,1). We define,
+��
+�(�; �|�)� ≔ max
+��
+min
+�∈�������
+� � ��
+�(�; �)���
+� ,
+where maximization is over all ��
+� = ∑
+��(�)|�⟩⟨�|�
+�∈�
+
+satisfying �(��
+� , ��) ≤ �.
+Fact 2: [24] Let ���� ∶= ∑
+��(�)|�⟩⟨�|�
+�∈�
+⊗ ���
+� be a
+tripartite classical-quantum state and � ∈ (0,1), the following
+relation holds,
+lim
+�→�
+1
+� ��
+�(�⨂�; �⨂�|��)�⨂� = �(�; �|�)�
+Definition 7: (Alternate smooth max-mutual information)
+Consider a bipartite state ��� and a parameter � ∈ (0,1). The
+alternate definition of the smooth max-mutual information
+between the systems � and � can be defined as follows:
+�����
+�
+(�; �) ∶=
+inf
+���
+�
+∈ℬ�(���) ����(���
+� ‖�� ⨂ ��
+� )
+Fact 3: (Relation between two definitions of the smooth
+max mutual information) [25]: Let � ∈ (0,1) and � ∈ (0, �)
+For a bipartite state ���, it holds that:
+
+�����
+�
+(�; �)� ≤ ����
+��� (�; �)� + log 3
+��
+Definition 8: (Conditional smooth max mutual information
+[23])
+Consider
+���� ∶= ∑
+��(�)|�⟩⟨�|�
+�∈�
+⊗ ���
+�
+be
+a
+tripartite classical-quantum state and � ∈ (0,1). We define,
+����
+�
+(�; �|�)� ≔ max
+��
+min
+�∈�������
+� � ����
+�
+(�; �)���
+� ,
+where maximization is over all ��
+� = ∑
+��(�)|�⟩⟨�|�
+�∈�
+
+satisfying �(��
+� , ��) ≤ �.
+Fact 4: [24] ���� ∶= ∑
+��(�)|�⟩⟨�|�
+�∈�
+⊗ ���
+� be a
+tripartite classical-quantum state and � ∈ (0,1), the following
+relation holds,
+lim
+�→�
+1
+� ����
+�
+(�⨂�; �⨂�|��)�⨂� = �(�; �|�)�
+Definition 9: (Quantum Rényi relative entropy of order �
+[17]) For a state � ∈ �(ℋ) and a positive semidefinite operator
+�, the quantum Rényi relative entropy of order �, where � ∈
+�0,1) ∪ (1, +∞) is defined as:
+��(�‖�) ≡
+1
+� − 1 log� ��{������}
+Also, Rényi entropy of order � can be defined as follows:
+��(�)� ≡
+1
+1 − � log� ��{��
+�}
+Definition 10: (One-shot inner bound of a classical-
+quantum multiple access channel) [14] A two-user classical-
+quantum multiple access channel (C-QMAC) under the one-
+shot
+setting
+is
+a
+triple
+(�� × ��, �����→�(��, ��) ≡
+�����
+�
+, ℋ�), where �� and �� are the alphabet sets of two
+classical inputs, and � is the output system. �����
+�
+ is a quantum
+state, and the channel has a completely positive trace-
+preserving map (CPTP) �����→�.
+Considering the joint typicality lemma introduced in
+[Corollary 4, 14], the one-shot inner bound of a C-QMAC is as
+follows:
+�� ≤ ��
+�(��: ���)� − 2 + log �
+�� ≤ ��
+�(��: ���)� − 2 + log �
+�� + �� ≤ ��
+�(����: �)� − 2 + log �
+with decoding error at most 49√�, where ��
+�(. ) is the hypothesis
+testing mutual information defined in Definition 1 with respect
+to the controlling state:
+������� ∶=
+� �(�)�(��|�)�(��|�)|�����⟩⟨�����|�����
+�����
+⊗ �����
+�
+
+(2)
+and � is a time-sharing variable.
+Note that ��
+�(: ) is the difference between a Rényi entropy
+of order two and a conditional quantum entropy.
+Lemma 1: [16] Given the control state in (2) (without time-
+sharing variable), �� > 0 and 0 < �� < ��,let ���, … , ���� and
+���, … , ���� be i.i.d. samples from the distributions �� and ��.
+Then, if
+log|��| ≥ ����
+�����(�: �)� + log 3
+��� − 1
+4 log ��
+log|��| ≥ ����
+�����(�: ��)� + log 3
+��� − 1
+4 log �� + �(1)
+the following holds,
+���,…,���~��
+��,…,���~��
+�
+1
+|��||��| � � �����
+�
+− ��
+|��|
+���
+|��|
+���
+�
+�
+≤ 20���
+�
+Proof: see [16].
+Lemma 2: : (Convex split lemma) [19,20] Let ��� be an
+arbitrary state and suppose that ���…��� be the following state:
+���…��� = 1
+� � ��� ⊗ … ⊗ ����� ⊗ ���� ⊗ ����� ⊗ …
+�
+���
+⊗ ���
+Let � ∈ (0,1) and � ∈ �0, √��, if
+log� � = �����
+√���(�; �)� + 2 log� �1
+��
+then,
+�����…���, ��� ⊗ … ⊗ ��� ⊗ ���� ≤ √�
+for some state ��� such that �(��, ���) ≤ √� − �.
+Proof: see [20].
+Lemma 3: (Hayashi-Nagaoka inequality [26]) Suppose that
+�,� ∈ �(ℋ�) such that (� − �) ∈ �(ℋ�) are operators such
+that � ≥ 0 and 0 ≤ � ≤ �, then for all positive constant �, the
+following relation holds:
+� − (� + �)��
+� � (� + �)��
+�
+≤ (1 + �)(� − �) + (2 + � + ���)�
+Proof: see [26].
+III. CHANNEL MODEL
+A
+two-user
+CQ-MA-WTC
+is
+a
+triple
+(�� ×
+��, �����→��(��, ��) ≡ �����
+�� , ℋ� ⊗ ℋ�), where ��, � ∈
+{1,2} denote the input alphabet sets, and �, � denote the output
+systems (� denotes the channel output at the legitimate
+receiver (Charlie), and � is the channel output at the
+eavesdropper). �����
+�� is the system output’s quantum state.
+Both users want to transmit their messages as secure as
+possible over a CQ-MA-WTC to the receiver.
+
+
+Figure 1. The CQ-MA-WTC model
+The main channel is illustrated in Figure 1.
+Consider the main channel illustrated in Figure 1. Each
+user chooses its message ��; � ∈ {1,2} from its message set
+ℳ� = �1: |ℳ�| = 2���; � ∈ {1,2}
+(��
+and
+��
+are
+the
+transmitting rates corresponding to the first and the second
+messages, respectively), and sends it over a CQ-MA-WTC.
+The users also use two junk variables ��; � ∈ {1,2} from two
+amplification
+sets
+�� = �1: |��| = 2����; � ∈ {1,2}
+for
+randomizing Eve’s knowledge. We have two doubly indexed
+codebooks ��(��, ��), and ��(��, ��), for user-1 and user-2,
+respectively.
+IV. MAIN RESULTS
+In this section, we present the main results.
+Corollary 1 gives a one-shot achievable secrecy rate region
+for sending classical messages over a CQ-MA-WTC based on
+Sen’s quantum joint typicality lemma [14]. The second theorem
+presents a novel approach to decode both messages over a CQ-
+MA-WTC reliably and confidentially (simultaneous position-
+based decoder). It should be noted that Corollary 1 and
+Theorem 1 use the same method to prove the security
+requirements. Also, we present a theorem that tries to overcome
+the bottlenecks connected to Theorem 1.
+Corollary 1: (One-shot achievable rate region for CQ-MA-
+WTC) Consider a two-user CQ-MA-WTC which accepts �� and
+�� as inputs and �, and � as outputs. �����
+�� is the channel
+density operator. For any fixed � ∈ (0,1), �� ∈ (0, ��) and �, ��
+such that �� > 0, the rate pair ���, ��, 49√� + 20��
+�
+�� is
+achievable to satisfy the following inequalities:
+�� ≤ ��
+�(��: ���|�)� − ����
+�
+(��: �|�)� + log � − 2 − log 3
+���
++ 1
+4 log ��
+�� ≤ ��
+�(��: ���|�)� − ����
+�
+(��: ���|�)� + log � − 2
+− log 3
+��� + 1
+4 log �� + �(1)
+�� + �� ≤ ��
+�(����: �|�)� − ����
+�
+(��: �|�)�
+− ����
+�
+(��: ���|�)� + log � − 2 − 2 log 3
+���
++ 1
+2 log �� + �(1)
+where � = �� − �� and the union is taken over input
+distribution ��(�)���|�(��|�)���|�(��|�). Q is the time-
+sharing random variable, and all of the mutual information
+quantities are taken with respect to the following state:
+
+�������� ≡
+� ���(�)���|�(��|�)���|�(��|�)|�⟩⟨�|�
+�,��,��
+⊗ |��⟩⟨��|�� ⊗ |��⟩⟨��|��
+⊗ �����
+��
+
+
+
+
+
+(3)
+Proof: See Appendix A.
+Sketch of proof: The proof has two steps: 1- Reliable
+decoding based on Sen’s quantum one-shot joint typicality
+(Definition 9). 2- Secure decoding based on Lemma 1.
+Theorem 1: (one-shot lower bound for CQ-MA-WTC) For
+any fixed � ∈ (0,1), �� ∈ (0,1) and �, �� such that � ∈ (0, �),
+and �� ∈ (0, ��), there exists a one-shot code for the channel
+�����→��, if rate pair ���, ��, � + 2� + 20��
+�
+�� satisfies the
+following bounds:
+�� ≤ ��
+�(��: ���|�)� − ����
+�
+(��: �|�)� − log� �4�
+���
+− log 3
+��� + 1
+4 log ��
+�� ≤ ��
+�(��: ���|�)� − ����
+�
+(��: ���|�)� − log� �4�
+���
+− log 3
+��� + 1
+4 log �� + �(1)
+�� + �� ≤ ��
+�(����: �|�)� − ����
+�
+(��: �|�)�
+− ����
+�
+(��: ���|�)� − log� �4�
+���
+− 2 log 3
+��� + 1
+2 log �� + �(1)
+where � = �� − �� and the union is taken over input
+distribution ��(�)���|�(��|�)���|�(��|�). Q is the time-
+sharing random variable, and all mutual information quantities
+are taken with respect to the state (3).
+Proof: See Appendix B.
+Sketch of proof: The proof has two steps: 1- Reliable
+decoding based on the simultaneous position-based technique:
+for simplicity of analysis, we merge reliability and
+confidentiality criteria into a single criterion [20]. 2- Secure
+decoding based on the Lemma 1.
+Remark 1: It should be noted that, both of the above
+theorems tend to the same result if and only if � = �.
+As mentioned before, the simultaneous position-based
+decoder tends to a multiple hypothesis testing problem which is
+unsolvable in the general case. Also, the convex split lemma
+(Lemma 2) does not make sense in the simultaneous decoding.
+Because it runs to the famous smoothing bottleneck of quantum
+information theory.
+Now, consider the channel illustrated in Figure 2.
+
+α2.YZ
+Z.
+wiretapper
+(M', M2)
+Px1x2
+m-+→(Mi, M2)m
+V
+Figure 2. The PP-QWTC model.
+This channel accepts two or more messages from one user.
+We call this channel a point-to-point quantum wiretap channel
+with multiple messages (PP-QWTC).
+Consider PP-QWTC with classical messages. This channel
+is studied in [27] under a different scenario wherein a sender
+wants to send classical and quantum messages simultaneously
+to a legitimate receiver.
+Information processing task: Two classical messages
+(��, ��) ∈ ℳ� × ℳ� are possessed by a sender (Alice) and
+be transmitted to a receiver (Bob) in the presence of a passive
+wiretapper over a point-to-point quantum channel under the
+one-shot scenario. Both of the messages, should be kept as
+secure as possible from the wiretapper. The PP-QWTC is a
+triple (�, ��→��(��, ��) ≡ ��(��,��)
+��
+, ℋ� ⊗ ℋ�), where �
+denotes the input alphabet sets, and �, � denote the output
+systems (� denotes the channel output at the legitimate
+receiver (Bob), and � is the channel output at the
+eavesdropper). ��(��,��)
+��
+≡ �����
+�� is the system output’s
+quantum state.
+Alice chooses its message ��; � ∈ {1,2} from its message
+set ℳ� = �1: |ℳ�| = 2���; � ∈ {1,2}, and sends it over a PP-
+QWTC. Alice also uses two junk variables ��; � ∈ {1,2} from
+two amplification sets �� = �1: |��| = 2����; � ∈ {1,2} for
+randomizing Eve’s knowledge. We have two doubly indexed
+codebooks ��(��, ��), and ��(��, ��).
+Encoding: An encoding operation by Alice ℰ: ���� →
+�(ℋ�)
+∀��, �� ∈ ��, ��
+1
+2 ������� − ����� ⊗ ����� ≤ ��
+(4)
+where for each message ��, ��, ������ and ����� are
+appropriate
+marginal
+of
+the
+state
+������� =
+�
+|ℳ�||ℳ�| ∑
+∑
+|��⟩⟨��| ⊗ |��⟩⟨��| ⊗
+|ℳ�|
+����
+|ℳ�|
+����
+��ℰ(��, ��)�. Also, ��� can be any arbitrary state.
+Decoding: Decoding operation by Bob �: �(ℋ�) → ������
+such that:
+�� �����, ���� ≠ (��, ��)� ≤ �1
+(5)
+A rate pair (��, ��) is (�1, �2)-achievable if, for such encoding
+and decoding maps (ℰ, �), the conditions stated in (4) and (5)
+are satisfied.
+As it can be understood from criterion (4), the reliability and
+confidentiality conditions are merged into a single criterion.
+This idea is used in [28] and [20] for the first time.
+
+Figure 3. The QBC model.
+Theorem 2: (An inner bound on the one-shot capacity
+region of PP-QWTC) For any fixed �� ∈ (0,1), �� ∈ (0,1) and
+��, �� such that �� ∈ (0, ��) and �� ∈ (0, ��), there exists a one-
+shot code for the channel ��→��, if rate pair (��, ��, 3�� +
+2√�� + 2√��, 2(�� + √��) + √��) satisfies the following
+bounds:
+�� ≤ ��
+�����(��; �|��)� − �����
+√�����(��; �)� − log 4��
+��
+�
+− 2 log 1
+��
+
+�� ≤ ��
+�����(��; �|��)� − �����
+√�����(��; �|��)� − log 4��
+��
+�
+− 2 log 1
+��
+
+with
+respect
+to
+state
+������� =
+∑
+∑
+�(��, ��)|��⟩⟨��| ⊗ |��⟩⟨��| ⊗ ���
+����
+|��|
+����
+|��|
+����
+.
+Proof: In Appendix C.
+Remark 2: The proof of Theorem 2 has two advantages over
+the proof of Theorem 1: The first is that the proof of Theorem
+2 is based on solving a binary hypothesis testing problem
+against the proof of Theorem 1, which is based on solving a
+multiple hypothesis testing problem. The second is that in the
+privacy proof of Theorem 1, Lemma 1 [16] is used. But, in the
+proof of Theorem 2 the convex split lemma (Lemma 2) can be
+used.
+Remark 3: From a comparison between the results of Theorem
+1 and Theorem 2, it can be understood that the proof of
+Theorem 3 does not give the sum-rate (�� + ��) . This is
+because of using the successive decoding technique. This issue
+should not cause doubts about whether PP-QWTC is a dual for
+CQ-MA-WTC or not. To solve this doubt, we propose the issue
+of quantum broadcast channels.
+Quantum broadcast channels
+The quantum broadcast channel (QBC) accepts one user and
+two or more receivers. In the basic case, the sender (Alice)
+wishes to transmit three separate messages: �� is the personal
+message for the first receiver ��, �� is the personal message for
+the second receiver ��, and �� is the common message for
+both of the receivers.
+The basic QBC is illustrated in Figure 3. It should be noted
+that, for ease of calculation, we removed the security constraint
+from the problem.
+
+
+→(Mc)X(m1,mc )
+(M1,Mc
+UXY Y2OBCYZ
+Z.
+wiretapper
+-(M", M*)
+Puru21
+di
+4-J
+→(Mi, M2)
+XV���������� ≡ � ���(��)|��⟩⟨��|�� ⊗ |��⟩⟨��|��� ⊗ |��⟩⟨��|����
+��
+
+(6)
+���������� ≡ � ���(��)|��⟩⟨��|�� ⊗ |��⟩⟨��|��� ⊗ |��⟩⟨��|����
+��
+
+(7)
+��������� ≡ � ���(��)|��⟩⟨��|�� ⊗ ���
+���� ⊗ |��⟩⟨��|����
+��
+
+(8)
+��������� ≡ � ���(��)|��⟩⟨��|�� ⊗ ���
+���� ⊗ |��⟩⟨��|����
+��
+
+(9)
+������� ≡ �������→������������ ⊗ ���������� = � ���(��)���(��)
+����
+|��⟩⟨��|�� ⊗ |��⟩⟨��|�� ⊗ ���
+����
+(10)
+1
+|ℳ�||ℳ�| � � 1
+2
+|ℳ�|
+����
+|ℳ�|
+����
+���→������ ��
+��������
+⊗ |ℳ�||��|��������
+⊗ |ℳ�||��|��
+(��,��),(��,��)
+− �|��⟩⟨��|��� ⊗ |��⟩⟨��|����
+⊗ ��
+����⊗|ℳ�||��|����⊗|ℳ�||��|���
+�
+≤ � + 2� + 20���
+�
+where ��
+����⊗|ℳ�||��|����⊗|ℳ�||��|� ≔ �
+����⊗|ℳ�||��|����⊗|ℳ�||��|
+(��,��),(��,��)
+⊗ ���.
+
+
+
+(11)
+��→������ ����⊗|ℳ�||��|��⊗|ℳ�||��|��
+(��,��),(��,��)
+� ≔ �
+� ����(��)����(��)|��⟩⟨��|��� ⊗ |��⟩⟨��|���
+|ℳ�|
+����
+|ℳ�|
+����
+= �
+� �� �Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+����
+���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+� |��⟩⟨��|��� ⊗ |��⟩⟨��|���
+|ℳ�|
+����
+|ℳ�|
+����
+
+
+
+
+(12)
+ The problem of QBC is widely studied in the i.i.d. case in
+[2-3] and in the one-shot case in [29]. In the following, we want
+to achieve a one-shot inner bound for QBC with classical
+messages. Suppose that Alice has not a personal message for
+the second receiver �� (�� = ∅ → �� = 0).
+The QBC under the one-shot setting is a triple
+( �, ��→���� ≡ ��
+����, ℋ�� ⊗ ℋ��) , where � denotes the
+input alphabet set, and ��, �� denote the output systems. ��
+����
+is the system output’s quantum state.
+Theorem 3: (one-shot inner bound for QBC) Let � be an
+auxiliary random variable, � = ��|�(�|�)��(�) be the code
+probability function. The one-shot achievable rate consists of
+all rate pairs (��, ��) such that:
+�� ≤ ��
+�(�; ��|�)� − 2 + log �
+�� ≤ ��
+�(�; ��)� − 2 + log �
+�� + �� ≤ ��
+�(�; ��)� − 2 + log �
+is achievable, and all information quantities are taken with
+respect to the following state:
+������� = � ��(�)��|�(�|�)
+�,�
+|�⟩⟨�|� ⊗ |�⟩⟨�|� ⊗ ��
+����
+(13)
+Proof: In Appendix D.
+Now, consider the extended version of the above theorem:
+Corollary 2: (one-shot inner bound for QBC with three
+personal messages for the first receiver) Let � be an auxiliary
+random variable, � = ��(�)���|�(��|�)���|���(��|���) be
+the code probability function. The one-shot achievable rate
+region consists of all rate tuples (��, ��, ��) in order to
+sending (��, ��, ��) such that:
+�� ≤ ��
+�(��; ��|�)� − 2 + log �
+�� ≤ ��
+�(��; ��|���)� − 2 + log �
+�� ≤ ��
+�(�; ��)� − 2 + log �
+�� + �� ≤ ��
+�(����; ��|�)� − 2 + log �
+�� + �� ≤ ��
+�(��; ��)� − 2 + log �
+�� + �� ≤ ��
+�(��; ��|��)� − 2 + log �
+is achievable, and all information quantities are taken with
+respect to the following state:
+���������� = � ��(�)���|�(��|�)
+�,�
+���|���(��|���)|�⟩⟨�|�
+⊗ |��⟩⟨��|�� ⊗ |��⟩⟨��|�� ⊗ �����
+����
+
+Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+(��,��),(��,��)
+≔ �� � � � Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,���,��� ,���
+���
+���
+���
+���
+�
+��
+�
+Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+�� � � � Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,���,��� ,���
+���
+���
+���
+���
+�
+��
+�
+
+
+
+
+(14)
+Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+≔ �����
+(�,�),(�,�) ⊗ … ⊗ �����
+(�,�),(�,��) ⊗ … ⊗ �����
+(�,�),(��,����) ⊗ … ⊗ �����
+(�,��),(��,��) ⊗ … ⊗ �����
+(��,����),(��,��)
+⊗ ������
+(��,��),(��,��) ⊗ �����
+(��,��),(��,����) ⊗ … ⊗ �����
+(|ℳ�|,|��|),(|ℳ�|,|��|)
+
+
+
+(15)
+�� ��� − Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+� = �� ��� − ��������������→��������� ⊗ ��������
+(16)
+�� �����, ���� ≠ (��, ��)�
+≤ (1 + �)�� ��� − Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
++ (2 + � + ���)
+�
+�
+�
+� �� �Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,���,��� ,���
+���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+������
+������
+��� ���
+��� ���
+= (1 + �)�� ��� − Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
++ (2 + � + ���)
+�
+� �� �Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,���,��,��
+���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+������
+��� ���
++ (2 + � + ���)
+�
+� �� �Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��� ,���
+���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+������
+��� ���
++ (2 + � + ���)
+�
+�
+�
+� �� �Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,���,��� ,���
+���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+������
+������
+��� ���
+��� ���
+= (1 + �)�� ��� − ��������������→��������� ⊗ ��������
++ (2 + � + ���)(|ℳ�||��| − 1)�� ��� − ��������������→������ ⊗ ���� ⊗ ��������
++ (2 + � + ���)(|ℳ�||��| − 1)�� ��� − ��������������→��������� ⊗ ��� ⊗ ������
++ (2 + � + ���)(|ℳ�||��| − 1)(|ℳ�||��|
+− 1)�� ��� − ��������������→������ ⊗ ���� ⊗ ��� ⊗ ������
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+(17)
+Proof: The proof follows the extended version of Theorem 3’s
+proof.
+The channel described in Corollary 2 will be converted to
+the channel described in Theorem 2 (PP-QWTC) without
+secrecy constraint by choosing �� = ∅. Set �� = 0 in
+Corollary 2:
+�� ≤ ��
+�(��; ��)� − 2 + log �
+�� ≤ ��
+�(��; ��|��)� − 2 + log �
+�� + �� ≤ ��
+�(����; ��)� − 2 + log �
+�� ≤ ��
+�(��; ��)� − 2 + log �
+�� ≤ ��
+�(��; ��|��)� − 2 + log �
+where the above last two rates are redundant. Then, we have the
+following region:
+�� ≤ ��
+�(��; ��)� − 2 + log �
+
+�� ≤ ��
+�(��; ��|��)� − 2 + log �
+�� + �� ≤ ��
+�(����; ��)� − 2 + log �
+(18)
+Consider the results of Theorem 2 without the leaked
+information terms. By choosing �� = ��, we have:
+�� ≤ ��
+�(��; ��|��)� − 2 + log ��
+�� ≤ ��
+�(��; ��|��)� − 2 + log ��
+(19)
+Comparing (18), (19), and (33), the argument stated in Remark
+3 is proved. Also, the region (18) is a near-optimal achievable
+rate region compared to Corollary 1. As it can be understood
+from a comparison between the results of Corollary 1, Theorem
+2, and Corollary 2 by considering (31), this idea can be proved
+that converting the CQ-MA-WTC to PP-QWTC can be a
+helpful approach to bypass the bottlenecks connected to the
+multiple hypothesis testing problem (Theorem 1) and the
+smoothing bottlenecks of quantum information theory
+(Corollary 1 and Theorem 1).
+
+Asymptotic analysis
+In this subsection, we want to evaluate our secrecy rate
+region in the asymptotic i.i.d. case (asymptotic limit of many
+uses of a memoryless channel). Consider PP-QWTC
+(�(��, ��) → ��
+��). The capacity region of the channel can be
+expressed as follows:
+��(�) ≔
+lim
+��,��→� lim
+�→�
+1
+� ���,��(�⊗�)
+(20)
+
+where ���,��(�⊗�) ≡
+max
+�(��,��) ℛ��,��(�⊗�).
+Let ℛ(�) be the set of the maximum rate pairs (��
+�, ��
+� ),
+ℛ(�) = � ��
+� ≤ �(��; �|��)� − �(��; �)�
+��
+� ≤ �(��; �|��)� − �(��; �|��)�
+
+(21)
+Then the capacity region ��(�) is the union over � uses of
+the channel �:
+��(�) ≔
+max
+�(��,��)
+1
+� � ℛ(�⊗�)
+�
+���
+
+(22)
+Our aim is to prove the expression above. Consider both of
+single rates. Applying Fact 2 (and its conditional version) we
+have,
+�� ≥ ��
+�����(��; �|��)� − ����
+√�������(��; �)� − log 4��
+��
+�
+− 2 log 1
+��
+− log 3
+��
+�� ≥ ��
+�����(��; �|��)� − ����
+√�������(��; �|��)� − log 4��
+��
+�
+− 2 log 1
+��
+− log 3
+��
+To prove the achievability, consider the one-shot lower bounds
+presented in Theorem 2, and apply quantum AEP [30] for the
+conditional smooth hypothesis testing, and max-mutual
+information. From Theorem 2, for � uses of the channel �, the
+following lower bound ���,��(�⊗�) can be obtained:
+� ℛ(�⊗�)
+�
+���
+⊆ ���,��(�⊗�)
+where ℛ(�⊗�) is the set of all rate pairs (��
+�, ��
+� ) satisfying:
+��
+� ≤ ��
+�����(��
+�; �⨂�|��
+�)� − ����
+√�������(��
+�; �⨂�)�
+− log 4��
+��
+� − 2 log 1
+��
+− log 3
+��
+
+(23)
+��
+� ≤ ��
+�����(��
+�; �⨂�|��
+�)�
+− ����
+√�������(��
+�; �⨂�|��
+�)�
+− log 4��
+��
+� − 2 log 1
+��
+− log 3
+��
+
+
+(24)
+We can assume that the sequences of the random variables are
+generated in an i.i.d. fashion according to their distributions.
+This is due to the fact that the region above is basically a lower
+bound on the capacity region. This empowers us to make use of
+quantum AEP as described below. From Fact 3, we have,
+
+lim
+��→� lim
+�→�
+1
+� ��
+�����(��
+�; �⨂�|��
+�)�⨂� = �(��; �|��)�
+(25)
+lim
+��→� lim
+�→�
+1
+� ��
+�����(��
+�; �⨂�|��
+�)�⨂� = �(��; �|��)�
+(26)
+Also, using Fact 4, we have the following:
+lim
+��→� lim
+�→�
+1
+� ����
+√�������(��
+�; �⨂�)�⨂� = �(��; �|��)�
+(27)
+lim
+��→� lim
+�→�
+1
+� ����
+√�������(��
+�; �⨂�|��
+�)�⨂�
+= �(��; �|��)�
+(28)
+Putting (25), (26), (27), and (28) into (23), and (24) gives (21):
+ℛ(�⊗�) ⊆
+lim
+��,��→� lim
+�→�
+1
+� ���,��(�⊗�)
+Given the argument above and using (20), and (22) completes
+the proof.
+V. DISCUSSION
+In this paper, we studied the problem of secure
+communication over a CQ-MA-WTC using three techniques:
+1- Sen’s joint typicality lemma. 2-simultaneous position-based
+decoding and, 3-successive position-based decoding. The first
+and the second decoding techniques use a newly introduced
+smooth technique [16] to analyze the privacy, while the third
+technique uses convex splitting [19]. We realized that the
+simultaneous position-based decoder tends to a multiple
+hypothesis testing problem which is unsolvable in the general
+case. We introduced a new channel (PP-QWTC) which can be
+considered as a dual for CQ-MA-WTC. Also, this channel can
+be derivate from the quantum broadcast channel. The results
+show that the PP-QWTC has a near-optimal achievable rate
+region to CQ-MA-WTC.
+APPENDIX
+Appendix A: (Proof of Corollary 1)
+As mentioned, the proof has two steps: Reliable decoding
+and secure decoding. To these ends, consider two junk variables
+��; � ∈ {1,2} for each of users ��, � ∈ {1,2}. These junk
+variables are used to make two doubly indexed codebooks
+{��(��, ��)}��∈ℳ�,��∈�� and {��(��, ��)}��∈ℳ�,��∈��. Bob
+should be able to detect the pair messages (��, ��), and the
+junk variables ��, and �� with high probability.
+Using Definition 10 (Sen’s inner bound for QMAC), we have
+the following relation:
+ℛ������������ = ℛ��� − ℛ������
+with decoding error at most 49√�, and privacy leakage at most
+20��
+�
+� (Lemma 1). Also, ℛ��� refers to Sen’s inner bound for
+QMAC (Definition 9), and ℛ������ refers to the leaked
+information from senders to Eve.
+From Lemma 1, we have the following:
+
+��������� ≤ ����
+�����(��: �)� + log 3
+��� − 1
+4 log ��
+��������� ≤ ����
+�����(��: ���)� + log 3
+��� − 1
+4 log �� + �(1)
+This completes the proof.
+Appendix B: (Proof of Theorem 1)
+Both of the messages are uniformly distributed on their sets.
+The receiver has to be able to decode both messages with
+negligible error probability. Before communication begins,
+Alice (A) and Bob (B) share randomness with Charlie (C), and
+wiretapper (Z). Let ���������� (6) and ���������� (7) be shared-
+randomness between (A,C,Z) and shared-randomness between
+(B,C,Z), respectively. Alice has ��
+� system, Bob has ��
+� system,
+and Charlie has (��, ��) system, and wiretapper has (��
+��, ��
+��)
+system. Let ��������� (8) and ��������� (9) denote the state
+resulting from sending ��
+� and ��
+� over the channel, respectively.
+Then the overall controlling state of the channel is as stated in
+(10).
+Sketch of the coding scheme: For each of the messages
+(��), � ∈ {1,2}, there exist local keys �� ∈ �1: |��|�, � ∈ {1,2}
+as uniform randomness for randomizing Eve’s knowledge
+about the sent messages. These local keys are not accessible to
+Charlie or Eve. Before the communication begins, assume that
+Alice, Charlie, and Eve share |ℳ�||��| copies of the state in
+(6) and Bob, Charlie and Eve share |ℳ�||��| copies of the state
+in (7):
+�
+��
+|ℳ�||��|���|ℳ�||��|����|ℳ�||��| = ����������
+⊗|ℳ�||��|
+�
+��
+|ℳ�||��|���|ℳ�||��|����|ℳ�||��| = ����������
+⊗|ℳ�||��|
+To send the pair messages ��, and ��, Alice and Bob pick
+�� ∈ �1: |��|� and �� ∈ �1: |��|�, respectively, and uniformly
+at random. They send (��, ��)-th system ��
+� and (��, ��)-th
+system ��
+� through the channel �������→��.
+There exists a simultaneous decoder for communication
+over a CQ-MA-WTC with the upper bound on the average error
+probability, as stated in (11). As it can be understood from (11),
+the security criterion is merged into the reliability criterion [20].
+The simultaneous position-based decoder can be constructed as
+stated in (12), where,
+Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+����
+= � � Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+(��,��),(��,��)
+|��|
+����
+|��|
+����
+
+Now, we consider the error term. Charlie constructs her
+position-based decoder to decode ��, ��, ��, and ��. Let
+Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+(��,��),(��,��)
+ denotes the POVM:
+�� �����
+|ℳ�||��|��|ℳ�||��|�
+− Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+(��,��),(��,��)
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+� ≤ �
+where Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+(��,��),(��,��)
+ is expressed in (14), and for
+�� ∈ �1: |ℳ�|� and �� ∈ �1: |��|�, Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+ is
+expressed in (15), in which ������
+(��,��,��,��) is a test operator used
+to discriminate between hypotheses ������, ����� ⊗ ��, ��� ⊗
+���� and ��� ⊗ ��� ⊗ �� with an error of �. Note that, this
+hypothesis testing problem is equal to discriminating between
+hypotheses �������→��������� ⊗ �������, �������→��������� ⊗
+��� ⊗ �����,
+�������→������ ⊗ ���� ⊗ �������
+and
+�������→������ ⊗ ���� ⊗ ��� ⊗ �����. Therefore, if Charlie
+checks for message pair (��, ��) when message pair (��, ��)
+is actually transmitted, then the probability of incorrectly
+decoding is as stated in (16).
+Similarly, other kinds of error probabilities can be considered
+as:
+
+If Charlie checks for message pair (��, ��) when
+message pair (��
+� , ��) is indeed transmitted, then the
+probability of incorrectly decoding is:
+�� ��� − Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,��,��,��
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+= ����� − ��������������→������
+⊗ ���� ⊗ ��������
+
+
+
+(29)
+
+If Charlie checks for message pair (��, ��) when
+message pair (��, ��
+� ) is indeed transmitted, then the
+probability of incorrectly decoding is:
+�� ��� − Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��� ,��
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+= ����� − ��������������→���������
+⊗ ��� ⊗ ������
+
+(30)
+
+If Charlie checks for message pair (��, ��) when
+message pair (��
+� , ��
+� ) is indeed transmitted, then the
+probability of incorrectly decoding is:
+Due to the code construction, the error probability under the
+position-based coding scheme is the same for each message pair
+(��, ��):
+�� �����, ���� ≠ (��, ��)�
+= �� ��� − Λ��⊗|ℳ�||��|��⊗|ℳ�||��|�
+(��,��),(��,��)
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+Applying Lemma 3 with � = Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��,��,��,��
+ and � =
+∑
+∑
+∑
+∑
+Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,���,��� ,���
+������
+������
+��� ���
+��� ���
+, we have a
+�� ��� − Γ
+��
+|ℳ�||��|��
+|ℳ�||��|�
+��� ,��,��� ,��
+� ���⊗|ℳ�||��|��⊗|ℳ�||��|�
+��������
+�
+= ����� − ��������������→������
+⊗ ���� ⊗ ��� ⊗ ������
+
+
+(31)
+
+chain of equalities and inequalities as stated in (17). Note that,
+we used (29)-(31).
+Multiple quantum hypothesis testing: As mentioned before,
+the problem of the existence of a simultaneous decoder for a
+general QMAC (more than two users) remained an open
+problem in the i.i.d. case. In [17], the authors presented a helpful
+discussion about the multiple quantum hypothesis testing and
+its relation with QMACs. In summary, the problem of multiple
+hypothesis testing is an open problem too. There are two
+possible
+hypothesis
+testing
+schemes:
+Symmetric
+and
+asymmetric. Chernoff distance from symmetric hypothesis
+testing gives a lower bound on the randomness-assisted error
+exponent [31]; In contrast, the application of the results from
+asymmetric hypothesis testing leads to a lower bound on the
+one-shot randomness-assisted capacity (for QMAC without
+secrecy constraint) and in turn on the second-order coding rate
+for randomness-assisted communication.
+In other words, from [7], we know that there exists a general
+simultaneous decoder to decode more than two messages
+simultaneously in the case of commutative version of outputs,
+and from [17], we know that the multiple hypothesis testing
+problem can be solved if the composite alternative hypothesis
+forms a commutative set of operators. This means that, for a test
+operator �, a finite set of positive semi-definite operators � ≡
+{��: 1 ≤ � ≤ �},
+for
+which
+����(�) ⊆ ����(��)
+and
+min
+�
+�(�‖��) > 0, there are two hypotheses, and we have:
+��{(� − �)�} ≤ �
+(32)
+− log� �� {���} ≥ �min
+�
+�(�‖��)� − �
+(33)
+where � is a positive integer.
+The last inequality holds when the set � forms a
+commutative set of operators. More information can be found
+in [17].
+With these explanations, we use asymmetric hypothesis
+testing for our problem. Note that we want to decode two
+messages simultaneously. Consider the upper bound on error
+probability in (17). Then, we rewrite that as follows:
+�� �����, ���� ≠ (��, ��)�
+≤ (1 + �)��{(� − �)�}
++ (2 + � + ���)��{�(�� + �� + ��)}
+where,
+� = �������→��������� ⊗ �������
+�� = �������→������ ⊗ ���� ⊗ �������
+�� = �������→��������� ⊗ ��� ⊗ �����
+�� = �������→������ ⊗ ���� ⊗ ��� ⊗ �����
+This is called asymmetric hypothesis testing, which tries to
+minimize all other probabilities subject to a constraint on the
+error probability ��{(� − �)�}. Note that we consider all three
+hypotheses (�� + �� + ��) as a unique composite alternative
+hypothesis.
+We can say for such a sequence of test operators, as stated
+in (32), and (33), the above multiple hypothesis testing problem
+can be solved as:
+�� �����, ���� ≠ (��, ��)�
+≤ (1 + �)��{(� − �)�}
++ (2 + � + ���)��{�(�� + �� + ��)}
+= (1 + �)�
++ (2 + � + ���)�|��|2�����
+� (�‖��)
++ |��|2�����
+� (�‖��)
++ |��||��|2��������
+� (�‖��)�
+= (1 + �)�
++ (2 + � + ���)�|��|2�����
+� (��:���)
++ |��|2�����
+� (��:���)
++ |��||��|2��������
+� (����:�)�
+Let |��| = 2��� and |��| = 2���. Then, by setting the above
+term equal to �, with a straightforward simplification, we have:
+�� + ��� = ��
+�(��: ���) + log� �� − (1 + �)�
+2 + � + ��� �
+�� + ��� = ��
+�(��: ���) + log� �� − (1 + �)�
+2 + � + ��� �
+�� + ��� + �� + ��� = ��
+�(����: �) + log� �� − (1 + �)�
+2 + � + ��� �
+The global maximum of the above expression with respect
+to � occurs at � =
+�
+�:
+�� + ��� = ��
+�(��: ���) − log� �4�
+���
+(34)
+�� + ��� = ��
+�(��: ���) − log� �4�
+���
+(35)
+�� + ��� + �� + ��� = ��
+�(����: �) − log� �4�
+���
+(36)
+and for such a �, we have:
+�� �����, ���� ≠ (��, ��)� ≤ � + 2�
+(37)
+Now, we turn our attention to the secrecy criterion. Using
+Lemma 1, we have:
+��� ≤ ����
+�����(��: �)� + log 3
+��� − 1
+4 log ��
+(38)
+��� ≤ ����
+�����(��: ���)� + log 3
+��� − 1
+4 log �� + �(1)
+(39)
+Substituting (38) and (39) in (34)-(36) completes the proof.
+Appendix C: (Proof of Theorem 2)
+The proof uses two successive position-based decoders. The
+first decoder tries to decode the first message ��, and the
+
+second decoder tries to decode the second message �� given
+the true decoded ��. This means that if the first decoder fails,
+the second decoder fails too. This decoding order can be shown
+as �� → ��.
+Constructing the first position-based decoder is the same as
+that presented in [20]. To decode ��, Bob performs his second
+position-based decoder conditioned on ��, which works for all
+�� ∈ ��. It should be noted that, the feeding state of the second
+decoder differs from the main state of the channel.
+Alice, Bob, and Eve are allowed to pre-share some quantum
+state as randomness. Also, Alice has access to two sources of
+uniform junk randomness ��; � ∈ {1,2} . The pre-shared
+randomness is as follows:
+�
+�������������
+⊗|ℳ�||��|
+⊗|ℳ�||��|
+≔ �� �(��)|��⟩⟨��|��
+��
+⊗ |��⟩⟨��|��� �� �(��)|��⟩⟨��|��
+��
+⊗ |��⟩⟨��|����
+⊗|ℳ�||��|
+�
+⊗|ℳ�||��|
+
+
+
+
+
+
+
+
+
+(40)
+The arguments connected to the decoding process for ��
+are listed as follows:
+
+The probability of error for decoding ��:
+��� = ����� ≠ ���
+≔
+1
+|ℳ�| � 1
+2
+|ℳ�|
+����
+�����→���
+��
+����⊗|ℳ�||��|�
+(��,��),(��,��)�
+− |��⟩⟨��|��� ⊗ ����
+� ≤ �� + ���
+
+
+
+
+(41)
+where ����→���
+��
+����⊗|ℳ�||��|�
+(��,��),(��,��)� is decoding map for ��:
+����→���
+��
+����⊗|ℳ�||��|�
+(��,��),(��,��)�
+≔ � � �� �Λ��|ℳ�||��|�
+��,��
+���⊗|ℳ�||��|�
+(��,��),(��,��)�
+|ℳ�|
+����
+|��|
+����
+⊗
+�Λ��|ℳ�||��|�
+��,��
+���⊗|ℳ�||��|�
+(��,��),(��,��)�Λ��|ℳ�||��|�
+��,��
+�� �Λ��|ℳ�||��|�
+��,��
+���⊗|ℳ�||��|�
+(��,��),(��,��)�
+
+
+Λ��|ℳ�||��|�
+��,��
+ is a pretty good measurement (POVM) for
+�� ∈ �1: |ℳ�|�:
+��
+|ℳ�||��|�
+��,��
+≔ � � � Γ��
+|ℳ�||��|�
+���,���
+|ℳ�|
+��
+���
+|��|
+��
+���
+�
+�� �
+�
+��
+|ℳ�||��|�
+��,��
+� � � ��
+|ℳ�||��|�
+���,���
+|ℳ�|
+��
+���
+|��|
+��
+���
+�
+�� �
+�
+
+where ��
+|ℳ�||��|�
+��,��
+ is the element of the first POVM:
+��
+|ℳ�||��|�
+��,��
+≔ ���
+(�,�) ⊗ … ⊗ ���
+(�,|��|) ⊗ … ⊗ ����
+��,�� ⊗ …
+⊗ ���
+(|ℳ�|,|��|)
+and ����
+��,�� is a test operator in order to discriminate between two
+hypotheses ����, and ��� ⊗ ��. Also, it is obvious that to decode ��,
+it does not matter for the second position-based decoder, which copy
+is selected by Alice among |ℳ�||��| copies.
+
+We face a hypothesis testing problem. Null hypothesis is
+���� and alternative hypothesis is ��� ⊗ ��. Therefore the
+probability of success in guessing null and alternative
+hypotheses are ������� ����� and �������� − ����� ���� ⊗
+����.
+The rest of the decoding process for �� is analogous to [20].
+Therefore, we have:
+�� ≤ ��
+�����(��; �)� − �����
+√�����(��; �)� − log 4��
+��
+�
+− 2 log 1
+��
+
+
+
+
+(42)
+Now, we turn our attention to decoding the second message. As
+mentioned before, the channel state changes after the first
+measurement. There is a detailed discussion in [32].
+Let ������(�����)⊗|ℳ�||��|��
+(�1,�1),(�2,�2)
+ denote the disturbed state after applying the
+first measurement (POVM):
+������(�����)⊗|ℳ�||��|��
+(�1,�1),(�2,�2)
+≔ � ���(��)|�1⟩⟨�1|�1 ⊗ |�1⟩⟨�1|���
+��
+⊗ ������
+��(�1,�1) ⊗ … ⊗ ��������
+��(�1,�1),(�2,�2) ⊗ …
+⊗ ������
+��(�1,�1),(|ℳ�|,|��|)
+Also, Bob’s second POVM is as follows:
+
+����
+|ℳ�||��|�
+��,��
+≔ � � � λ����
+|ℳ�||��|�
+���,���
+|ℳ�|
+��
+���
+|��|
+��
+���
+�
+�� �
+�
+���
+|ℳ�||��|�
+��,��
+
+� � � ���
+|ℳ�||��|�
+���,���
+|ℳ�|
+��
+���
+|��|
+��
+���
+�
+�� �
+�
+
+���
+|ℳ�||��|�
+��,��
+ is the element of the second POVM:
+���
+|ℳ�||��|�
+��,��
+≔ |�1⟩⟨�1|�1 ⊗ ���
+(�,�) ⊗ … ⊗ ���
+(�,|��|) ⊗ … ⊗ ����
+��,��
+⊗ … ⊗ ���
+(|ℳ�|,|��|)
+
+����
+��,�� is a binary test operator to discriminate between two hypotheses
+����
+�� and ���
+�� ⊗ ��
+�� with an error of �� − ��; i.e.,
+�����������
+�� � ≥ 1 − (�� − ��)
+�� ∈ (0,1),
+�� ∈ (0, ��)
+In other words, Bob has to be able to discriminate between the
+following states:
+� ���(��)|�1⟩⟨�1|�1 ⊗ ����
+��
+��
+
+� ���(��)|�1⟩⟨�1|�1 ⊗ ���
+�� ⊗ ��
+��
+��
+
+Similar to what mentioned in [20], and [27], we have the following
+rate:
+�� ≤ ��
+�����(��; �|��)� − �����
+√�����(��; �|��)�
+− log 4��
+��
+� − 2 log 1
+��
+
+
+
+(43)
+The probability of error for �� is as follows:
+��� = ����� ≠ ���
+≔
+1
+|ℳ�| � 1
+2
+|ℳ�|
+����
+��������→���
+��
+��
+������������⊗|ℳ�||��|��
+(��,��),(��,��)
+�|��⟩⟨��|���
+⊗ ��������⊗|ℳ�||��|��
+�
+≤ 2��� + ���� + ���
+�
+
+
+
+(44)
+Also, the error probability exponents stated in (41) and (44)
+are proved. See [20, 27].
+This process can be repeated for another decoding order. In
+other words, we can first decode �� , and then decode ��
+( �� → �� ). Then taking the intersection of the regions
+resulting from both orders, we give:
+�� ≤ ��
+�����(��; �|��)� − �����
+√�����(��; �)� − log 4��
+��
+�
+− 2 log 1
+��
+
+�� ≤ ��
+�����(��; �|��)� − �����
+√�����(��; �|��)� − log 4��
+��
+�
+− 2 log 1
+��
+
+This completes the proof.
+Appendix D: (Proof of Theorem 3)
+The proof uses superposition coding. Assume that the first
+receiver �� has a better reception signal than the second receiver
+��. In this setting, Alice is able to encode a further message
+superimposed on top of the common message. Using the
+successive decoding can be helpful.
+Codebook generation: Randomly and independently
+generate 2�� sequence �(��) according to the distribution
+��(�). For each sequence �(��), randomly and conditionally
+independently generate 2�� sequence �(��, ��) according to
+the distribution ��|�����(��)� . The �� ’s state can be
+calculated by tracing out �� from (13):
+����� = � ��(�)��|�(�|�)
+�,�
+|�⟩⟨�|� ⊗ |�⟩⟨�|� ⊗ ��
+��
+Similar to what mentioned for Theorem 2, we construct the
+POVM for the first receiver as:
+��,��
+≔ � � � Γ��� ,���
+|ℳ�|
+��� ��
+|ℳ�|
+�����
+�
+�� �
+�
+��,�� � � � ��� ,���
+|ℳ�|
+��� ��
+|ℳ�|
+�����
+�
+�� �
+�
+
+Also, the POVM for the second receiver can be constructed as
+follows:
+Λ�� ≔ � � λ���
+|ℳ�|
+�����
+�
+�� �
+�
+� � � ��
+|ℳ�|
+�����
+�
+�� �
+�
+
+Consider the probability of error for ��:
+��� = ������, ���� ≠ (��, ��)�
+≔
+1
+|ℳ�||ℳ�| � � �� ���
+��
+��
+− Λ��,�����(��,��)
+��
+�
+and for ��:
+ ��� = ����� ≠ ��� ≔
+1
+|ℳ�| � �� ��� − Λ�����(��)
+��
+�
+��
+
+By a straightforward calculation analogous to [3] for i.i.d. case
+and in [29] (to calculate one-shot Marton inner bound for QBC),
+the above error probability exponents can be calculated as
+follows:
+��� + ��� ≤ 2
+���
+� ��; ������������ � + 2���
+� (�;��)������� �
++ 2���
+� (�;��)������� � + �(�)
+This completes the proof.
+REFERENCES
+[1] A. Winter, “The capacity of the quantum multiple-access channel,” IEEE
+Trans. Inf. Theory, vol. 47, no. 7, pp. 3059–3065, July 2001.
+[2] J. Yard, P. Hayden, and I. Devetak. "Quantum broadcast channels," IEEE
+Trans. on Inf. Theory, vol. 57, no. 10, pp. 7147-7162, October 2011.
+[3] I. Savov, “Network information theory for classical-quantum channels,”
+Ph.D. dissertation, McGill University, Montreal, 2012.
+[4] A. El Gamal and Y.-H. Kim, “Lecture notes on network information
+theory,” January 2010, available online at http://arxiv.org/abs/1001.3404
+[5] N. Cai, A. Winter, and R. W. Yeung. “Quantum Privacy and Quantum
+Wiretap Channels,” Problems of Information Transmission, vol. 40, no.
+4, pp. 318-336, 2004.
+[6] I. Devetak, “The Private Classical Capacity and Quantum Capacity of a
+Quantum Channel,” IEEE Trans. Inf. Theory, vol. 51, no. 1, pp. 44-55,
+January 2005.
+[7] O. Fawzi, P. Hayden, I. Savov, P. Sen, and M. M. Wilde, “Classical
+communication over a quantum interference channel,” IEEE Trans. on
+Inf. Theory, vol. 58, no. 6, pp. 3670-3691, June 2012.
+[8] H. Aghaee, B. Akhbari, “One-Shot Achievable Secrecy Rate Regions for
+Quantum Interference Wiretap Channel ,” The ISC International Journal
+of Information Security (IseCure), vol.14, no. 3, pp 71-80, 2022.
+
+[9] H. Aghaee, B. Akhbari, “Classical-Quantum Multiple Access Wiretap
+Channel,”
+in
+Proc.
+16th
+International
+ISC
+Conference
+on Information Security and Cryptology (ISCISC'19), Mashhad, Iran,
+August 2019.
+[10] H. Aghaee, B. Akhbari, “Private Classical Information over a Quantum
+Multiple Access Channel: One-shot Secrecy Rate Region,” in Proc 10th
+International Symposium on Telecommunications (IST'2020), Iran, 2020.
+[11] H. Aghaee, B. Akhbari, “Classical-Quantum Multiple Access Channel
+with Secrecy Constraint: One-shot Rate Region,” International Journal
+of Information and Communication Technology Research (IJICTR), vol.
+12, no. 2, pp. 1-10, Spring 2020.
+[12] H. Aghaee, B. Akhbari, “Classical-Quantum Multiple Access Wiretap
+Channel with Common Message: One-shot Rate Region,” in Proc. 11th
+Conference on Information and Knowledge Technology (IKT'2020), Iran,
+2020.
+[13] H. Aghaee, B. Akhbari, “Entanglement assisted Classical-Quantum
+Multiple Access Wiretap Channel: One-shot Achievable Rate Region,” in
+Proc. 30th International Conference on Electrical Engineering
+(ICEE2022), Iran, 2022.
+[14] P. Sen, “Unions, intersections and a one-shot quantum joint typicality
+lemma,” Sadhana, 46(1):1–44, 2021.
+[15] A. D. Wyner. “The wire-tap channel,” Bell Syst. Tech. J., vol. 54, no. 8,
+pp. 2–10, October 1975.
+[16] S. Chakraborty, A. Nema, P. Sen, “One-shot inner bounds for sending
+private classical information over a quantum MAC,” 2021 IEEE Inf.
+Theory Workshop (ITW), pp. 1-6, 2021.
+[17] H. Qi, Q. Wang, and M. M. Wilde, “Applications of position-based
+coding
+to
+classical
+communication
+over
+quantum
+channels,”
+https://arxiv.org/abs/1704.01361, 2017.
+[18] A. Anshu. "One-shot protocols for communication over quantum
+networks: Achievability and limitations," PhD diss., National University
+of Singapore (Singapore), 2018.
+[19] A. Anshu, R. Jain and N. A. Warsi, "A Generalized Quantum Slepian–
+Wolf," in IEEE Trans. on Inf. Theory, vol. 64, no. 3, pp. 1436-1453,
+March 2018.
+[20] M. M. Wilde, “Position-based coding and convex splitting for private
+communication over quantum channels,” Quantum Information
+Processing. 16(10):264, October 2017.
+[21] L. Wang and R. Renner, “One-shot classical-quantum capacity and
+hypothesis testing,” Phys. Rev. Lett., vol. 108, no. 20, p. 200501, 2012.
+[22] H. Umegaki, “Conditional expectations in an operator algebra IV (entropy
+and information),” Kodai Math. Sem. Rep., vol. 14, pp. 59-85, 1962.
+[23] M. Berta, M. Christandl and R. Renner, “The quantum reverse Shannon
+theorem based on one-shot information theory,” Commun. Math. Phys.,
+vol. 306, no. 3, pp. 579-615, 2011.
+[24] F. Salek, A. Anshu, M. -H. Hsieh, R. Jain and J. R. Fonollosa, "One-Shot
+Capacity Bounds on the Simultaneous Transmission of Classical and
+Quantum Information," IEEE Trans. on Inf. Theory, vol. 66, no. 4, pp.
+2141-2164, April 2020.
+[25] N. Ciganovi´c, N. J. Beaudry and R. Renner, “Smooth max-information
+as one-shot generalization for mutual information,” IEEE Trans. Inf.
+Theory, vol. 60, pp. 1537-1581, March 2014.
+[26] M. Hayashi, H. Nagaoka, “General formulas for capacity of classical-
+quantum channels,” IEEE Trans. Inf. Theory, vol. 49, pp. 1753–1768, July
+2003.
+[27] F. Salek, A. Anshu, M. -H. Hsieh, R. Jain and J. R. Fonollosa, "One-shot
+Capacity Bounds on the Simultaneous Transmission of Public and Private
+Information Over Quantum Channels," 2018 IEEE International
+Symposium on Information Theory (ISIT), 2018, pp. 296-300.
+[28] M. M. Wilde, M. Tomamichel and M. Berta, "Converse Bounds for
+Private Communication Over Quantum Channels," in IEEE Trans. on Inf.
+Theory, vol. 63, no. 3, pp. 1792-1817, March 2017.
+[29] P. Sen. “Inner bounds via simultaneous decoding in quantum network
+information
+theory,”
+Sādhanā
+46,
+18
+(2021).
+https://doi.org/10.1007/s12046-020-01517-9
+[30] M. Wilde, Quantum Information Theory, Cambridge Univ. Press, 2013.
+[31] K. Li, “Discriminating quantum states: the multiple Chernoff distance,”
+The
+Annals
+of
+Statistics,
+44(4):1661{1679,
+August
+2016.
+arXiv:1508.06624.
+[32] M. M. Wilde, “Sequential decoding of a general classical-quantum
+channel,” Proceedings of the Royal Society A: Mathematical, Physical
+and Engineering Sciences, 469(2157), 2013. 20130259.
+
+
diff --git a/StE0T4oBgHgl3EQflAEY/content/tmp_files/load_file.txt b/StE0T4oBgHgl3EQflAEY/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cc7d91597123140be3fe071673fffbf75900375a
--- /dev/null
+++ b/StE0T4oBgHgl3EQflAEY/content/tmp_files/load_file.txt
@@ -0,0 +1,866 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf,len=865
+page_content='Quantum Multiple Access Wiretap Channel: On the One-Shot Achievable Secrecy Rate Regions Hadi Aghaee Faculty of Electrical Engineering K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Toosi University of Technology Tehran, Iran Email: Aghaee_Hadi@email.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='kntu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='ir Bahareh Akhbari Faculty of Electrical Engineering K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Toosi University of Technology Tehran, Iran Email: akhbari@eetd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='kntu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='ir Abstract— In this paper, we want to investigate classical- quantum multiple access wiretap channels (CQ-MA-WTC) under one-shot setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In this regard, we analyze the CQ-MA-WTC using simultaneous position-based decoder for reliable decoding and using a newly introduced technique in order to decode securely.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, for the sake of comparison, we analyze the CQ-MA- WTC using Sen’s one-shot joint typicality lemma for reliable decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The simultaneous position-based decoder tends to a multiple hypothesis testing problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, using convex splitting to analyze the privacy criteria in a simultaneous scenario becomes problematic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' To overcome both problems, we first introduce a new channel that can be considered as a dual to the CQ-MA-WTC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This channel is called a point-to-point quantum wiretap channel with multiple messages (PP-QWTC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In the following, as a strategy to solve the problem, we also investigate and analyze quantum broadcast channels (QBCs) under the one-shot setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Keywords—Quantum Channel;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Mutual Information;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Secrecy Capacity;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Multiple Access Channel I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' INTRODUCTION The quantum multiple access channel (QMAC) was first introduced by Winter [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' A QMAC can accept two or more messages (classical or quantum) as inputs and one output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Similar to the classical world, decoding messages over a QMAC is based on two main techniques: successive cancelation decoding and simultaneous decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In [1], the author employs the successive cancelation decoding technique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' A quantum broadcast channel (QBC) is a channel with a sender and two or more receivers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The sender wishes to transmit two or more messages (classical or quantum) over the channel to the receivers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The QBC was first introduced by Yard et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In [2], the authors derived an inner bound for QBC for i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' (independent and identical) case, and in [3], the authors derived the same inner bound using a more straightforward method and more in the spirit of its classical analogous [4] than the method in [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In recent decades, with development of quantum data processing and its applications, the necessity to study the security of quantum channels has increased.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In this regard, the quantum wiretap channel (QWTC) was first introduced in [5] and [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then, the secrecy constraints are extended to multi-user quantum channels such as quantum interference channel (QIC) [7,8], and quantum multiple access channel (QMAC) [9-13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' There are two bottlenecks in studying the security of quantum channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The first is decoding three or more messages simultaneously (reliability), and the second is about how we can securely decode two or more messages (confidentiality).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The first bottleneck arises from the nonexistence of a general quantum joint typicality lemma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' However, this problem has been solved in some cases, such as the min-entropy case and QMACs with commutative output [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Therefore, in the independent and identical distributed (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=') case, successive decoding combined with time-sharing techniques should be used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In this setting, transmitters are allowed to transmit their messages by only one use channel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sen proved a joint typicality lemma which helps to decode any number of messages simultaneously in the one-shot case [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=" Obtaining secrecy against the eavesdropper by Wyner's technique [15] of randomizing over a block becomes problematic in the quantum setting." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=" Wyner's technique has been shown to work for point- to-point quantum channels by Devetak [6] and explained further in [16]." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' However, there are no easy generalizations to multiple senders for a quantum channel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This issue is discussed in detail in [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In this paper, we want to investigate the secrecy problem of quantum multiple access channel (QMAC) with classical inputs under one-shot setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, we have investigated some bottlenecks connected to decoding process for CQ-MA-WTC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The achievement of this paper is about analyzing bottlenecks in decoding process and providing solutions to overcome them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, we present two techniques for quantum multiple access wiretap channel with classical inputs (CQMA-WTC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The first approach is based on the method presented in [14], and another technique is the simultaneous position-based decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' From [17], we know that the simultaneous position-based decoder tends to a multiple quantum hypothesis testing problem which is solvable in a special case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, from [18], we know that the convex split lemma could not be used to analyze the privacy of multiple messages in simultaneous decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The paper is organized as follows: In Section II, some seminal definitions are presented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In Section III, the main channel and information processing tasks are presented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In Section IV, the results and main theorems are presented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Section V is dedicated to discussion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' PRELIMINARIES Let A (Alice), B (Bob), and C (Charlie) be three quantum systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' These quantum systems can be denoted by their corresponding Hilbert spaces as ℋ�, ℋ�, and ℋ�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The states of the above quantum systems are presented as density operators ��, ��, and ��, respectively, while the shared state between Alice, Bob, and Charlie is denoted by ����.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' A density operator is a positive semidefinite operator with a unit trace.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Alice, Bob, or Charlie’s state can be defined by a partial trace operator over the shared state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The partial trace is used to model the lack of access to a quantum system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Thus, Alice’s density operator using partial trace is �� = ����{����}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' |�⟩� denotes the pure state of system A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The corresponding density operator is �� = |�⟩⟨�|�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The von Neumann entropy of the state �� is defined by �(�)� = −��{�� log ��}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' For an arbitrarily state such as ���, the quantum conditional entropy is defined by �(�|�)� = �(�, �)� − �(�)�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The quantum mutual information is defined by �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� = �(�)� + �(�)� − �(�, �)�, and the conditional quantum mutual information is defined by: �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|�)� = �(�|�)� + �(�|�)� − �(�, �|�)� Quantum operations can be denoted by completely positive trace-preserving (CPTP) maps ��→�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The CPTP maps accept input states in A and output states in B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The distance between two quantum states, such as A and B, is defined by trace distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The trace distance between two arbitrary states, such as � and � is: ‖� − �‖� = ��|� − �| (1) where |Ψ| = √Ψ�Ψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This quantity is zero for two similar and perfectly distinguishable states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fidelity is defined as �(�, �) = ���√��� �, and purified distance is a metric on �(ℋ) and is defined as �(�, �) ≔ �1 − �(�, �)�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Most of the above definitions are given in [19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Definition 1: (Hypothesis testing mutual information) Hypothesis testing mutual information is denoted by �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �) ∶= �� � (���‖�� ⊗ ��), � ∈ (0,1) and is considered as quantum hypothesis testing divergence [17] where �� � (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ‖.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=') is hypothesis testing relative entropy [17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � is the smoothing variable, �ℋ�ℋ� is the joint classical-quantum state of input and output over their Hilbert spaces (ℋ�, ℋ�), and it can be shown as ���: ��� = � ��(�)|�⟩⟨�|� ⊗ �� � � where �� is the input distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Definition 2: (Quantum relative entropy [20]): Consider states ��, �� ∈ �(ℋ�).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The Quantum relative entropy is defined as: �(��‖��) ≔ ���{���log� �� − log� ���} ����(��) ⊆ ����(��) +∞ ��ℎ������ where ����(��) refers to the set-theoretic support of �.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ����(�) is the subspace of ℋ spanned by all eigenvectors of � with non-zero eigenvalues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fact 1: The following relation exists between the quantum relative entropy and hypothesis testing relative entropy for � ∈ (0,1) [21]: �� �(��‖��) ≤ 1 1 − � ��(��‖��) + ℎ�(�)� where ℎ�(�) ≔ −� log� � − (1 − �) log�(1 − �) is a binary entropy function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Definition 3: (Max mutual information [21]) Consider a bipartite state ��� and a parameter � ∈ (0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The max mutual information can be defined as follows: ����(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� ≔ ����(��� ‖��⨂�� )� where � refers to the state ��� and ����(∣∣) is the max-relative entropy [22] for ��, �� ∈ ℋ�: ����(�� ‖��) ≔ inf{� ∈ ℝ: �� ≤ 2���} Definition 4: (Quantum smooth max relative entropy [22]) Consider states ��, �� ∈ �(ℋ�), and � ∈ (0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The quantum smooth max relative entropy is defined as: ���� � (��‖��) ∶= inf �� � ∈ℬ�(��) ����(�� � ‖�� ) where ℬ�(��) ≔ {�� � ∈ �(ℋ�): �(�� � , ��) ≤ �} is �-ball for ���.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Definition 5: (Quantum smooth max mutual information [21]) Consider ��� ∶= ∑ ��(�)|�⟩⟨�|� ⊗ �∈� �� � as a classical- quantum state and a parameter � ∈ (0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The smooth max mutual information between the systems � and � can be defined as follows: ���� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �) ∶= inf ��� � ∈ℬ�(���) ����(��� � ‖��⨂�� ) = inf ��� � ∈ℬ�(���)����(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)�� , where ℬ�(���) ≔ {��� � ∈ �(ℋ� ⊗ ℋ�): �(��� � , ���) ≤ �} is �-ball for ���.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Definition 6: (Conditional smooth hypothesis testing mutual information [23]) Consider ���� ∶= ∑ ��(�)|�⟩⟨�|� �∈� ⊗ ��� � be a tripartite classical-quantum state and � ∈ (0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We define, �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|�)� ≔ max �� min �∈������� � � �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)��� � , where maximization is over all �� � = ∑ ��(�)|�⟩⟨�|� �∈� satisfying �(�� � , ��) ≤ �.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fact 2: [24] Let ���� ∶= ∑ ��(�)|�⟩⟨�|� �∈� ⊗ ��� � be a tripartite classical-quantum state and � ∈ (0,1), the following relation holds, lim �→� 1 � �� �(�⨂�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|��)�⨂� = �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|�)� Definition 7: (Alternate smooth max-mutual information) Consider a bipartite state ��� and a parameter � ∈ (0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The alternate definition of the smooth max-mutual information between the systems � and � can be defined as follows: ����� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �) ∶= inf ��� � ∈ℬ�(���) ����(��� � ‖�� ⨂ �� � ) Fact 3: (Relation between two definitions of the smooth max mutual information) [25]: Let � ∈ (0,1) and � ∈ (0, �) For a bipartite state ���, it holds that: ����� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� ≤ ���� ��� (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� + log 3 �� Definition 8: (Conditional smooth max mutual information [23]) Consider ���� ∶= ∑ ��(�)|�⟩⟨�|� �∈� ⊗ ��� � be a tripartite classical-quantum state and � ∈ (0,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We define, ���� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|�)� ≔ max �� min �∈������� � � ���� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)��� � , where maximization is over all �� � = ∑ ��(�)|�⟩⟨�|� �∈� satisfying �(�� � , ��) ≤ �.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fact 4: [24] ���� ∶= ∑ ��(�)|�⟩⟨�|� �∈� ⊗ ��� � be a tripartite classical-quantum state and � ∈ (0,1), the following relation holds, lim �→� 1 � ���� � (�⨂�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|��)�⨂� = �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|�)� Definition 9: (Quantum Rényi relative entropy of order � [17]) For a state � ∈ �(ℋ) and a positive semidefinite operator �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' the quantum Rényi relative entropy of order �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' where � ∈ �0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='1) ∪ (1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' +∞) is defined as: ��(�‖�) ≡ 1 � − 1 log� ��{������} Also,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Rényi entropy of order � can be defined as follows: ��(�)� ≡ 1 1 − � log� ��{�� �} Definition 10: (One-shot inner bound of a classical- quantum multiple access channel) [14] A two-user classical- quantum multiple access channel (C-QMAC) under the one- shot setting is a triple (�� × ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �����→�(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��) ≡ ����� � ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ℋ�),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' where �� and �� are the alphabet sets of two classical inputs,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' and � is the output system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ����� � is a quantum state, and the channel has a completely positive trace- preserving map (CPTP) �����→�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Considering the joint typicality lemma introduced in [Corollary 4, 14], the one-shot inner bound of a C-QMAC is as follows: �� ≤ �� �(��: ���)� − 2 + log � �� ≤ �� �(��: ���)� − 2 + log � �� + �� ≤ �� �(����: �)� − 2 + log � with decoding error at most 49√�, where �� �(.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ) is the hypothesis testing mutual information defined in Definition 1 with respect to the controlling state: ������� ∶= � �(�)�(��|�)�(��|�)|�����⟩⟨�����|����� ����� ⊗ ����� � (2) and � is a time-sharing variable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Note that �� �(: ) is the difference between a Rényi entropy of order two and a conditional quantum entropy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Lemma 1: [16] Given the control state in (2) (without time- sharing variable), �� > 0 and 0 < �� < ��,let ���, … , ���� and ���, … , ���� be i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' samples from the distributions �� and ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then, if log|��| ≥ ���� �����(�: �)� + log 3 ��� − 1 4 log �� log|��| ≥ ���� �����(�: ��)� + log 3 ��� − 1 4 log �� + �(1) the following holds, ���,…,���~�� ��,…,���~�� � 1 |��||��| � � ����� � − �� |��| ��� |��| ��� � � ≤ 20��� � Proof: see [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Lemma 2: : (Convex split lemma) [19,20] Let ��� be an arbitrary state and suppose that ���…��� be the following state: ���…��� = 1 � � ��� ⊗ … ⊗ ����� ⊗ ���� ⊗ ����� ⊗ … � ��� ⊗ ��� Let � ∈ (0,1) and � ∈ �0, √��, if log� � = ����� √���(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� + 2 log� �1 �� then, �����…���, ��� ⊗ … ⊗ ��� ⊗ ���� ≤ √� for some state ��� such that �(��, ���) ≤ √� − �.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Proof: see [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Lemma 3: (Hayashi-Nagaoka inequality [26]) Suppose that �,� ∈ �(ℋ�) such that (� − �) ∈ �(ℋ�) are operators such that � ≥ 0 and 0 ≤ � ≤ �, then for all positive constant �, the following relation holds: � − (� + �)�� � � (� + �)�� � ≤ (1 + �)(� − �) + (2 + � + ���)� Proof: see [26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' CHANNEL MODEL A two-user CQ-MA-WTC is a triple (�� × ��, �����→��(��, ��) ≡ ����� �� , ℋ� ⊗ ℋ�), where ��, � ∈ {1,2} denote the input alphabet sets, and �, � denote the output systems (� denotes the channel output at the legitimate receiver (Charlie), and � is the channel output at the eavesdropper).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ����� �� is the system output’s quantum state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Both users want to transmit their messages as secure as possible over a CQ-MA-WTC to the receiver.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The CQ-MA-WTC model The main channel is illustrated in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Consider the main channel illustrated in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Each user chooses its message ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} from its message set ℳ� = �1: |ℳ�| = 2���;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} (�� and �� are the transmitting rates corresponding to the first and the second messages, respectively), and sends it over a CQ-MA-WTC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The users also use two junk variables ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} from two amplification sets �� = �1: |��| = 2����;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} for randomizing Eve’s knowledge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We have two doubly indexed codebooks ��(��, ��), and ��(��, ��), for user-1 and user-2, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' MAIN RESULTS In this section, we present the main results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Corollary 1 gives a one-shot achievable secrecy rate region for sending classical messages over a CQ-MA-WTC based on Sen’s quantum joint typicality lemma [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The second theorem presents a novel approach to decode both messages over a CQ- MA-WTC reliably and confidentially (simultaneous position- based decoder).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' It should be noted that Corollary 1 and Theorem 1 use the same method to prove the security requirements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, we present a theorem that tries to overcome the bottlenecks connected to Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Corollary 1: (One-shot achievable rate region for CQ-MA- WTC) Consider a two-user CQ-MA-WTC which accepts �� and �� as inputs and �, and � as outputs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ����� �� is the channel density operator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' For any fixed � ∈ (0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� ∈ (0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��) and �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� such that �� > 0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' the rate pair ���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 49√� + 20�� � �� is achievable to satisfy the following inequalities: �� ≤ �� �(��: ���|�)� − ���� � (��: �|�)� + log � − 2 − log 3 ��� + 1 4 log �� �� ≤ �� �(��: ���|�)� − ���� � (��: ���|�)� + log � − 2 − log 3 ��� + 1 4 log �� + �(1) �� + �� ≤ �� �(����: �|�)� − ���� � (��: �|�)� − ���� � (��: ���|�)� + log � − 2 − 2 log 3 ��� + 1 2 log �� + �(1) where � = �� − �� and the union is taken over input distribution ��(�)���|�(��|�)���|�(��|�).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Q is the time- sharing random variable, and all of the mutual information quantities are taken with respect to the following state: �������� ≡ � ���(�)���|�(��|�)���|�(��|�)|�⟩⟨�|� �,��,�� ⊗ |��⟩⟨��|�� ⊗ |��⟩⟨��|�� ⊗ ����� �� (3) Proof: See Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sketch of proof: The proof has two steps: 1- Reliable decoding based on Sen’s quantum one-shot joint typicality (Definition 9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 2- Secure decoding based on Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theorem 1: (one-shot lower bound for CQ-MA-WTC) For any fixed � ∈ (0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� ∈ (0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='1) and �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� such that � ∈ (0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' and �� ∈ (0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' there exists a one-shot code for the channel �����→��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' if rate pair ���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � + 2� + 20�� � �� satisfies the following bounds: �� ≤ �� �(��: ���|�)� − ���� � (��: �|�)� − log� �4� ��� − log 3 ��� + 1 4 log �� �� ≤ �� �(��: ���|�)� − ���� � (��: ���|�)� − log� �4� ��� − log 3 ��� + 1 4 log �� + �(1) �� + �� ≤ �� �(����: �|�)� − ���� � (��: �|�)� − ���� � (��: ���|�)� − log� �4� ��� − 2 log 3 ��� + 1 2 log �� + �(1) where � = �� − �� and the union is taken over input distribution ��(�)���|�(��|�)���|�(��|�).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Q is the time- sharing random variable, and all mutual information quantities are taken with respect to the state (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Proof: See Appendix B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sketch of proof: The proof has two steps: 1- Reliable decoding based on the simultaneous position-based technique: for simplicity of analysis, we merge reliability and confidentiality criteria into a single criterion [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 2- Secure decoding based on the Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Remark 1: It should be noted that, both of the above theorems tend to the same result if and only if � = �.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' As mentioned before, the simultaneous position-based decoder tends to a multiple hypothesis testing problem which is unsolvable in the general case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, the convex split lemma (Lemma 2) does not make sense in the simultaneous decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Because it runs to the famous smoothing bottleneck of quantum information theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Now, consider the channel illustrated in Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' α2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='YZ Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=" wiretapper (M', M2) Px1x2 m-+→(Mi, M2)m V Figure 2." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The PP-QWTC model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This channel accepts two or more messages from one user.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We call this channel a point-to-point quantum wiretap channel with multiple messages (PP-QWTC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Consider PP-QWTC with classical messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This channel is studied in [27] under a different scenario wherein a sender wants to send classical and quantum messages simultaneously to a legitimate receiver.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Information processing task: Two classical messages (��, ��) ∈ ℳ� × ℳ� are possessed by a sender (Alice) and be transmitted to a receiver (Bob) in the presence of a passive wiretapper over a point-to-point quantum channel under the one-shot scenario.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Both of the messages, should be kept as secure as possible from the wiretapper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The PP-QWTC is a triple (�, ��→��(��, ��) ≡ ��(��,��) �� , ℋ� ⊗ ℋ�), where � denotes the input alphabet sets, and �, � denote the output systems (� denotes the channel output at the legitimate receiver (Bob), and � is the channel output at the eavesdropper).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��(��,��) �� ≡ ����� �� is the system output’s quantum state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Alice chooses its message ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} from its message set ℳ� = �1: |ℳ�| = 2���;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2}, and sends it over a PP- QWTC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Alice also uses two junk variables ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} from two amplification sets �� = �1: |��| = 2����;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} for randomizing Eve’s knowledge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We have two doubly indexed codebooks ��(��, ��), and ��(��, ��).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Encoding: An encoding operation by Alice ℰ: ���� → �(ℋ�) ∀��, �� ∈ ��, �� 1 2 ������� − ����� ⊗ ����� ≤ �� (4) where for each message ��, ��, ������ and ����� are appropriate marginal of the state ������� = � |ℳ�||ℳ�| ∑ ∑ |��⟩⟨��| ⊗ |��⟩⟨��| ⊗ |ℳ�| ���� |ℳ�| ���� ��ℰ(��, ��)�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, ��� can be any arbitrary state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Decoding: Decoding operation by Bob �: �(ℋ�) → ������ such that: �� �����, ���� ≠ (��, ��)� ≤ �1 (5) A rate pair (��, ��) is (�1, �2)-achievable if, for such encoding and decoding maps (ℰ, �), the conditions stated in (4) and (5) are satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' As it can be understood from criterion (4), the reliability and confidentiality conditions are merged into a single criterion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This idea is used in [28] and [20] for the first time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The QBC model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theorem 2: (An inner bound on the one-shot capacity region of PP-QWTC) For any fixed �� ∈ (0,1), �� ∈ (0,1) and ��, �� such that �� ∈ (0, ��) and �� ∈ (0, ��), there exists a one- shot code for the channel ��→��, if rate pair (��, ��, 3�� + 2√�� + 2√��, 2(�� + √��) + √��) satisfies the following bounds: �� ≤ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ����� √�����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� − log 4�� �� � − 2 log 1 �� �� ≤ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ����� √�����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − log 4�� �� � − 2 log 1 �� with respect to state ������� = ∑ ∑ �(��, ��)|��⟩⟨��| ⊗ |��⟩⟨��| ⊗ ��� ���� |��| ���� |��| ���� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Proof: In Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Remark 2: The proof of Theorem 2 has two advantages over the proof of Theorem 1: The first is that the proof of Theorem 2 is based on solving a binary hypothesis testing problem against the proof of Theorem 1, which is based on solving a multiple hypothesis testing problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The second is that in the privacy proof of Theorem 1, Lemma 1 [16] is used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' But, in the proof of Theorem 2 the convex split lemma (Lemma 2) can be used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Remark 3: From a comparison between the results of Theorem 1 and Theorem 2, it can be understood that the proof of Theorem 3 does not give the sum-rate (�� + ��) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This is because of using the successive decoding technique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This issue should not cause doubts about whether PP-QWTC is a dual for CQ-MA-WTC or not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' To solve this doubt, we propose the issue of quantum broadcast channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Quantum broadcast channels The quantum broadcast channel (QBC) accepts one user and two or more receivers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In the basic case, the sender (Alice) wishes to transmit three separate messages: �� is the personal message for the first receiver ��, �� is the personal message for the second receiver ��, and �� is the common message for both of the receivers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The basic QBC is illustrated in Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' It should be noted that, for ease of calculation, we removed the security constraint from the problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' →(Mc)X(m1,mc ) (M1,Mc UXY Y2OBCYZ Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' wiretapper (M",' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M*) Puru21 di 4-J →(Mi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M2) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='XV���������� ≡ � ���(��)|��⟩⟨��|�� ⊗ |��⟩⟨��|��� ⊗ |��⟩⟨��|���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(6) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���������� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='≡ � ���(��)|��⟩⟨��|�� ⊗ |��⟩⟨��|��� ⊗ |��⟩⟨��|���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(7) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��������� ≡ � ���(��)|��⟩⟨��|�� ⊗ ��� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���� ⊗ |��⟩⟨��|���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(8) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��������� ≡ � ���(��)|��⟩⟨��|�� ⊗ ��� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���� ⊗ |��⟩⟨��|���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(9) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='������� ≡ �������→������������ ⊗ ���������� = � ���(��)���(��) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��⟩⟨��|�� ⊗ |��⟩⟨��|�� ⊗ ��� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(10) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|ℳ�||ℳ�| � � 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|ℳ�| ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|ℳ�| ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���→������ �� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�������� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='⊗ |ℳ�||��|�������� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='⊗ |ℳ�||��|�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) − �|��⟩⟨��|��� ⊗ |��⟩⟨��|���� ⊗ �� ����⊗|ℳ�||��|����⊗|ℳ�||��|��� � ≤ � + 2� + 20��� � where �� ����⊗|ℳ�||��|����⊗|ℳ�||��|� ≔ � ����⊗|ℳ�||��|����⊗|ℳ�||��| (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) ⊗ ���.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' (11) ��→������ ����⊗|ℳ�||��|��⊗|ℳ�||��|�� (��,��),(��,��) � ≔ � � ����(��)����(��)|��⟩⟨��|��� ⊗ |��⟩⟨��|��� |ℳ�| ���� |ℳ�| ���� = � � �� �Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� ���� ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � |��⟩⟨��|��� ⊗ |��⟩⟨��|��� |ℳ�| ���� |ℳ�| ���� (12) The problem of QBC is widely studied in the i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' case in [2-3] and in the one-shot case in [29].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In the following, we want to achieve a one-shot inner bound for QBC with classical messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Suppose that Alice has not a personal message for the second receiver �� (�� = ∅ → �� = 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The QBC under the one-shot setting is a triple ( �, ��→���� ≡ �� ����, ℋ�� ⊗ ℋ��) , where � denotes the input alphabet set, and ��, �� denote the output systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� ���� is the system output’s quantum state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theorem 3: (one-shot inner bound for QBC) Let � be an auxiliary random variable, � = ��|�(�|�)��(�) be the code probability function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The one-shot achievable rate consists of all rate pairs (��, ��) such that: �� ≤ �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|�)� − 2 + log � �� ≤ �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� + �� ≤ �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � is achievable, and all information quantities are taken with respect to the following state: ������� = � ��(�)��|�(�|�) �,� |�⟩⟨�|� ⊗ |�⟩⟨�|� ⊗ �� ���� (13) Proof: In Appendix D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Now, consider the extended version of the above theorem: Corollary 2: (one-shot inner bound for QBC with three personal messages for the first receiver) Let � be an auxiliary random variable, � = ��(�)���|�(��|�)���|���(��|���) be the code probability function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The one-shot achievable rate region consists of all rate tuples (��, ��, ��) in order to sending (��, ��, ��) such that: �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|�)� − 2 + log � �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|���)� − 2 + log � �� ≤ �� �(�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� + �� ≤ �� �(����;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|�)� − 2 + log � �� + �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� + �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|��)� − 2 + log � is achievable,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' and all information quantities are taken with respect to the following state: ���������� = � ��(�)���|�(��|�) �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='� ���|���(��|���)|�⟩⟨�|� ⊗ |��⟩⟨��|�� ⊗ |��⟩⟨��|�� ⊗ ����� ���� Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) ≔ �� � � � Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ��� ��� ��� ��� � �� � Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� �� � � � Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ��� ��� ��� ��� � �� � (14) Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ≔ ����� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�) ⊗ … ⊗ ����� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) ⊗ … ⊗ ����� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='����) ⊗ … ⊗ ����� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) ⊗ … ⊗ ����� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='����),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) ⊗ ������ (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) ⊗ ����� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='����) ⊗ … ⊗ ����� (|ℳ�|,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(|ℳ�|,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|) (15) �� ��� − Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � = �� ��� − ��������������→��������� ⊗ �������� (16) �� �����,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ���� ≠ (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� ≤ (1 + �)�� ��� − Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � + (2 + � + ���) � � � � �� �Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � ������ ������ ��� ��� ��� ��� = (1 + �)�� ��� − Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � + (2 + � + ���) � � �� �Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � ������ ��� ��� + (2 + � + ���) � � �� �Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � ������ ��� ��� + (2 + � + ���) � � � � �� �Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � ������ ������ ��� ��� ��� ��� = (1 + �)�� ��� − ��������������→��������� ⊗ �������� + (2 + � + ���)(|ℳ�||��| − 1)�� ��� − ��������������→������ ⊗ ���� ⊗ �������� + (2 + � + ���)(|ℳ�||��| − 1)�� ��� − ��������������→��������� ⊗ ��� ⊗ ������ + (2 + � + ���)(|ℳ�||��| − 1)(|ℳ�||��| − 1)�� ��� − ��������������→������ ⊗ ���� ⊗ ��� ⊗ ������ (17) Proof: The proof follows the extended version of Theorem 3’s proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The channel described in Corollary 2 will be converted to the channel described in Theorem 2 (PP-QWTC) without secrecy constraint by choosing �� = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Set �� = 0 in Corollary 2: �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|��)� − 2 + log � �� + �� ≤ �� �(����;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|��)� − 2 + log � where the above last two rates are redundant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then, we have the following region: �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|��)� − 2 + log � �� + �� ≤ �� �(����;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� − 2 + log � (18) Consider the results of Theorem 2 without the leaked information terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' By choosing �� = ��, we have: �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|��)� − 2 + log �� �� ≤ �� �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��|��)� − 2 + log �� (19) Comparing (18), (19), and (33), the argument stated in Remark 3 is proved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, the region (18) is a near-optimal achievable rate region compared to Corollary 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' As it can be understood from a comparison between the results of Corollary 1, Theorem 2, and Corollary 2 by considering (31), this idea can be proved that converting the CQ-MA-WTC to PP-QWTC can be a helpful approach to bypass the bottlenecks connected to the multiple hypothesis testing problem (Theorem 1) and the smoothing bottlenecks of quantum information theory (Corollary 1 and Theorem 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Asymptotic analysis In this subsection, we want to evaluate our secrecy rate region in the asymptotic i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' case (asymptotic limit of many uses of a memoryless channel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Consider PP-QWTC (�(��, ��) → �� ��).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The capacity region of the channel can be expressed as follows: ��(�) ≔ lim ��,��→� lim �→� 1 � ���,��(�⊗�) (20) where ���,��(�⊗�) ≡ max �(��,��) ℛ��,��(�⊗�).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Let ℛ(�) be the set of the maximum rate pairs (�� �, �� � ), ℛ(�) = � �� � ≤ �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� �� � ≤ �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� (21) Then the capacity region ��(�) is the union over � uses of the channel �: ��(�) ≔ max �(��,��) 1 � � ℛ(�⊗�) � ��� (22) Our aim is to prove the expression above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Consider both of single rates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Applying Fact 2 (and its conditional version) we have, �� ≥ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ���� √�������(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� − log 4�� �� � − 2 log 1 �� − log 3 �� �� ≥ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ���� √�������(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − log 4�� �� � − 2 log 1 �� − log 3 �� To prove the achievability, consider the one-shot lower bounds presented in Theorem 2, and apply quantum AEP [30] for the conditional smooth hypothesis testing, and max-mutual information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' From Theorem 2, for � uses of the channel �, the following lower bound ���,��(�⊗�) can be obtained: � ℛ(�⊗�) � ��� ⊆ ���,��(�⊗�) where ℛ(�⊗�) is the set of all rate pairs (�� �, �� � ) satisfying: �� � ≤ �� �����(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|�� �)� − ���� √�������(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�)� − log 4�� �� � − 2 log 1 �� − log 3 �� (23) �� � ≤ �� �����(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|�� �)� − ���� √�������(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|�� �)� − log 4�� �� � − 2 log 1 �� − log 3 �� (24) We can assume that the sequences of the random variables are generated in an i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' fashion according to their distributions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This is due to the fact that the region above is basically a lower bound on the capacity region.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This empowers us to make use of quantum AEP as described below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' From Fact 3, we have, lim ��→� lim �→� 1 � �� �����(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|�� �)�⨂� = �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� (25) lim ��→� lim �→� 1 � �� �����(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|�� �)�⨂� = �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� (26) Also, using Fact 4, we have the following: lim ��→� lim �→� 1 � ���� √�������(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�)�⨂� = �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� (27) lim ��→� lim �→� 1 � ���� √�������(�� �;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �⨂�|�� �)�⨂� = �(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� (28) Putting (25), (26), (27), and (28) into (23), and (24) gives (21): ℛ(�⊗�) ⊆ lim ��,��→� lim �→� 1 � ���,��(�⊗�) Given the argument above and using (20), and (22) completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' DISCUSSION In this paper, we studied the problem of secure communication over a CQ-MA-WTC using three techniques: 1- Sen’s joint typicality lemma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 2-simultaneous position-based decoding and, 3-successive position-based decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The first and the second decoding techniques use a newly introduced smooth technique [16] to analyze the privacy, while the third technique uses convex splitting [19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We realized that the simultaneous position-based decoder tends to a multiple hypothesis testing problem which is unsolvable in the general case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We introduced a new channel (PP-QWTC) which can be considered as a dual for CQ-MA-WTC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, this channel can be derivate from the quantum broadcast channel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The results show that the PP-QWTC has a near-optimal achievable rate region to CQ-MA-WTC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' APPENDIX Appendix A: (Proof of Corollary 1) As mentioned, the proof has two steps: Reliable decoding and secure decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' To these ends, consider two junk variables ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} for each of users ��, � ∈ {1,2}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' These junk variables are used to make two doubly indexed codebooks {��(��, ��)}��∈ℳ�,��∈�� and {��(��, ��)}��∈ℳ�,��∈��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Bob should be able to detect the pair messages (��, ��), and the junk variables ��, and �� with high probability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Using Definition 10 (Sen’s inner bound for QMAC), we have the following relation: ℛ������������ = ℛ��� − ℛ������ with decoding error at most 49√�, and privacy leakage at most 20�� � � (Lemma 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, ℛ��� refers to Sen’s inner bound for QMAC (Definition 9), and ℛ������ refers to the leaked information from senders to Eve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' From Lemma 1, we have the following: ��������� ≤ ���� �����(��: �)� + log 3 ��� − 1 4 log �� ��������� ≤ ���� �����(��: ���)� + log 3 ��� − 1 4 log �� + �(1) This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Appendix B: (Proof of Theorem 1) Both of the messages are uniformly distributed on their sets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The receiver has to be able to decode both messages with negligible error probability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Before communication begins, Alice (A) and Bob (B) share randomness with Charlie (C), and wiretapper (Z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Let ���������� (6) and ���������� (7) be shared- randomness between (A,C,Z) and shared-randomness between (B,C,Z), respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Alice has �� � system, Bob has �� � system, and Charlie has (��, ��) system, and wiretapper has (�� ��, �� ��) system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Let ��������� (8) and ��������� (9) denote the state resulting from sending �� � and �� � over the channel, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then the overall controlling state of the channel is as stated in (10).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sketch of the coding scheme: For each of the messages (��), � ∈ {1,2}, there exist local keys �� ∈ �1: |��|�, � ∈ {1,2} as uniform randomness for randomizing Eve’s knowledge about the sent messages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' These local keys are not accessible to Charlie or Eve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Before the communication begins, assume that Alice, Charlie, and Eve share |ℳ�||��| copies of the state in (6) and Bob, Charlie and Eve share |ℳ�||��| copies of the state in (7): � �� |ℳ�||��|���|ℳ�||��|����|ℳ�||��| = ���������� ⊗|ℳ�||��| � �� |ℳ�||��|���|ℳ�||��|����|ℳ�||��| = ���������� ⊗|ℳ�||��| To send the pair messages ��, and ��, Alice and Bob pick �� ∈ �1: |��|� and �� ∈ �1: |��|�, respectively, and uniformly at random.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' They send (��, ��)-th system �� � and (��, ��)-th system �� � through the channel �������→��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' There exists a simultaneous decoder for communication over a CQ-MA-WTC with the upper bound on the average error probability, as stated in (11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' As it can be understood from (11), the security criterion is merged into the reliability criterion [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The simultaneous position-based decoder can be constructed as stated in (12), where, Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� ���� = � � Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� (��,��),(��,��) |��| ���� |��| ���� Now, we consider the error term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Charlie constructs her position-based decoder to decode ��, ��, ��, and ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Let Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) denotes the POVM: �� ����� |ℳ�||��|��|ℳ�||��|� − Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � ≤ � where Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) is expressed in (14),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' and for �� ∈ �1: |ℳ�|� and �� ∈ �1: |��|�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� is expressed in (15),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' in which ������ (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) is a test operator used to discriminate between hypotheses ������,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ����� ⊗ ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��� ⊗ ���� and ��� ⊗ ��� ⊗ �� with an error of �.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Note that, this hypothesis testing problem is equal to discriminating between hypotheses �������→��������� ⊗ �������, �������→��������� ⊗ ��� ⊗ �����, �������→������ ⊗ ���� ⊗ ������� and �������→������ ⊗ ���� ⊗ ��� ⊗ �����.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Therefore, if Charlie checks for message pair (��, ��) when message pair (��, ��) is actually transmitted, then the probability of incorrectly decoding is as stated in (16).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Similarly,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' other kinds of error probabilities can be considered as: If Charlie checks for message pair (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��) when message pair (�� � ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��) is indeed transmitted,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' then the probability of incorrectly decoding is: �� ��� − Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � = ����� − ��������������→������ ⊗ ���� ⊗ �������� (29) If Charlie checks for message pair (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��) when message pair (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� � ) is indeed transmitted,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' then the probability of incorrectly decoding is: �� ��� − Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � = ����� − ��������������→��������� ⊗ ��� ⊗ ������ (30) If Charlie checks for message pair (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��) when message pair (�� � ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �� � ) is indeed transmitted,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' then the probability of incorrectly decoding is: Due to the code construction,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' the error probability under the position-based coding scheme is the same for each message pair (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��): �� �����,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ���� ≠ (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� = �� ��� − Λ��⊗|ℳ�||��|��⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � Applying Lemma 3 with � = Γ �� |ℳ�||��|�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� and � = ∑ ∑ ∑ ∑ Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ������ ������ ��� ��� ��� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' we have a �� ��� − Γ �� |ℳ�||��|�� |ℳ�||��|� ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � ���⊗|ℳ�||��|��⊗|ℳ�||��|� �������� � = ����� − ��������������→������ ⊗ ���� ⊗ ��� ⊗ ������ (31) chain of equalities and inequalities as stated in (17).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Note that, we used (29)-(31).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Multiple quantum hypothesis testing: As mentioned before, the problem of the existence of a simultaneous decoder for a general QMAC (more than two users) remained an open problem in the i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In [17], the authors presented a helpful discussion about the multiple quantum hypothesis testing and its relation with QMACs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In summary, the problem of multiple hypothesis testing is an open problem too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' There are two possible hypothesis testing schemes: Symmetric and asymmetric.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Chernoff distance from symmetric hypothesis testing gives a lower bound on the randomness-assisted error exponent [31];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In contrast, the application of the results from asymmetric hypothesis testing leads to a lower bound on the one-shot randomness-assisted capacity (for QMAC without secrecy constraint) and in turn on the second-order coding rate for randomness-assisted communication.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In other words, from [7], we know that there exists a general simultaneous decoder to decode more than two messages simultaneously in the case of commutative version of outputs, and from [17], we know that the multiple hypothesis testing problem can be solved if the composite alternative hypothesis forms a commutative set of operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This means that, for a test operator �, a finite set of positive semi-definite operators � ≡ {��: 1 ≤ � ≤ �}, for which ����(�) ⊆ ����(��) and min � �(�‖��) > 0, there are two hypotheses, and we have: ��{(� − �)�} ≤ � (32) − log� �� {���} ≥ �min � �(�‖��)� − � (33) where � is a positive integer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The last inequality holds when the set � forms a commutative set of operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' More information can be found in [17].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' With these explanations, we use asymmetric hypothesis testing for our problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Note that we want to decode two messages simultaneously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Consider the upper bound on error probability in (17).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then, we rewrite that as follows: �� �����, ���� ≠ (��, ��)� ≤ (1 + �)��{(� − �)�} + (2 + � + ���)��{�(�� + �� + ��)} where, � = �������→��������� ⊗ ������� �� = �������→������ ⊗ ���� ⊗ ������� �� = �������→��������� ⊗ ��� ⊗ ����� �� = �������→������ ⊗ ���� ⊗ ��� ⊗ ����� This is called asymmetric hypothesis testing, which tries to minimize all other probabilities subject to a constraint on the error probability ��{(� − �)�}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Note that we consider all three hypotheses (�� + �� + ��) as a unique composite alternative hypothesis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We can say for such a sequence of test operators, as stated in (32), and (33), the above multiple hypothesis testing problem can be solved as: �� �����, ���� ≠ (��, ��)� ≤ (1 + �)��{(� − �)�} + (2 + � + ���)��{�(�� + �� + ��)} = (1 + �)� + (2 + � + ���)�|��|2����� � (�‖��) + |��|2����� � (�‖��) + |��||��|2�������� � (�‖��)� = (1 + �)� + (2 + � + ���)�|��|2����� � (��:���) + |��|2����� � (��:���) + |��||��|2�������� � (����:�)� Let |��| = 2��� and |��| = 2���.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' by setting the above term equal to �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' with a straightforward simplification,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' we have: �� + ��� = �� �(��: ���) + log� �� − (1 + �)� 2 + � + ��� � �� + ��� = �� �(��: ���) + log� �� − (1 + �)� 2 + � + ��� � �� + ��� + �� + ��� = �� �(����: �) + log� �� − (1 + �)� 2 + � + ��� � The global maximum of the above expression with respect to � occurs at � = � �: �� + ��� = �� �(��: ���) − log� �4� ��� (34) �� + ��� = �� �(��: ���) − log� �4� ��� (35) �� + ��� + �� + ��� = �� �(����: �) − log� �4� ��� (36) and for such a �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' we have: �� �����,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ���� ≠ (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� ≤ � + 2� (37) Now,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' we turn our attention to the secrecy criterion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Using Lemma 1, we have: ��� ≤ ���� �����(��: �)� + log 3 ��� − 1 4 log �� (38) ��� ≤ ���� �����(��: ���)� + log 3 ��� − 1 4 log �� + �(1) (39) Substituting (38) and (39) in (34)-(36) completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Appendix C: (Proof of Theorem 2) The proof uses two successive position-based decoders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The first decoder tries to decode the first message ��, and the second decoder tries to decode the second message �� given the true decoded ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This means that if the first decoder fails, the second decoder fails too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This decoding order can be shown as �� → ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Constructing the first position-based decoder is the same as that presented in [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' To decode ��, Bob performs his second position-based decoder conditioned on ��, which works for all �� ∈ ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' It should be noted that, the feeding state of the second decoder differs from the main state of the channel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Alice, Bob, and Eve are allowed to pre-share some quantum state as randomness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, Alice has access to two sources of uniform junk randomness ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' � ∈ {1,2} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The pre-shared randomness is as follows: � ������������� ⊗|ℳ�||��| ⊗|ℳ�||��| ≔ �� �(��)|��⟩⟨��|�� �� ⊗ |��⟩⟨��|��� �� �(��)|��⟩⟨��|�� �� ⊗ |��⟩⟨��|���� ⊗|ℳ�||��| � ⊗|ℳ�||��| (40) The arguments connected to the decoding process for �� are listed as follows: The probability of error for decoding ��: ��� = ����� ≠ ��� ≔ 1 |ℳ�| � 1 2 |ℳ�| ���� �����→��� �� ����⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)� − |��⟩⟨��|��� ⊗ ���� � ≤ �� + ��� (41) where ����→��� �� ����⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)� is decoding map for ��: ����→��� �� ����⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)� ≔ � � �� �Λ��|ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ���⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)� |ℳ�| ���� |��| ���� ⊗ �Λ��|ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ���⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)�Λ��|ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� �� �Λ��|ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ���⊗|ℳ�||��|� (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)� Λ��|ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� is a pretty good measurement (POVM) for �� ∈ �1: |ℳ�|�: Λ�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ≔ � � � Γ�� |ℳ�||��|� ���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� |ℳ�| �� ��� |��| �� ��� � �� � � Γ�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � � � Γ�� |ℳ�||��|� ���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� |ℳ�| �� ��� |��| �� ��� � �� � � where Γ�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� is the element of the first POVM: Γ�� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ≔ ��� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�) ⊗ … ⊗ ��� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|) ⊗ … ⊗ ���� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ⊗ … ⊗ ��� (|ℳ�|,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|) and ���� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� is a test operator in order to discriminate between two hypotheses ����,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' and ��� ⊗ ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Also, it is obvious that to decode ��, it does not matter for the second position-based decoder, which copy is selected by Alice among |ℳ�||��| copies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' We face a hypothesis testing problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Null hypothesis is ���� and alternative hypothesis is ��� ⊗ ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Therefore the probability of success in guessing null and alternative hypotheses are ������� ����� and �������� − ����� ���� ⊗ ����.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The rest of the decoding process for �� is analogous to [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Therefore, we have: �� ≤ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� − ����� √�����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� − log 4�� �� � − 2 log 1 �� (42) Now, we turn our attention to decoding the second message.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' As mentioned before, the channel state changes after the first measurement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' There is a detailed discussion in [32].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Let ������(�����)⊗|ℳ�||��|�� (�1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(�2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�2) denote the disturbed state after applying the first measurement (POVM): ������(�����)⊗|ℳ�||��|�� (�1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(�2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�2) ≔ � ���(��)|�1⟩⟨�1|�1 ⊗ |�1⟩⟨�1|��� �� ⊗ ������ ��(�1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�1) ⊗ … ⊗ �������� ��(�1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(�2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�2) ⊗ … ⊗ ������ ��(�1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�1),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='(|ℳ�|,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|) Also,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Bob’s second POVM is as follows: Λ���� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ≔ � � � λ���� |ℳ�||��|� ���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� |ℳ�| �� ��� |��| �� ��� � �� � � λ���� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � � � λ���� |ℳ�||��|� ���,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� |ℳ�| �� ��� |��| �� ��� � �� � � λ���� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� is the element of the second POVM: λ���� |ℳ�||��|� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ≔ |�1⟩⟨�1|�1 ⊗ ��� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�) ⊗ … ⊗ ��� (�,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|) ⊗ … ⊗ ���� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ⊗ … ⊗ ��� (|ℳ�|,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='|��|) ���� ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� is a binary test operator to discriminate between two hypotheses ���� �� and ��� �� ⊗ �� �� with an error of �� − ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=', ����������� �� � ≥ 1 − (�� − ��) �� ∈ (0,1), �� ∈ (0, ��) In other words, Bob has to be able to discriminate between the following states: � ���(��)|�1⟩⟨�1|�1 ⊗ ���� �� �� � ���(��)|�1⟩⟨�1|�1 ⊗ ��� �� ⊗ �� �� �� Similar to what mentioned in [20], and [27], we have the following rate: �� ≤ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ����� √�����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − log 4�� �� � − 2 log 1 �� (43) The probability of error for �� is as follows: ��� = ����� ≠ ��� ≔ 1 |ℳ�| � 1 2 |ℳ�| ���� ��������→��� �� �� ������������⊗|ℳ�||��|�� (��,��),(��,��) �|��⟩⟨��|��� ⊗ ��������⊗|ℳ�||��|�� � ≤ 2��� + ���� + ��� � (44) Also, the error probability exponents stated in (41) and (44) are proved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' See [20, 27].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' This process can be repeated for another decoding order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In other words, we can first decode �� , and then decode �� ( �� → �� ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Then taking the intersection of the regions resulting from both orders, we give: �� ≤ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ����� √�����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �)� − log 4�� �� � − 2 log 1 �� �� ≤ �� �����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − ����� √�����(��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' �|��)� − log 4�� �� � − 2 log 1 �� This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Appendix D: (Proof of Theorem 3) The proof uses superposition coding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Assume that the first receiver �� has a better reception signal than the second receiver ��.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' In this setting, Alice is able to encode a further message superimposed on top of the common message.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Using the successive decoding can be helpful.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Codebook generation: Randomly and independently generate 2�� sequence �(��) according to the distribution ��(�).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' For each sequence �(��), randomly and conditionally independently generate 2�� sequence �(��, ��) according to the distribution ��|�����(��)� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' The �� ’s state can be calculated by tracing out �� from (13): ����� = � ��(�)��|�(�|�) �,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='� |�⟩⟨�|� ⊗ |�⟩⟨�|� ⊗ �� �� Similar to what mentioned for Theorem 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' we construct the POVM for the first receiver as: ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� ≔ � � � Γ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� |ℳ�| ��� �� |ℳ�| ����� � �� � � Γ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�� � � � ��� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��� |ℳ�| ��� �� |ℳ�| ����� � �� � � Also,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' the POVM for the second receiver can be constructed as follows: Λ�� ≔ � � λ��� |ℳ�| ����� � �� � � λ�� � � λ��� |ℳ�| ����� � �� � � Consider the probability of error for ��: ��� = ������,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ���� ≠ (��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ��)� ≔ 1 |ℳ�||ℳ�| � � �� ��� �� �� − Λ��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='�����(��,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��) �� � and for ��: ��� = ����� ≠ ��� ≔ 1 |ℳ�| � �� ��� − Λ�����(��) �� � �� By a straightforward calculation analogous to [3] for i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' case and in [29] (to calculate one-shot Marton inner bound for QBC), the above error probability exponents can be calculated as follows: ��� + ��� ≤ 2 ��� � ��;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' ������������ � + 2��� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)������� � + 2��� � (�;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='��)������� � + �(�) This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' REFERENCES [1] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Winter, “The capacity of the quantum multiple-access channel,” IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 47, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 7, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 3059–3065, July 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [2] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Yard, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Hayden, and I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Devetak.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' "Quantum broadcast channels," IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' on Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 57, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 10, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 7147-7162, October 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [3] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Savov, “Network information theory for classical-quantum channels,” Ph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' dissertation, McGill University, Montreal, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [4] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' El Gamal and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Kim, “Lecture notes on network information theory,” January 2010, available online at http://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='org/abs/1001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='3404 [5] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Cai, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Winter, and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Yeung.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' “Quantum Privacy and Quantum Wiretap Channels,” Problems of Information Transmission, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 40, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 4, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 318-336, 2004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [6] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Devetak, “The Private Classical Capacity and Quantum Capacity of a Quantum Channel,” IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 51, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 44-55, January 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [7] O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fawzi, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Hayden, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Savov, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sen, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wilde, “Classical communication over a quantum interference channel,” IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' on Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 58, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 6, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 3670-3691, June 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [8] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Aghaee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Akhbari, “One-Shot Achievable Secrecy Rate Regions for Quantum Interference Wiretap Channel ,” The ISC International Journal of Information Security (IseCure), vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='14, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 3, pp 71-80, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [9] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Aghaee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Akhbari, “Classical-Quantum Multiple Access Wiretap Channel,” in Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=" 16th International ISC Conference on Information Security and Cryptology (ISCISC'19), Mashhad, Iran, August 2019." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [10] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Aghaee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=" Akhbari, “Private Classical Information over a Quantum Multiple Access Channel: One-shot Secrecy Rate Region,” in Proc 10th International Symposium on Telecommunications (IST'2020), Iran, 2020." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [11] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Aghaee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Akhbari, “Classical-Quantum Multiple Access Channel with Secrecy Constraint: One-shot Rate Region,” International Journal of Information and Communication Technology Research (IJICTR), vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 12, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 2, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1-10, Spring 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [12] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Aghaee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Akhbari, “Classical-Quantum Multiple Access Wiretap Channel with Common Message: One-shot Rate Region,” in Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=" 11th Conference on Information and Knowledge Technology (IKT'2020), Iran, 2020." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [13] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Aghaee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Akhbari, “Entanglement assisted Classical-Quantum Multiple Access Wiretap Channel: One-shot Achievable Rate Region,” in Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 30th International Conference on Electrical Engineering (ICEE2022), Iran, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [14] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sen, “Unions, intersections and a one-shot quantum joint typicality lemma,” Sadhana, 46(1):1–44, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [15] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wyner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' “The wire-tap channel,” Bell Syst.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Tech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=', vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 54, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 8, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 2–10, October 1975.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [16] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Chakraborty, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Nema, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sen, “One-shot inner bounds for sending private classical information over a quantum MAC,” 2021 IEEE Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory Workshop (ITW), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1-6, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [17] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Qi, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wang, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wilde, “Applications of position-based coding to classical communication over quantum channels,” https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='org/abs/1704.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='01361, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [18] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Anshu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' "One-shot protocols for communication over quantum networks: Achievability and limitations," PhD diss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=', National University of Singapore (Singapore), 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [19] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Anshu, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Jain and N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Warsi, "A Generalized Quantum Slepian– Wolf," in IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' on Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 64, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 3, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1436-1453, March 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [20] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wilde, “Position-based coding and convex splitting for private communication over quantum channels,” Quantum Information Processing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 16(10):264, October 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [21] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wang and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Renner, “One-shot classical-quantum capacity and hypothesis testing,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=', vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 108, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 20, p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 200501, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [22] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Umegaki, “Conditional expectations in an operator algebra IV (entropy and information),” Kodai Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Rep.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=', vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 14, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 59-85, 1962.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [23] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Berta, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Christandl and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Renner, “The quantum reverse Shannon theorem based on one-shot information theory,” Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=', vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 306, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 3, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 579-615, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [24] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Salek, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Anshu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' -H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Hsieh, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Jain and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fonollosa, "One-Shot Capacity Bounds on the Simultaneous Transmission of Classical and Quantum Information," IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' on Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 66, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 4, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 2141-2164, April 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [25] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Ciganovi´c, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Beaudry and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Renner, “Smooth max-information as one-shot generalization for mutual information,” IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 60, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1537-1581, March 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [26] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Hayashi, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Nagaoka, “General formulas for capacity of classical- quantum channels,” IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 49, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1753–1768, July 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [27] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Salek, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Anshu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' -H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Hsieh, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Jain and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Fonollosa, "One-shot Capacity Bounds on the Simultaneous Transmission of Public and Private Information Over Quantum Channels," 2018 IEEE International Symposium on Information Theory (ISIT), 2018, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 296-300.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [28] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wilde, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Tomamichel and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Berta, "Converse Bounds for Private Communication Over Quantum Channels," in IEEE Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' on Inf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Theory, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 63, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 3, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 1792-1817, March 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [29] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Sen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' “Inner bounds via simultaneous decoding in quantum network information theory,” Sādhanā 46, 18 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='1007/s12046-020-01517-9 [30] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wilde, Quantum Information Theory, Cambridge Univ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Press, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [31] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Li, “Discriminating quantum states: the multiple Chernoff distance,” The Annals of Statistics, 44(4):1661{1679, August 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' arXiv:1508.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content='06624.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' [32] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' Wilde, “Sequential decoding of a general classical-quantum channel,” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 469(2157), 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
+page_content=' 20130259.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/StE0T4oBgHgl3EQflAEY/content/2301.02479v1.pdf'}
diff --git a/T9E0T4oBgHgl3EQflQGf/content/tmp_files/2301.02484v1.pdf.txt b/T9E0T4oBgHgl3EQflQGf/content/tmp_files/2301.02484v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..ab15de5a3a7e8000ff0c90cef7264b0eac51f74a
--- /dev/null
+++ b/T9E0T4oBgHgl3EQflQGf/content/tmp_files/2301.02484v1.pdf.txt
@@ -0,0 +1,2969 @@
+1
+Graph-Collaborated Auto-Encoder Hashing for
+Multi-view Binary Clustering
+Huibing Wang, Mingze Yao, Guangqi Jiang, Zetian Mi, Xianping Fu
+Abstract—Unsupervised
+hashing
+methods
+have
+attracted
+widespread attention with the explosive growth of large-scale
+data, which can greatly reduce storage and computation by
+learning compact binary codes. Existing unsupervised hashing
+methods attempt to exploit the valuable information from sam-
+ples, which fails to take the local geometric structure of unlabeled
+samples into consideration. Moreover, hashing based on auto-
+encoders aims to minimize the reconstruction loss between the
+input data and binary codes, which ignores the potential consis-
+tency and complementarity of multiple sources data. To address
+the above issues, we propose a hashing algorithm based on auto-
+encoders for multi-view binary clustering, which dynamically
+learns affinity graphs with low-rank constraints and adopts col-
+laboratively learning between auto-encoders and affinity graphs
+to learn a unified binary code, called Graph-Collaborated Auto-
+Encoder Hashing for Multi-view Binary Clustering (GCAE).
+Specifically, we propose a multi-view affinity graphs learning
+model with low-rank constraint, which can mine the underlying
+geometric information from multi-view data. Then, we design
+an encoder-decoder paradigm to collaborate the multiple affinity
+graphs, which can learn a unified binary code effectively. Notably,
+we impose the decorrelation and code balance constraints on
+binary codes to reduce the quantization errors. Finally, we
+utilize an alternating iterative optimization scheme to obtain the
+multi-view clustering results. Extensive experimental results on
+5 public datasets are provided to reveal the effectiveness of the
+algorithm and its superior performance over other state-of-the-
+art alternatives.
+Index Terms—Graph-collaborated, Auto-encoder, Multi-view
+clustering, Binary code
+I. INTRODUCTION
+W
+ITH the development of information digitization [1],
+[2], [3] and computer technology, researchers have pro-
+posed a large number of feature extraction methods to extract
+features from multiple views of the same sample [4], [5], [6],
+[7]. For example, an image can be extracted as different feature
+representations by multiple descriptors, i.e., LBP [8], Gabor
+[9], HOG [10] and SIFT [11]. However, these multi-view data
+extracted from different feature descriptors have properties that
+are large-scale and heterogeneous, which cry out for reliable
+mining methods to explore the discriminative information
+from multiple views. In order to effectively process the large-
+scale data, most existing researches introduce hash methods
+H. Wang, M. Yao, M. Ze and X. Fu are with College of Information
+Science and Technology, Dalian Maritime University, Liaoning, 116026,
+China, e-mail: (huibing.wang@dlmu.edu.cn; ymz0284@dlmu.edu.cn; mize-
+tian@dlmu.edu.cn; fxp@dlmu.edu.cn). Both Huibing Wang and Mingze Yao
+are first authors.
+G. Jiang is with School of Computer Science and Artifical Intelli-
+gence, Changzhou University, Jiangsu, 213164, China, e-mail: (guangqi-
+jiang@cczu.edu.cn).
+Mingze Yao, Guangqi Jiang and Xianping Fu are corresponding authors.
+due to its fast running speed and economical storage cost.
+Specifically, hash methods encode the large-scale data by a
+set of compact binary codes in a low-dimensional Hamming
+space. Therefore, existing hash algorithms have been widely
+applied to various large-scale visual application tasks, such as
+cross-modal retrieval [12], object re-identification [13], image
+detection [14] and multi-view learning [15], [16], [17], [18]
+etc.
+Considering the effectiveness of binary codes for various
+vision tasks with large-scale data, several methods have been
+proposed to explore the more discriminative binary code
+representation. Over the past few decades, several supervised
+hashing methods have been proposed, such as Supervised
+Discrete Hashing (SDH) [19], Strongly Constrained Discrete
+Hashing (SCDH) [20] and Fast Discriminative Discrete Hash-
+ing (FDDH) [21]. Note that while these aforementioned ap-
+proaches have achieved great performance with hashing, most
+of them deeply depend on the manual labels, which is time-
+consuming and less effective process the large-scale unlabeled
+data. Therefore, some unsupervised hashing methods have
+been proposed to deal with the unlabeled problem. The typical
+unsupervised hashing is Locality-Sensitive Hashing (LSH)
+[22] which adopts random projections to generate discrete
+binary codes. Based on LSH, Spectral Hashing (SH) [23],
+Discrete Graph Hashing (DGH) [24] and Scalable Graph
+Hashing (SGH) [25] have been proposed to explore similar
+information from the large-scale data. Even though the above
+methods have effectively learned compact binary codes in an
+unsupervised manner, most existing hashing methods usually
+utilize the data from single source. For multi-view data,
+these hashing methods are difficult to uncover the multi-
+view information holistically and ignore the consistent and
+complementary information from different views.
+Compared with the data from a single source, multi-view
+data usually contain more compatible and complementary
+information hidden in different views, which are extracted
+from same samples. Therefore, multi-view clustering methods
+have been proposed to explore the latent structure of different
+views and integrate complementary information from multi-
+view data. Kumar et al. [26] introduced a co-regularized model
+to complete spectral clustering with a centroid-based algo-
+rithm and pairwise algorithm which can mine the underlying
+structure from original data. Zhan et al. [27] proposed a
+graph-learning method with the rank constraint to integrate
+different graphs into a global graph for multi-view clustering
+tasks. Wang et al. [28] proposed a multi-graph laplacian
+regularized LRR model, which can separately impose a low-
+rank constraint on each graph to achieve agreeable results.
+arXiv:2301.02484v1 [cs.CV] 6 Jan 2023
+
+2
+Besides, Wang et al. [29] performed reinforcement learning
+on the graph of each view and the unified graph of all views
+by considering the weights of different views. Xiao et al.
+[30] proposed a graph-based multi-view clustering framework
+with knowledge elements, which can combine knowledge and
+language for clustering. Moreover, Shi et al. [31] proposed
+a common joint graph learning strategy, which utilizes non-
+negative constraint to fully explore the structure information
+from multi-view data. This strategy aims to directly obtain
+cluster results and avoid post-processing. The above methods
+mostly measure the distance between features in Euclidean
+space, while they still need a high computational cost and low
+efficiency for processing large-scale data.
+Some researchers proposed multi-view hash methods to
+learn compact binary codes and utilize efficient XOR operation
+[32], which can improve the speed and accuracy of the
+algorithm. Jin et al. [33] proposed a binary function clustering
+scheme that captures the function semantics as semantic hash-
+ing to quickly cluster the high degree of similarity samples.
+Tian et al. [34] provided a variant of the LRR [35] model
+to recover the latent structure of original data, which can
+effectively learn similarity graphs for binary code learning.
+Wang et al. [36] utilized l2,1-norm to learn compact binary
+codes, which can improve the robustness of the model. Shen
+et al. [37] constructed a novel semantic-rebased model, which
+adopted a sparse graph setting and rebased the similarity
+graph. Notably, most related hashing works focus on re-
+trieval tasks, which ignore the complementary information and
+underlying cluster structure from multi-view data. Recently,
+several hashing algorithms have been proposed to solve large-
+scale image clustering problems. Wang et al. [38] provided a
+cluster-wise unsupervised hashing framework, which projects
+the multi-view original data into latent low-dimensional space
+to learn cluster centroid for searching. Zhang et al. [39]
+explored a highly-economized algorithm for image clustering,
+which jointly learned binary representation and binary cluster
+results. Even though the above methods can process large-
+scale data effectively and achieve great performance, most
+of them heavily rely on affinity graphs from original data
+directly and fail to mine the local structures. Meanwhile, some
+studies simplified the optimization problem by relaxing binary
+constraints, which may cause quantization errors. Therefore,
+it is essential to compose an effective graph collaboration
+framework to explore the local geometric information from
+multiple views and utilize suitable binary constraints.
+To address the above limitations, this paper proposes a
+novel method, termed as Graph-Collaborated Auto-Encoder
+Hashing for Multi-view Binary Clustering (GCAE). GCAE
+constructs auto-encoders to learn binary codes for processing
+multi-view data, which emphasizes collaboratively learning
+between affinity graphs and auto-encoders to learn a uni-
+fied binary codes for multi-view clustering. Firstly, GCAE
+constructs affinity graphs from each view by imposing a
+low-rank constraint on the original data, which can preserve
+essential information and the latent structure from the multi-
+view data. Secondly, to effectively explore the compatible
+and complementary information from multi-view data, GCAE
+adopts auto-encoders to collaborate multiple affinity graphs,
+which aim to learn unified binary codes for clustering and
+preserve the discrete binary constraint. Subsequently, GCAE
+utilizes the matrix factorization strategy to directly obtain
+cluster results without post-processing, which can avoid error
+accumulation. Finally, an alternating iterative optimization
+strategy is adopted to update each variable of the objective
+function. The whole model of GCAE has been shown in
+Fig. 1. The major contributions of the proposed method are
+summarized as follows:
+• We propose Graph-Collaborated Auto-Encoder Hashing
+for Multi-view Binary Clustering (GCAE), which utilizes
+affinity graphs and auto-encoders collaboratively to learn
+compact binary codes for multi-view clustering.
+• In particular, GCAE imposes the low-rank constraint
+on graphs to mine essential information effectively and
+utilizes auto-encoders to collaborate multiple graphs for
+learning unified binary codes, which can explore comple-
+mentary information from multi-view data and guide the
+learning of binary codes. Besides, our proposed GCAE
+directly obtain cluster results to avoid the accumulation
+of errors caused by post-processing.
+The remainder of the paper is outlined as follows. Section
+2 introduces the related work. Section 3 presents the proposed
+GCAE model and the optimization process. Extensive experi-
+ments including complexity analysis and convergence analysis
+are conducted to verify our proposed model in Section 4.
+Finally, Section 5 concludes this paper.
+II. RELATED WORK
+In this section, we briefly review the related studies about
+graph-based multi-view clustering and multi-view hashing
+methods with graphs.
+Graph-based multi-view clustering methods mostly aim to
+integrate information from multiple views and calculate simi-
+larity graphs in Euclidean distance for clustering. For example,
+Nie et al. [40] proposed a framework based on standard
+spectral learning which learns weights for multiple graphs
+automatically without introducing additive parameters. Hou et
+al. [41] presented an automatic method to learn a common
+similarity graph to characterize the structures across different
+views and tune balance weights. However, the above methods
+require an additional clustering step to obtains the final clusters
+by utilizing K-means [42]. In order to avoid the impact of
+post-processing for obtain the cluster results, Wang et al. [29]
+proposed a model which can produce clusters directly without
+post-processing for clustering and construct each view graph
+and fusion graph simultaneously. Besides, Zhang et al. [43]
+utilize Hadamard product to integrate multiple graphs into a
+global graph which can recover the graph structure effectively.
+Shi et al. [44] proposed a unified framework for jointly learn-
+ing multiple similarity graphs and spectral embedding, which
+can obtain cluster results in a unified framework. Although the
+above methods achieved great clustering results, they directly
+build graphs by original data, which may make the learned
+similarity graph inaccurate and ignore the underlying cluster
+structure of multi-view data. Meanwhile, the real-value multi-
+view data requires high computational costs for generating
+
+3
+Multi-view Data
+View 1
+View 2
+.
+.
+.
+View M
+Kernelized Representation of Multiple Views
+Clustering Result
+×
+Nonlinear Kernel Mapping
+1
+Clustering Centroids
+ Indicator Vectors
+Affinity Graphs
+Expend to Multiple Views
+Z1
+Z2
+.
+.
+.
+.
+.
+.
+Projection Matrix
+Feature
+Encoder
+Feature
+Encoder
+Inverse Projection Matrix
+Feature
+Decoder
+Feature
+Decoder
+Projection Matrix
+Feature
+Encoder
+Inverse Projection Matrix
+Feature
+Decoder
+Common Binary Codes Matrix
+-1
+1
+1
+1
+1
+1
+-1
+-1
+-1
+1
+1
+1
+-1
+-1
+1
+-1
+-1
+1
+1
+1
+1
+1
+-1
+-1
+-1
+1
+1
+1
+-1
+-1
+1
+-1
+-1
+1
+1
+1
+1
+-1
+1
+1
+-1
+-1
+1
+-1
+1
+-1
+1
+-1
+-1
+1
+1
+1
+1
+-1
+1
+1
+-1
+-1
+1
+-1
+1
+-1
+1
+-1
+1
+1
+1
+-1
+1
+1
+-1
+-1
+-1
+1
+-1
+1
+-1
+1
+-1
+-1
+1
+1
+1
+-1
+1
+1
+-1
+-1
+-1
+1
+-1
+1
+-1
+1
+-1
+-1
+1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+-1
+1
+-1
+1
+1
+1
+1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+-1
+1
+-1
+1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+1
+-1
+1
+1
+1
+1
+1
+-1
+1
+-1
+-1
+-1
+1
+1
+-1
+1
+1
+1
+-1
+1
+1
+1
+1
+1
+-1
+1
+-1
+-1
+-1
+1
+1
+-1
+1
+1
+-1
+1
+1
+1
+1
+1
+-1
+-1
+-1
+1
+1
+1
+-1
+-1
+1
+-1
+-1
+1
+1
+1
+1
+-1
+1
+1
+-1
+-1
+1
+-1
+1
+-1
+1
+-1
+1
+1
+1
+-1
+1
+1
+-1
+-1
+-1
+1
+-1
+1
+-1
+1
+-1
+-1
+1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+-1
+1
+-1
+1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+-1
+1
+-1
+1
+1
+1
+1
+1
+-1
+1
+-1
+-1
+-1
+1
+1
+-1
+1
+1
+Collaborative Learning
+Collaborative Learning
+Binary Codes Learning by Auto-encoders
+1
+1
+1
+1
+1
+-1
+-1
+-1
+1
+1
+1
+1
+1
+-1
+-1
+-1
+1
+1
+1
+-1
+1
+-1
+1
+1
+1
+1
+-1
+-1
+1
+-1
+-1
+-1
+1
+1
+-1
+-1
+1
+-1
+-1
+-1
+-1
+1
+-1
+-1
+1
+-1
+-1
+-1
+-1
+1
+-1
+-1
+1
+-1
+-1
+-1
+1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+1
+1
+1
+1
+-1
+-1
+-1
+1
+1
+1
+-1
+1
+-1
+1
+1
+1
+1
+-1
+-1
+1
+-1
+-1
+-1
+-1
+1
+-1
+-1
+1
+-1
+-1
+-1
+1
+1
+-1
+1
+1
+-1
+1
+-1
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+0
+0
+0
+1
+0
+0
+1
+0
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+0
+0
+0
+1
+0
+0
+1
+0
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+1
+0
+0
+0
+0
+0
+One Step Clustering
+Binary Matrix Factorization
+View 1
+View 1
+Kernelized Representation
+ZM
+Fig. 1: The whole model of Graph-Collaborated Auto-Encoder Hashing for Multi-view Binary Clustering (GCAE), which
+learns affinity graphs by low-rank constraint to integrate specific information from multiple views and adopts auto-encoder to
+generate unified binary code for clustering.
+graphs with large-scale multi-view data. Some researchers
+explore hash methods which utilize binary codes rather than
+real-value features. Therefore, multi-view hashing methods
+are widely used to integrate large-scale data with the recent
+advances.
+Existing multi-view hashing methods with graphs can be
+roughly divided into supervised and unsupervised methods
+based on the usage of semantic presentation (e.g., class
+labels). For supervised methods, Jin et al. [45] construct a
+semantic graph by jointly taking the semantic presentation
+and the local similarity structure into consideration. Guan et
+al. [46] proposed a supervised learning model to construct
+similarity graphs that capture the intrinsic manifold structure
+from semantic supervision. Despite these supervised methods
+can achieve great performance, it is not practical to label infor-
+mation manually for large-scale data. Therefore, Liu et al. [24]
+utilized anchor graphs to capture the latent structure inherent in
+a given massive dataset and calculated hash codes in Hamming
+space. Xiang et al. [47] proposed a novel hashing method
+that adopted a quantization regularization term to reduce the
+distortion error during constructing similarity graphs. Besides,
+Fang et al. [48] jointly learned the intra-modal similarity
+graph and reconstructed the similarity graph by symmetric
+nonnegative matrix factorization and then utilized binary code
+inner product to learn binary codes. The above approaches
+have obtained good results, but relaxing the discrete constraint
+on binary codes will cause quantization errors. Meanwhile, for
+the multi-view clustering tasks, existing multi-view hashing
+methods have not effectively mined the underlying cluster
+structure from different views.
+III. GRAPH-COLLABORATED AUTO-ENCODER HASHING
+In this paper, boldface lowercase letters and boldface up-
+percase letters are used to denote the vectors and matrices,
+receptively, e.g., x and X. For any matrix X, xij means
+Table I: The Descriptions of Some Important Formula Sym-
+bols.
+Scalar
+N
+N ∈ R
+Vector
+a
+a ∈ RN
+ai
+i-th row
+av
+i
+i-th row with v-th view
+1
+all-1 vector
+0
+all-0 vector
+||a||2
+l2 norm as
+�� N
+i=1a2
+i
+Martix
+Av
+Av ∈ RN×dv
+Ai,j
+(i, j)-th element
+I
+the identity matrix
+tr(A)
+the trace of matrix A
+AT
+the transpose of matrix A
+σ
+σ = diag(S)
+||A||F
+Frobenius norm as
+��
+i,jA2
+i,j, ∥σ∥2
+||A||∗
+nuclear norm as tr(
+√
+ATA)
+φ(Av)
+kernelized representation of (Av)
+the (i, j)-element of X. Besides some important mathematic
+notations are listed in Table I. For given multi-data sets,
+X = {X1, X2, ..., Xv} where Xv ∈ RN×dv is the v-th view
+data. dv denotes the feature dimension and N is the number
+of sample.
+We propose a novel model of graph-collaborated auto-
+encoder hashing for clustering. Specifically, we jointly learned
+the affinity graphs from each view and constructed auto-
+encoders for the unified hash code learning which can min-
+imize reconstruction loss between affinity graphs and hash
+code. Besides, our method utilized matrix factorization strat-
+egy to obtain cluster results.
+
+r
+G me4
+A. Multi-view Affinity Graph Learning
+For given data matrix X = {X1, X2, ..., Xv}, where Xv
+presents the v-th view data matrix. Different from existing
+methods that directly utilize original multi-view data for gen-
+erating multiple graphs. GCAE firstly adopts nonlinear kernel
+mapping for each view, which aims to unify the dimension of
+multi-view data so that we can learn the underlying geometric
+structure from data. Inspired by [32], we employ the nonlinear
+RBF kernel mapping method for each view as follows:
+φ(Xv) = [exp(−∥xv
+1 − av
+1∥2/η), ..., exp(−∥xv
+N − av
+t ∥2/η)]T
+(1)
+where η is the kernel width, φ(Xv) ∈ RN×t indicates the t-
+dimensional nonlinear mapping from the v-th view. Besides,
+to ensure that the original data structure is not destroyed after
+nonlinear projection, we randomly select t anchor samples av
+t
+from the v-th view.
+After φ(Xv) yielded by Eq. 1, the affinity graph Zv is
+constructed by processing each view of φ(Xv). Above this, we
+propose a new variant of the LRR [49] model that can make
+the learned affinity graph retain the important information of
+the original data instead of noise and redundant information.
+Then, we give the following model:
+min
+Zv
+M
+�
+v=1
+∥φ(Xv) − Zvφ(Xv)∥2
+F + ∥Zv∥∗
+(2)
+where Zv ∈ RN×N represents the learned affinity graph
+for v-th view data. ||Zv||∗ indicates the nuclear norm which
+calculates the sum of singular values for low-rank constraint.
+Minimizing the objective function in Eq. 2 aims to preserve
+important information while exploring the low-rank structure
+from the original data.
+B. Graph-collaborated Auto-Encoders Hashing for Clustering
+In order to encourage collaborative learning of affinity
+graphs from different views, we proposed an auto-encoders
+hashing by the multi-graphs model which can effectively
+mine the consistency and complementarity information and
+generate binary code for clustering. Different from the existing
+hash methods, the auto-encoders hashing model preserves the
+discrete constraint and the uncorrelated constraint on binary
+bits.
+min
+Wv,B,Zv
+M
+�
+v=1
+(∥WvZv − B∥2
+F + ∥Zv − WvT B∥2
+F )
+s.t.
+WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI
+(3)
+Eq. 3 aims to minimize the difference between low-rank
+affinity graphs mapping and discrete hash code. b represents
+binary bits. The decorrelation and code balance constraints
+B ∈ {−1, +1}b×N, BBT = NI are imposed to generate
+mutually independent binary codes and reduce quantization
+errors. Wv presents the projection matrix. For simplicity, we
+utilize the transpose matrix of Wv to replace the inverse map-
+ping matrix in auto-encoder, and add a constraint WvWvT =
+I on projection matrix. Meanwhile, the regularization term
+||Wv||2
+F is the constant due to ||Wv||2
+F = tr(WvWvT ) =
+tr(I) = const. Finally, we construct a matrix factorization
+model with the generated unified binary codes, which can
+avoid the suboptimal results caused by the two-step clustering
+methods. Above all, Eq. 3 can be reformulated as follows:
+min
+Zv,Wv,B,Q,H
+M
+�
+v=1
+(∥WvZv − B∥2
+F + ∥Zv − WvT B∥2
+F ))
++ λ ∥B − QH∥2
+F
+s.t.
+WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI,
+QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N,
+c
+�
+i=1
+his = 1
+(4)
+where λ is the regularization parameter and c is the number of
+clusters. Q and H represent the clustering centroids and cluster
+indicator matrix, respectively. Meanwhile, in order to generate
+the efficient binary code and maximize the information of each
+bits for clustering, we add the balance constraint on clustering
+centroids Q, which can produce efficient code and make our
+model to adapt binary clustering task.
+C. Overall Objective Function of GCAE
+In order to collaborating affinity graphs and auto-encoders
+to learn a unified binary code for clustering, we summarize
+the above graph-collaborated auto-encoders hashing and the
+low-rank affinity graphs learning, which are both crucial to
+multi-view clustering. The above consideration can be fulfilled
+as follows:
+min L(Zv, Wv, B, Q, H, pv)
+=
+M
+�
+v=1
+∥φ(Xv) − Zvφ(Xv)∥2
+F + ∥Zv∥∗
+�
+��
+�
+Multi−view Affinity Graph Learning
++
+M
+�
+v=1
+(pv)k(∥WvZv−B∥2
+F +∥Zv−WvTB∥2
+F )+λ∥B−QH∥2
+F
+�
+��
+�
+Auto−Encoders Hashing by Multi−Graphs for Clustering
+s.t.
+WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI,
+QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N,
+M
+�
+v=1
+pv = 1, pv > 0,
+c
+�
+i=1
+his = 1
+(5)
+where pv indicates the normalized weighting coefficient,
+which aims to balance the affinity graphs from each view ac-
+cording to the contribution. Besides. we also add the constraint
+pv > 0 to ensure nonnegative vector. Above all, we summarize
+multi-view affinity graph learning and auto-encoders hashing
+by multi-graphs in a unified framework. The proposed GCAE
+model learns the low-rank affinity graph from each view,
+which can effectively preserve important information from
+original data to improve the quality of affinity graphs. And
+then, the proposed model utilizes binary matrix factorization
+model for the unified binary codes, which can effectively gen-
+erate cluster results in a one-step model. By jointly optimizing
+
+5
+the above equation, we obtain the graph-collaborated binary
+code representation to obtain the cluster results. We propose
+a novel optimization algorithm for the objective function Eq.
+5 in the following section.
+D. Optimization Process for GCAE
+In this section, we describe the optimization process for
+GCAE in detail. It is obvious that Eq. 5 is a nonconvex
+optimization problem that can not directly obtain the holistic
+optimal solution in Eq. 5. We develop an auxiliary matrices
+strategy, which separates the problem into two sub-problems
+include low-rank affinity graphs learning and auto-encoders
+hash for obtaining optimized cluster results. Based on the aux-
+iliary matrices strategy, we introduce two auxiliary matrices to
+relax the nuclear norm and control the rank of affinity graphs.
+And then the proposed auto-encoders utilize the affinity graphs
+from each view to generate binary code for clustering. Finally,
+we develop an iterative alternative strategy, which alternatively
+updates each variable when fixing others.
+The proposed auxiliary matrices strategy introduces the
+multiplication of two auxiliary matrices Fv and Gv (i.e.,
+Zv = FvGvT ) to replace the low-rank affinity graph Zv, and
+then problem Eq. 2 can be converted as:
+min
+Fv,Gv
+���φ(Xv) − FvGvT φ(Xv)
+���
+2
+F
+(6)
+where Fv ∈ RN×r and Gv ∈ RN×r. r is a parameter
+that we use to relax nuclear norm and control the rank of
+Zv. And this strategy is based on the fact that rank(Zv) =
+rank(FvGvT ) ≤ min(rank(Fv), rank(Gv)) ≤ r (generally
+r ≪ N). The updating process of Fv and Gv are shown as
+follows:
+Updating
+Fv: Based on the operational rules of matrix
+trace, Eq. 6 can be unfolded as:
+O(Fv) =
+tr((φ(Xv) − FvGvT φ(Xv))(φ(Xv) − FvGvT φ(Xv))T )
+(7)
+then we calculate Fv iteratively by setting the derivation
+∂O(Fv)
+∂Fv
+= 0:
+∂O(Fv)
+∂Fv
+=2FvGvT φ(Xv)φ(Xv)T Gv
+− 2φ(Xv)φ(Xv)T Gv = 0
+(8)
+Obviously, the close solution of Fv is written as:
+Fv = φ(Xv)φ(Xv)T Gv(GvT φ(Xv)φ(Xv)T Gv)−1
+(9)
+Updating
+Gv: With Fv fixed, we can solve the opti-
+mization problem of Gv which is similar with above method.
+The solution of Gv can be easily obtained:
+Gv = Fv(FvT Fv)−1
+(10)
+Based on the above solution of Fv and Gv, we sum-
+marize the above iteratively method in Algorithm 1. No-
+tably, in order to avoid obtaining the singular value, we
+add a small smooth item in optimization. That is, Fv =
+φ(Xv)φ(Xv)T Gv(GvT φ(Xv)φ(Xv)T Gv +θI)−1 and Gv =
+Algorithm 1 Low-Rank Affinity Graphs Learning Algorithm
+Input: Kernelized dataset φ(Xv); the parameter r, and the
+maximum number of iterations Iter
+Output: Fv, Gv
+1: set t = 1 and randomly initialize F(0)v and G(0)v
+2: while t < Iter do
+3:
+Compute F(t)v according to Eq. 9
+4:
+Compute G(t)v according to Eq. 10
+5:
+t = t + 1
+6: end while
+7: return F(t)v and G(t)v
+Fv(FvT Fv + θI)−1 where θ is set in range of [1e−4, 1e−6] in
+our paper.
+After the above low-rank affinity graphs learning with
+auxiliary matrices strategy, we have rewritten the formula
+to auto-encoders hash for clustering as Eq. 11. In order to
+obtain the optimal solution of Eq. 11, we propose an iterative
+optimization method. The proposed method can effectively
+maintain discrete constraints for binary code and get more
+efficient binary code for clustering.
+min
+Zv,B,Q,H,pv
+M
+�
+v=1
+(∥FvGvT − Zv∥2
+F + (pv)k(∥WvZv − B∥2
+F
++ ∥Zv − WvT B∥2
+F )) + λ ∥B − QH∥2
+F
+s.t.
+WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI,
+QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N,
+M
+�
+v=1
+pv = 1,pv > 0,
+c
+�
+i=1
+his = 1
+(11)
+Updating
+Zv: By fixing all variables but Zv, problem(11)
+reduces to:
+min
+Z
+M
+�
+v=1
+(∥FvGvT − Zv∥2
+F
++ (pv)k(∥WvZv − B∥2
+F + ∥Zv − WvT B∥2
+F ))
+(12)
+In order to minimize the above equation, we convert the
+Frobenius norm in the above equation to the trace form of the
+matrix which is convenient for derivation. And then we take
+the derivative of the trace of the matrix with regard to Zv as
+follows:
+∂LZv
+∂Zv =
+M
+�
+v=1
+tr(ZvZvT − 2FvGvT ZvT
++(pv)k(WvZvZvTWvT +ZvZvT −4WvTBZvT ))
+(13)
+whose solution can be easily achieved by setting the derivation
+to 0:
+Zv =
+((pv)kWvT Wv + I + (lv)kI)−1(FvGvT + 2(lv)kWvT B)
+(14)
+
+6
+Updating
+Wv: In this stage, other variables are fixed.
+We take the constraint WvWvT = I in consider, the whole
+loss function related to Wv in Eq. 11 can be rewritten as:
+LWv = min
+Wv
+M
+�
+v=1
+(∥WvZv − B∥2
+F + ∥Zv − WvT B∥2
+F )
+= max tr(WvZvBT )
+s.t.
+WvWvT = I
+(15)
+where we also need to consider the condition BBT = I which
+is used during the optimization process. Specifically, we firstly
+convert Eq. 15 to trace of matrix form, and then we utilize the
+SVD algorithm [50] to solve the optimization problem.
+Wv = SDT
+(16)
+where S and D are the left and right singular vectors of the
+compact Singular Value Decomposition (SVD) of ZvBT .
+Updating
+B: The common gradient method is not suit-
+able for solving the discrete binary codes B. We rewrite Eq11
+related to B as follows:
+max
+B tr(BT (2
+M
+�
+v=1
+(pv)kWvZv + λQH))
+s.t.
+B ∈ {−1, +1}b×N, BBT = NI
+(17)
+We need to preserve constraints of B which can generate
+compact and effective binary code. Thus, the optimal binary
+code B can be obtained as follows, and sgn(·) means sym-
+bolic function.
+B = sgn(
+M
+�
+v=1
+(2(pv)kWvZv) + λQH)
+(18)
+Updating
+Q
+and
+H: In this part, we iteratively op-
+timize the binary clustering model by matrix factorization
+in Hamming space. By removing the irrelevant terms, the
+problem can be rewritten as:
+min
+Q,H ∥B − QH∥2
+F
+s.t.
+QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N,
+c
+�
+i=1
+his = 1,
+B ∈ {−1, +1}b×N
+(19)
+We simply reformulate Eq. 19 to:
+min
+Q,H ∥B − QH∥2
+F + ρ
+���QT 1
+���
+2
+s.t.
+Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N,
+c
+�
+i=1
+his = 1
+(20)
+which is equal Eq. 19 with adaptively large ρ. The proposed
+matrix factorization strategy iteratively optimizes the cluster
+centroids and indicators as follows.
+Updating
+Q
+: Due to the discrete constraint on Q, we
+utilize the discrete proximal linearized minimization (DPLM)
+[51] method, which can effectively obtain high-quality binary
+solutions. With H fixed, the optimal Q can be obtained as
+follow:
+minLQ = ∥B − QH∥2
+F + ρ
+���QT 1
+���
+2
+= −2tr(BT QH) + ρ
+���QT ���
+2
++ con
+s.t.
+Q ∈ {−1, +1}l×c
+(21)
+According to the DPLM, in the t + 1-th iteration Q can be
+updated as:
+Qt+1 = sgn(Qt − 1
+µ∇LQt)
+(22)
+where ∇LQt represents the gradient of LQ
+Updating
+H
+: We utilize vectors-based method to
+optimize the indicator matrix H, the solution to hi,j can be
+easily obtained by
+ht+1
+i,j =
+� 1,
+j = arg minsD(bi, qt+1
+s
+)
+0,
+otherwise
+(23)
+where D(bi, qt+1
+s
+) is the distance between i-th binary code bi
+and the s-th cluster centroid qs in Hamming space. Notably,
+we use binary code rather than real-value to calculate Eq. 23,
+which adopt Hamming distance that can save time compared
+to the Euclidean distance.
+Updating
+pv:
+For
+simplicity,
+we
+let
+av
+=
+∥WvZv − B∥2
+F + ∥Zv − WvT B∥2
+F . Based on the attributes
+of different views, the weighting coefficient pv can be
+equivalent as following equation:
+min
+pv
+M
+�
+v=1
+(pv)kav
+s.t.
+M
+�
+v=1
+pv = 1,pv > 0
+(24)
+Due to the constraint on pv, we can solve this problem
+by the Lagrange multiplier method. By setting the Lagrange
+multiplier Γ, Eq. 24 can be rewritten as:
+min L(pv, Γ) =
+M
+�
+v=1
+(pv)kav − Γ(
+M
+�
+v=1
+pv − 1)
+(25)
+and then, we calculate the partial derivative of L(pv, Γ) about
+pv and Γ, we can get:
+�
+�
+�
+∂L
+∂pv
+= k(pv)k−1av − Γ
+∂L
+∂Γ
+=
+M
+�
+v=1
+(pv)k − 1
+(26)
+Therefore, we set partial derivative to zero which can get:
+pv =
+(av)
+1
+1−k
+M
+�
+v=1
+(av)
+1
+1−k
+(27)
+We have presented the whole optimization process for Eq.
+11. And we summarize the whole process in Algorithm 2,
+which iteratively updates the variables until convergence.
+
+7
+Table II: The Comparison Result with Hash Method.
+Datasets
+Metrics
+SH
+DSH
+SP
+ITQ
+SGH
+RSSH
+RFDH
+HSIC
+BMVC
+GCAE
+100leaves
+ACC
+0.4713
+0.4494
+0.4750
+0.4569
+0.5088
+0.3631
+0.4513
+0.6563
+0.4981
+0.8888
+NMI
+0.7214
+0.7320
+0.7240
+0.7405
+0.7579
+0.6203
+0.7161
+0.8245
+0.7291
+0.9426
+Purity
+0.4950
+0.4988
+0.5138
+0.4944
+0.5338
+0.3888
+0.4838
+0.6788
+0.5331
+0.9031
+F-score
+0.3471
+0.3219
+0.3312
+0.3324
+0.3996
+0.2209
+0.3234
+0.5431
+0.3057
+0.8366
+Precision
+0.3184
+0.2653
+0.2942
+0.2528
+0.3626
+0.2094
+0.2708
+0.5128
+0.2547
+0.8166
+ARI
+0.3404
+0.3141
+0.3240
+0.3240
+0.3933
+0.2131
+0.3158
+0.5385
+0.2978
+0.8350
+Caltech-101
+ACC
+0.1747
+0.1610
+0.2077
+0.2472
+0.2256
+0.2860
+0.2196
+0.2429
+0.2930
+0.3005
+NMI
+0.3252
+0.3600
+0.4034
+0.4404
+0.4349
+0.4846
+0.4424
+0.4451
+0.4900
+0.4711
+Purity
+0.3089
+0.3439
+0.3878
+0.4231
+0.4051
+0.4729
+0.4228
+0.4125
+0.4907
+0.4421
+F-score
+0.1486
+0.1250
+0.1738
+0.2311
+0.2089
+0.2533
+0.2081
+0.2055
+0.2465
+0.3023
+Precision
+0.2604
+0.1975
+0.2897
+0.3544
+0.3085
+0.4071
+0.3419
+0.3430
+0.4147
+0.4229
+ARI
+0.1347
+0.1091
+0.1596
+0.2167
+0.1935
+0.2406
+0.1943
+0.1919
+0.2336
+0.2880
+Cifar-10
+ACC
+0.1708
+0.2189
+0.2275
+0.2245
+0.2205
+0.1960
+0.2252
+0.2153
+0.2350
+0.2498
+NMI
+0.0282
+0.0938
+0.0979
+0.0987
+0.0995
+0.0659
+0.1011
+0.0920
+0.1016
+0.1020
+Purity
+0.1727
+0.2271
+0.2285
+0.2297
+0.2218
+0.2046
+0.2333
+0.2236
+0.2368
+0.2549
+F-score
+0.1137
+0.1462
+0.1342
+0.1266
+0.1458
+0.1316
+0.1560
+0.1441
+0.1590
+0.1597
+Precision
+0.1129
+0.1520
+0.1495
+0.1544
+0.1408
+0.1284
+0.1457
+0.1420
+0.1533
+0.1563
+ARI
+0.0145
+0.0624
+0.0584
+0.0610
+0.0600
+0.0325
+0.0582
+0.0476
+0.0618
+0.0642
+SUNRGBD
+ACC
+0.1164
+0.1577
+0.1925
+0.1895
+0.1823
+0.1613
+0.1738
+0.1616
+0.1379
+0.2423
+NMI
+0.1319
+0.2198
+0.2180
+0.2108
+0.2172
+0.1980
+0.2032
+0.2202
+0.1545
+0.2207
+Purity
+0.2441
+0.3271
+0.3418
+0.3366
+0.3433
+0.3279
+0.3332
+0.3524
+0.2803
+0.3435
+F-score
+0.0650
+0.1033
+0.1223
+0.1274
+0.1184
+0.0976
+0.1145
+0.1059
+0.0822
+0.1541
+Precision
+0.1193
+0.1861
+0.1722
+0.1802
+0.1822
+0.1869
+0.1901
+0.2013
+0.1550
+0.1909
+ARI
+0.0309
+0.0699
+0.0895
+0.0951
+0.0859
+0.0661
+0.0823
+0.0744
+0.0496
+0.1076
+Caltech256
+ACC
+0.0622
+0.0782
+0.0856
+0.0865
+0.0826
+0.0937
+0.0783
+0.0678
+0.0932
+0.1071
+NMI
+0.2557
+0.2866
+0.2947
+0.2717
+0.2969
+0.3058
+0.2855
+0.2350
+0.3185
+0.2871
+Purity
+0.1096
+0.1255
+0.1399
+0.1386
+0.1363
+0.1512
+0.1278
+0.1064
+0.1534
+0.1415
+F-score
+0.0465
+0.0644
+0.0635
+0.0975
+0.0623
+0.0805
+0.0610
+0.0402
+0.0745
+0.1145
+Precision
+0.0527
+0.0681
+0.0658
+0.0950
+0.0649
+0.0932
+0.0644
+0.0305
+0.0811
+0.1037
+ARI
+0.0415
+0.0591
+0.0582
+0.0920
+0.0569
+0.0758
+0.0558
+0.0327
+0.0695
+0.1087
+Algorithm 2 GCAE Algorithm
+Input: Dataset by kernelized φ(Xv); the parameter λ; the
+auxiliary matrices Fv and Gv;
+Output: Binary code B; cluster indicator matrix H;
+1: Randomly initialize binary code B; affinity graph Zv;
+weights for different view pv; binary code length b;
+project matrix Wv
+2: repeat
+3:
+Update Zv according to Eq. 12
+4:
+Update Wv according to Eq. 16
+5:
+Update B according to Eq. 18
+6:
+Update Q and H according to Eq. 22 and Eq. 23
+7:
+Update pv according to Eq. 27
+8: until convergance
+IV. EXPERIMENTS
+In this section, we describe the datasets and comparison
+methods to verify the effectiveness of the proposed GCAE
+for clustering tasks. We evaluated the performance of GCAE
+by comparing several hash methods and multi-view clustering
+methods on 5 widely used datasets. Moreover, we analyze
+the parameter sensitivity of our model, which can affect the
+fluctuations of results. We also summarize the running times
+on all datasets and evaluate the convergence of our model.
+All the experiments are conducted using Matlab 2020a on a
+Windows PC with Intel 2.8-GHz CPU and 64-GB RAM.
+A. Experimental settings
+In this part, we describe the utilized datasets and the com-
+parison methods in detail. We also introduce some evaluation
+metrics which aim to verify the effectiveness of GCAE.
+1) Datasets: To evaluate the clustering performance of the
+proposed model and comparison models, five widely-accepted
+multi-view datasets are selected. The details of the selected
+datasets are as follows:
+Caltech1011 contains 9144 samples of 101 objects. 6 pub-
+licly available features are utilized as multiple views, including
+Gabor feature with 48 dimension, LBP feature with 928
+dimension, GIST feature with 512 dimension, CENTRIST
+feature with 254 dimension, wavelet moments(WM) with 40
+dimension and HOG feature with 1984 dimension.
+Caltech2562 contains 30607 images by 4 kinds of features.
+It includes 256 classes with more than 80 samples per class.
+Besides, each image is represented by color histogram with
+729 dimension, Gist feature with 1024 dimension, HOG
+feature with 1152 dimension and features based on convolution
+network with 1440 dimension, which are four different types
+of presentation.
+Cifar-103 is composed by 60000 tiny color images in 10
+kinds of classes. Specifically, we represent the features by
+1http://www.vision.caltech.edu/ImageDatasets/Caltech101/
+2http://www.vision.caltech.edu/Image-Datasets/Caltech256
+3https://www.cs.toronto.edu/kriz/cifar.html.
+
+8
+DSD feature with 220 dimension, HOG feature with 512
+dimension and Gist feature with 768 dimension. Besides, a
+subset of this dataset is selected in the experiment that contains
+10000 samples.
+100leaves4 [52] contains 1600 samples from each of 100
+plant species, which form the UCI repository with three
+different views, i.e., Shape, Texture and Margin features.
+SUNRGBD5 contains 10335 indoor scene images by 3D
+cameras in 45 classes. Similar with [53], we utilize two
+views with 4096 dimension and features extracted by different
+convolution network.
+2) Comparing Algorithms and Evaluation Metrics: In our
+experiments, we evaluate the performance of GCAE by com-
+paring several state-of-the-art multi-view algorithms and hash
+algorithms. Specifically, we utilize seven single-view hash
+methods and two multi-view hash clustering algorithms, in-
+cluding SH [23], DSH [54], SP [55], ITQ [56], SGH [25],
+RSSH [34], RFDH [36], HSIC [39], BMVC [32]. For single-
+view hash methods, we utilize K-means algorithm to process
+the generated binary code for clustering and we preserve
+the best result of each view clustering. Besides, we adopt
+eight real-value multi-view clustering algorithms as comparing
+algorithms, including K-means [42], SC [57], Co-regularize
+[26], AMGL [40], Mul-NMF [58], MLAN [59], mPAC [60],
+GMC [?]. We utilize the source code from the author’s public
+homepage for comparing. Notably, we set the length of binary
+code as 128-bits in our experiments, which can effectively
+preserve critical information.
+In order to generally evaluate the superiority of our method,
+we utilize six widely used evaluation metrics, i.e., clustering
+accuracy(ACC), Normalized Mutual Information(NMI), Pu-
+rity, F-score, Precision and ARI [?], [61]. For all algorithms,
+better performance is improved by the higher value of evalu-
+ation metrics.
+B. Experimental Results and Analysis
+In this section, we conducted experiments on 5 multi-view
+datasets to show the superiority of the proposed GCAE. The
+detailed clustering results are shown in Table II, III. We utilize
+the bold values to represent the best performance in each
+table corresponding to the dataset. Besides, this section also
+introduced the analysis of parameter sensitivity, which can
+reflect clustering results with different parameters. Finally, we
+provide the complexity analysis and convergence analysis of
+the proposed model to verify the stability of GCAE.
+1) Comparison with Hash Methods: In this section, we
+conduct experiments with hash methods on five multi-view
+datasets to verify the performance of our proposed model.
+We compare GCAE with seven single-view hash methods
+and two multi-view hash clustering methods. In single-view
+hash methods, we generate cluster results by adopting k-means
+algorithm to finish cluster task. Specifically, we adopt K-means
+algorithm with cluster binary code to obtain sample labels. The
+results with different datasets are reported in Table II.
+4https://archive.ics.uci.edu/ml/dataset
+5http://rgbd.cs.princeton.edu/
+We summarized the clustering performance with compared
+methods on all multimedia datasets in Table II, and the best
+results are highlighted in bold. As shown in Table II, we can
+obvious that GCAE obtains the best clustering accuracy on five
+multi-view datasets compared with hash methods. Specifically,
+we proposed GCAE method evidently outperforms other state-
+of-the-art methods for 100leaves and Cifar-10 datasets. For the
+100leaves dataset, the performance of GCAE improves around
+23.2%, 11.8%, and 22.4% with terms of ACC, NMI, and
+Purity over the second-best method. Moreover, for the Cifar-10
+dataset, the performance improvements over the second-best
+method are 1.5%, 0.4%, 1.81% with terms of ACC, NMI, and
+Purity.
+The experiments demonstrate that the multi-view hash clus-
+tering methods get better results than single-view methods,
+which proves multi-view methods can exploit complementary
+information from multi-view. For single-view methods, we
+perform clustering on each views and select the best results
+to report in the above table. Generally speaking, multi-view
+methods take the complementary information and consis-
+tent information between multi-view data into consideration.
+Therefore, multi-view methods can get better performance and
+single-view methods can not obtain satisfactory performance
+in most situations. The results in Table II also verify the low-
+rank affinity graphs construction is a vital part of our proposed
+GCAE, which aims to effectively generate binary code with
+essential information.
+Overall, the clustering performance about our proposed
+model outperforms the other hash methods in most situations.
+The phenomenon denotes that GCAE can effectively preserve
+essential information by auto-encoders structure and keep the
+discrete constraint on binary code. Besides, GCAE has the
+ability to integrate comprehensive information from multi-
+view data.
+2) Multi-view Methods Experimental Results and Analysis:
+In this section, we present the detailed comparison clustering
+results with multi-view clustering methods in Table III. As
+shown in this table, we compare GCAE with eight real-
+value clustering methods and two hash clustering methods.
+Moreover, for single-view methods i.e., K-means and SC, we
+concatenates all multiple views into one view for clustering.
+As shown in Table III, it is clear that GCAE obtains the
+best clustering accuracy with the comparison results on all
+multi-view datasets. Specifically, the proposed GCAE model
+outperforms all other methods on ACC, NMI, Purity, F-score,
+Precision, and ARI in most situations. We can obviously find
+the performance improvements over the second-best method
+are 1.9%, 1.4% and 14.1% respectively in metrics of ACC,
+NMI, and Purity for 100leaves dataset. Moreover, for the Cifar-
+10 dataset, our proposed model can improve 1.5%, 0.4%,
+and 1.81% in the above metrics which is compared with the
+second-best method.
+We conducted the experiments with comparing real-value
+based multi-view clustering method and hash methods, which
+aims to verify the stability of clustering in Hamming space.
+It is clear that hash methods can obtain satisfactory perfor-
+mance in most situations. This is because real-value methods
+utilize Euclidean distance to measure two samples, which have
+
+9
+Table III: The Clustering Results on Five Datasets.
+Datasets
+Metrics
+k-means
+SC
+Co-re-p
+Co-re-c
+AMGL
+Mul-NMF
+MLAN
+mPAC
+GMC
+HSIC
+BMVC
+GCAE
+100leaves
+ACC
+0.6200
+0.4894
+0.7253
+0.7939
+0.7631
+0.8694
+0.7356
+0.8238
+0.4338
+0.6563
+0.4981
+0.8888
+NMI
+0.8284
+0.7649
+0.8835
+0.9257
+0.9065
+0.9288
+0.8848
+0.9292
+0.7014
+0.8245
+0.7291
+0.9426
+Purity
+0.6550
+0.5575
+0.7565
+0.8272
+0.8063
+0.8981
+0.7625
+0.8506
+0.5350
+0.6788
+0.5331
+0.9031
+F-score
+0.5233
+0.2144
+0.6595
+0.7558
+0.4513
+0.8236
+0.6583
+0.5042
+0.3145
+0.5431
+0.3057
+0.8366
+Precision
+0.4689
+0.1307
+0.6107
+0.7011
+0.3086
+0.7832
+0.6082
+0.3521
+0.3744
+0.5128
+0.2547
+0.8166
+ARI
+0.5183
+0.2021
+0.6560
+0.7533
+0.4437
+0.8219
+0.6548
+0.4974
+0.3070
+0.5385
+0.2978
+0.8350
+Caltech-101
+ACC
+0.1331
+0.1751
+0.2611
+0.2587
+0.1476
+0.1908
+0.2274
+0.1950
+0.2672
+0.2429
+0.2940
+0.3005
+NMI
+0.3078
+0.3207
+0.4752
+0.4912
+0.3757
+0.3519
+0.4564
+0.3446
+0.4408
+0.4451
+0.4900
+0.4711
+Purity
+0.2907
+0.3107
+0.4622
+0.4664
+0.1681
+0.3184
+0.4401
+0.3012
+0.3497
+0.4125
+0.4907
+0.4421
+F-score
+0.0985
+0.1362
+0.2202
+0.2226
+0.0338
+0.0470
+0.1930
+0.0496
+0.2658
+0.2055
+0.2465
+0.3023
+Precision
+0.1351
+0.1440
+0.3954
+0.3999
+0.0175
+0.0248
+0.3185
+0.0261
+0.2337
+0.3430
+0.4147
+0.4229
+ARI
+0.0796
+0.1126
+0.2078
+0.2103
+0.0155
+-0.0068
+0.1790
+-0.0042
+0.2475
+0.1919
+0.2336
+0.2880
+Cifar-10
+ACC
+0.2036
+0.1712
+0.2169
+0.2084
+0.2232
+0.1217
+0.2304
+0.1038
+0.2312
+0.2153
+0.2350
+0.2498
+NMI
+0.0891
+0.0773
+0.0944
+0.0903
+0.0845
+0.0204
+0.0919
+0.0066
+0.1014
+0.0920
+0.1016
+0.1020
+Purity
+0.2055
+0.1751
+0.2214
+0.2180
+0.2276
+0.1235
+0.2522
+0.1043
+0.2462
+0.2236
+0.2368
+0.2549
+F-score
+0.1529
+0.1443
+0.1593
+0.1477
+0.1587
+0.1582
+0.1570
+0.1511
+0.1499
+0.1441
+0.1590
+0.1597
+Precision
+0.1344
+0.1080
+0.1453
+0.1393
+0.1393
+0.1005
+0.1537
+0.0999
+0.1495
+0.1420
+0.1533
+0.1563
+ARI
+0.0443
+0.0155
+0.0561
+0.0467
+0.0594
+0.0012
+0.0612
+0.0603
+0.0607
+0.0476
+0.0618
+0.0642
+SUNRGBD
+ACC
+0.1859
+0.1060
+0.1824
+0.1829
+0.1010
+0.1346
+0.1898
+0.1277
+0.2155
+0.1616
+0.1379
+0.2423
+NMI
+0.1866
+0.0084
+0.2109
+0.2161
+0.1883
+0.0976
+0.2101
+0.0728
+0.2204
+0.2202
+0.1545
+0.2207
+Purity
+0.3022
+0.1087
+0.3274
+0.3360
+0.1119
+0.1583
+0.3257
+0.1415
+0.2303
+0.3524
+0.2803
+0.3435
+F-score
+0.1293
+0.1213
+0.1172
+0.1221
+0.0637
+0.1200
+0.1209
+0.1215
+0.1387
+0.1059
+0.0822
+0.1541
+Precision
+0.1057
+0.0646
+0.1849
+0.1748
+0.0364
+0.0645
+0.1822
+0.0650
+0.1007
+0.2013
+0.1550
+0.1909
+ARI
+0.0976
+0.0323
+0.0865
+0.0916
+0.0262
+0.0251
+0.0891
+0.0008
+0.1026
+0.0744
+0.0496
+0.1076
+Caltech-256
+ACC
+0.0845
+0.0924
+0.0854
+0.1025
+0.0467
+0.0612
+0.0761
+0.0904
+0.0723
+0.0678
+0.0932
+0.1071
+NMI
+0.2748
+0.2764
+0.2997
+0.2786
+0.1070
+0.1486
+0.2820
+0.2198
+0.1347
+0.2350
+0.3185
+0.2871
+Purity
+0.1296
+0.1339
+0.1412
+0.1522
+0.0415
+0.1060
+0.1343
+0.1381
+0.1008
+0.1064
+0.1534
+0.1415
+F-score
+0.0562
+0.0628
+0.0596
+0.0713
+0.0466
+0.0106
+0.0556
+0.0498
+0.0355
+0.0402
+0.0745
+0.1145
+Precision
+0.0522
+0.0855
+0.0692
+0.0813
+0.0458
+0.0053
+0.0570
+0.0412
+0.0515
+0.0305
+0.0811
+0.1037
+ARI
+0.0501
+0.0846
+0.0549
+0.0689
+0.0488
+-0.0011
+0.0501
+0.0361
+0.0834
+0.0327
+0.0695
+0.1087
+1e-9
+1e-8
+1e-7
+1e-6
+1e-5
+1e-4
+1e-3
+0
+0.05
+0.1
+0.15
+0.2
+0.25
+0.3
+0.35
+0.4
+ACC
+SUNRGBD
+Caltech-101
+Cifar-10
+(a) Clustering accuracy with λ
+3
+4
+5
+6
+7
+8
+k
+0
+0.05
+0.1
+0.15
+0.2
+0.25
+0.3
+0.35
+0.4
+ACC
+SUNRGBD
+Caltech-101
+Cifar-10
+(b) Clustering accuracy with k
+200
+400
+600
+800
+1000
+1200
+1400
+t
+0
+0.05
+0.1
+0.15
+0.2
+0.25
+0.3
+0.35
+0.4
+ACC
+SUNRGBD
+Caltech-101
+Cifar-10
+(c) Clustering accuracy with t
+1e-8
+1e-7
+1e-6
+1e-5
+1e-4
+1e-3
+1e-2
+1e-1
+0
+0.05
+0.1
+0.15
+0.2
+0.25
+0.3
+0.35
+0.4
+ACC
+SUNRGBD
+Caltech-101
+Cifar-10
+(d) Clustering accuracy with θ
+Fig. 2: Clustering accuracy with parameter tuning for the Caltech-101, Cifar-10 and SUNRGBD datasets (a)λ (b)k (c)t (d)θ.
+problems of low efficiency and high computational complexity.
+And hash methods learn binary code and obtain cluster results
+in Hamming space, which can improve calculation efficiency.
+The adoption of the affinity graph is more conductive to pre-
+serving vital information of mining original data, but utilizing
+hash methods only improves significantly in computational
+complexity, which can not significantly improve evaluation
+metrics.
+It is worth noting that the proposed GCAE obtain the
+best cluster accuracy and satisfactory performance on other
+evaluation metrics for five multi-view datasets. This is because
+GCAE can explore consistency and complementary informa-
+tion in multi-view data structure by learning low-rank affinity
+graphs, which learn high-quality collaborative representation
+well and eliminates some redundant and noise information in
+the original real-valued features. Besides, GCAE can also im-
+prove the computational complexity of algorithm significantly.
+C. Parameter Sensitivity and Complexity Analysis
+In this section, we discover that the fluctuation of several
+parameters has a significant influence on the experiments.
+In order to evaluate the sensitivity of the parameters in the
+proposed GCAE, we select 100leaves, Caltech-101, Cifar-10,
+and SUNRGBD as the benchmark datasets. For GCAE, there
+are six parameters to be tuned, i.e., λ, k, t, θ, b and r. The
+above parameters represent regularization parameter for binary
+clustering, the power of normalized weighting coefficient,
+the dimension of nonlinear kernel, the small smooth item,
+binary bits, respectively. We firstly conducted experiments on
+the Caltech-101, Cifar-10, and SUNRGBD datasets to show
+the influence of parameters λ, k, t and θ change trend on
+the results. Specifically, we calculate the cluster accuracy
+values and display them in line charts, which can intuitively
+find the influence of parameters changing. Fig. 2 shows the
+different change trends on three benchmark datasets. In Fig.
+2(a), we can observe the different λ values have effected
+
+10
+8
+16
+32
+64
+128
+Length of binary code
+0
+0.1
+0.2
+0.3
+0.4
+0.5
+0.6
+0.7
+0.8
+0.9
+1
+Performance
+ACC
+NMI
+Purity
+(a) 100leaves
+8
+16
+32
+64
+128
+Length of binary code
+0
+0.1
+0.2
+0.3
+0.4
+0.5
+0.6
+Performance
+ACC
+NMI
+Purity
+(b) Caltech-101
+8
+16
+32
+64
+128
+Length of binary code
+0
+0.05
+0.1
+0.15
+0.2
+0.25
+0.3
+0.35
+0.4
+Performance
+ACC
+NMI
+Purity
+(c) Cifar-10
+Fig. 3: From left to right, the performance with binary code length tuning for the 100leaves, Caltech-101 and Cifar-10 datasets,
+respectively.
+(a) 100leaves
+04
+。
+。
+。
+uuv
+b
+
+r
+
+(b) Caltech-101
+(c) Cifar-10
+Fig. 4: Parameter tuning with respect to r and b for the 100leaves, Caltech-101 and Cifar-10 datasets.
+the cluster accuracy notably on three datasets. We conduct
+experiments with different parameters λ selected from the
+range of [1e − 9, 1e − 3]. It is obvious that the change trend
+of GCAE on the three datasets and GCAE can achieve ideal
+performance when λ is set to an appropriate value. Besides, it
+is obvious that the parameter k can effect cluster accuracy
+values selected from the range of [3, 8] in Fig. 2(b). The
+proposed GCAE can achieve excellent performance with a
+good selection of k. As shown in Fig. 2(c) and Fig. 2(d),
+the change trend of t and θ do not influence cluster results
+obviously. It is notable that GCAE can still achieve a relatively
+stable and good performance with the change in the dimension
+of nonlinear kernel t and the small smooth item θ, which are
+selected in range of [200, 1400] and [1e−8, 1e−1], respectively.
+As shown in Fig. 3, we adopt comparison experiments
+of the length of binary code to verify our proposed GCAE.
+Therefore, we change the binary code length and calculate
+ACC, NMI, and Purity on 100leaves, Caltech-101 and Cifar-
+10 datasets to show the performance of GCAE. Above the
+figure, we select 128-bits binary code in GCAE, which aims
+to fully preserve essential information for clustering. Besides,
+in order to demonstrate the relation between the rank of affinity
+graphs and the length of binary code. We conducted a three-
+dimensional statistical graph of clustering results by adjusting
+the values of r and b. The statistical results of 100leaves,
+Caltech-101 and Cifar-10 datasets are shown in Fig. 4, which
+clustering performance as ACC. We can obviously find GCAE
+can obtain excellent performance with a good selection of
+r and b, which are designed in the range of [20, 200] and
+[8, 128] respectively. In conclusion, we select 128-bits binary
+codes and r = 100 on these datasets, which achieves the best
+clustering performance with the proposed GCAE algorithm.
+We summarized the overall algorithm of the proposed
+GCAE method in Algorithm 1 and Algorithm 2, which present
+low-rank affinity graphs learning and graph-collaborated auto-
+encoder hashing respectively. The computational burden of
+GCAE mainly consists of the above two sub-process, and we
+provide the analysis of GCAE in detail. Specifically, the two
+closed-form solutions, i.e., Eq. 9 and Eq. 10 cost O(Ndvr +
+Nr2 + r3). Therefore, the whole computational complexity
+of Algorithm 1 is O(N × max(dv, r) × r × Iter), where we
+set the Iter to 80. For Algorithm 2, we adopt the iterative
+optimization method to obtain the solution for each parameter.
+When updating Zv, which is the most time-consuming part
+of GCAE. It costs O(N 3 + N 2b + N 2r) because it needs
+to calculate the inverse of the matrix. And then, we need
+O(N 2b) for updating W v, which contains SVD operation
+and matrix multiplication. After that, in order to update B
+and 23, we need O(N 2b + Nbc) which includes Frobenius
+norm and ADPLM method. Besides, solving Eq. 22 and Eq. 23
+needs O(Nc) and updating lv costs O(rN) for each iteration.
+Overall, the whole computational complexity of GCAE is
+O((N 3 +3N 2b+N 2r +Nbc)p+Nmax(dv, r)rIter), which
+p is the number of epoch. For clarity, we summarized the
+running time of GCAE for each dataset in Table IV.
+Table IV: Running time of GCAE.
+Datasets
+100leaves
+Caltech-101
+Cifar-10
+SUNRGBD
+Caltech-256
+Times(s)
+14.16
+1158.71
+1120.15
+1488.33
+47618
+D. Convergence Analysis
+In order to verify the proposed iteratively optimization al-
+gorithm for GCAE is convergent, we conduct the convergence
+
+3
+6-
+120
+4
+203
+2
+200
+0
+180
+160
+128
+140
+120
+64
+100
+32
+80
+门
+16
+60
+40
+80.4
+0.3
+0.2
+0.1
+200
+0
+180
+160
+128
+140
+64
+120
+32
+100
+80
+I
+b
+16
+60
+8
+40
+200.9
+0.8
+0.7
+0.6
+CC
+0.5
+0.4
+0.3
+0.2
+0.1
+200
+0
+180
+160
+128
+140
+64
+120
+32
+100
+80
+I
+b
+16
+60
+8
+40
+2011
+0
+5
+10
+15
+20
+25
+Iteration step
+181
+182
+183
+184
+185
+186
+187
+188
+Objective value
+(a) 100leaves
+0
+5
+10
+15
+Iteration step
+1785
+1790
+1795
+1800
+1805
+1810
+Objective value
+(b) Caltech-101
+0
+5
+10
+15
+20
+25
+30
+Iteration step
+939
+940
+941
+942
+943
+944
+945
+946
+947
+948
+949
+950
+Objective value
+(c) Cifar-10
+0
+5
+10
+15
+20
+25
+30
+35
+40
+45
+50
+Iteration step
+656
+658
+660
+662
+664
+666
+668
+670
+Objective value
+(d) SUNRGBD
+Fig. 5: Convergence results for the 100leaves, Caltech-101,
+Cifar-10 and SUNRGBD datasets, respectively.
+analysis of the proposed GCAE in this section. Fig. 5 shows
+the convergence curves of GCAE on the datasets 100leaves,
+Caltech-101 and Cifar-10. In Fig .5, the x-axis and y-axis
+present the number of iterations and the value of the objective
+function, respectively. It can obviously find that the values
+of the objective function in GCAE decrease monotonically
+with the increased iterations. Furthermore, the objective values
+converge very quickly after 5-10 iterations which verifies the
+effectiveness of our proposed optimized solution for each sub-
+process.
+V. CONCLUSION
+In this paper, we proposed a novel binary multi-view
+clustering method termed as Graph-Collaborated Auto-encoder
+Hashing for Multi-view Binary Clustering (GCAE), which
+can effectively construct low-rank affinity graphs from each
+view and jointly learn binary code by auto-encoders. GCAE
+constructs graphs by utilizing auxiliary matrices to control the
+low-rank constraint, which can reasonably preserve essential
+information from original data. And then GCAE adopts auto-
+encoders to collaborate multiple graphs for learning unified
+binary codes and obtaining cluster results. With the pro-
+posed optimization algorithm, the objective function converges
+quickly. The extensive experiments demonstrated the supe-
+riority of GCAE. It can be obviously seen that different
+from real-value multi-view clustering methods, GCAE can
+effectively obtain cluster results in Hamming space. Compared
+with hash methods, GCAE can reasonably utilize multi-view
+consistency and complementarity information. We conducted
+our experiments on five multi-view datasets for experimental
+examination. In the future work, we intend to further improve
+the speed of the algorithm, which is the deficiency of the
+proposed method.
+ACKNOWLEDGMENT
+This work was supported in part by the National Nat-
+ural Science Foundation of China Grant 62002041 and
+62176037, Liaoning Fundamental Research Funds for Uni-
+versities Grant LJKQZ2021010, Liaoning Doctoral Research
+Startup Fund Project Grant 2021-BS-075, Liaoning Province
+Applied Basic Research Project 22JH2/101300264 and Dalian
+Science and Technology Innovation Fund 2022JJ12GX019,
+2021JJ12GX028 and 2022JJ12GX016.
+REFERENCES
+[1] C. Zhang, Y. Geng, Z. Han, Y. Liu, H. Fu, and Q. Hu, “Autoencoder
+in autoencoder networks,” IEEE transactions on neural networks and
+learning systems, 2022.
+[2] Y. Zhang, Z. Zhang, Y. Wang, Z. Zhang, L. Zhang, S. Yan, and M. Wang,
+“Dual-constrained deep semi-supervised coupled factorization network
+with enriched prior,” International Journal of Computer Vision, vol. 129,
+no. 12, pp. 3233–3254, 2021.
+[3] Y. Wang, L. Wu, X. Lin, and J. Gao, “Multiview spectral clustering via
+structured low-rank matrix factorization,” IEEE transactions on neural
+networks and learning systems, vol. 29, no. 10, pp. 4833–4843, 2018.
+[4] B. Qian, Y. Wang, H. Yin, R. Hong, and M. Wang, “Switchable online
+knowledge distillation,” in European Conference on Computer Vision.
+Springer, 2022, pp. 449–466.
+[5] Y. Wang, “Survey on deep multi-modal data analytics: Collaboration,
+rivalry, and fusion,” ACM Transactions on Multimedia Computing,
+Communications, and Applications (TOMM), vol. 17, no. 1s, pp. 1–25,
+2021.
+[6] X. Peng, J. Feng, J. T. Zhou, Y. Lei, and S. Yan, “Deep subspace
+clustering,” IEEE transactions on neural networks and learning systems,
+vol. 31, no. 12, pp. 5509–5521, 2020.
+[7] Y. Wang, X. Lin, L. Wu, W. Zhang, Q. Zhang, and X. Huang, “Robust
+subspace clustering for multi-view data by exploiting correlation con-
+sensus,” IEEE Transactions on Image Processing, vol. 24, no. 11, pp.
+3939–3949, 2015.
+[8] Z. Guo, L. Zhang, and D. Zhang, “A completed modeling of local binary
+pattern operator for texture classification,” IEEE transactions on image
+processing, vol. 19, no. 6, pp. 1657–1663, 2010.
+[9] J. R. Movellan, “Tutorial on gabor filters,” Open Source Document,
+vol. 40, pp. 1–23, 2002.
+[10] N. Dalal and B. Triggs, “Histograms of oriented gradients for human
+detection,” in 2005 IEEE computer society conference on computer
+vision and pattern recognition (CVPR’05), vol. 1.
+Ieee, 2005, pp.
+886–893.
+[11] D. G. Lowe, “Distinctive image features from scale-invariant keypoints,”
+International journal of computer vision, vol. 60, no. 2, pp. 91–110,
+2004.
+[12] Y. Gu, S. Wang, H. Zhang, Y. Yao, W. Yang, and L. Liu, “Clustering-
+driven unsupervised deep hashing for image retrieval,” Neurocomputing,
+vol. 368, pp. 114–123, 2019.
+[13] Y. Wang, J. Peng, H. Wang, and M. Wang, “Progressive learning with
+multi-scale attention network for cross-domain vehicle re-identification,”
+Science China Information Sciences, vol. 65, no. 6, pp. 1–15, 2022.
+[14] W. Hu, Y. Fan, J. Xing, L. Sun, Z. Cai, and S. Maybank, “Deep
+constrained siamese hash coding network and load-balanced locality-
+sensitive hashing for near duplicate image detection,” IEEE Transactions
+on Image Processing, vol. 27, no. 9, pp. 4452–4464, 2018.
+[15] H. Wang, G. Jiang, J. Peng, R. Deng, and X. Fu, “Towards adaptive
+consensus graph: Multi-view clustering via graph collaboration,” IEEE
+Transactions on Multimedia, 2022.
+[16] G. Jiang, J. Peng, H. Wang, Z. Mi, and X. Fu, “Tensorial multi-view
+clustering via low-rank constrained high-order graph learning,” IEEE
+Transactions on Circuits and Systems for Video Technology, 2022.
+[17] H. Wang, Y. Wang, Z. Zhang, X. Fu, L. Zhuo, M. Xu, and M. Wang,
+“Kernelized multiview subspace analysis by self-weighted learning,”
+IEEE Transactions on Multimedia, vol. 23, pp. 3828–3840, 2020.
+[18] Y. Wang and L. Wu, “Beyond low-rank representations: Orthogonal
+clustering basis reconstruction with optimized graph structure for multi-
+view spectral clustering,” Neural Networks, vol. 103, pp. 1–8, 2018.
+[19] F. Shen, C. Shen, W. Liu, and H. Tao Shen, “Supervised discrete
+hashing,” in Proceedings of the IEEE conference on computer vision
+and pattern recognition, 2015, pp. 37–45.
+[20] Y. Chen, Z. Tian, H. Zhang, J. Wang, and D. Zhang, “Strongly
+constrained discrete hashing,” IEEE Transactions on Image Processing,
+vol. 29, pp. 3596–3611, 2020.
+
+12
+[21] X. Liu, X. Wang, and Y.-m. Cheung, “Fddh: Fast discriminative discrete
+hashing for large-scale cross-modal retrieval,” IEEE Transactions on
+Neural Networks and Learning Systems, 2021.
+[22] P. Indyk and R. Motwani, “Approximate nearest neighbors: towards
+removing the curse of dimensionality,” in Proceedings of the thirtieth
+annual ACM symposium on Theory of computing, 1998, pp. 604–613.
+[23] Y. Weiss, A. Torralba, and R. Fergus, “Spectral hashing,” Advances in
+neural information processing systems, vol. 21, 2008.
+[24] W. Liu, C. Mu, S. Kumar, and S.-F. Chang, “Discrete graph hashing,”
+Advances in neural information processing systems, vol. 27, 2014.
+[25] Q.-Y. Jiang and W.-J. Li, “Scalable graph hashing with feature transfor-
+mation,” in Twenty-Fourth International Joint Conference on Artificial
+Intelligence, 2015.
+[26] A. Kumar, P. Rai, and H. Daume, “Co-regularized multi-view spectral
+clustering,” Advances in neural information processing systems, vol. 24,
+2011.
+[27] K. Zhan, C. Zhang, J. Guan, and J. Wang, “Graph learning for multiview
+clustering,” IEEE Transactions on Cybernetics, vol. 48, no. 10, pp.
+2887–2895, 2018.
+[28] Y. Wang, W. Zhang, L. Wu, X. Lin, M. Fang, and S. Pan, “Iterative
+views agreement: an iterative low-rank based structured optimization
+method to multi-view spectral clustering,” in Proceedings of the Twenty-
+Fifth International Joint Conference on Artificial Intelligence, 2016, pp.
+2153–2159.
+[29] H. Wang, Y. Yang, and B. Liu, “Gmc: Graph-based multi-view cluster-
+ing,” IEEE Transactions on Knowledge and Data Engineering, vol. 32,
+no. 6, pp. 1116–1129, 2020.
+[30] H. Xiao, Y. Chen, and X. Shi, “Knowledge graph embedding based
+on multi-view clustering framework,” IEEE Transactions on Knowledge
+and Data Engineering, vol. 33, no. 2, pp. 585–596, 2021.
+[31] S. Shi, F. Nie, R. Wang, and X. Li, “Multi-view clustering via nonnega-
+tive and orthogonal graph reconstruction,” IEEE Transactions on Neural
+Networks and Learning Systems, pp. 1–14, 2021.
+[32] Z. Zhang, L. Liu, F. Shen, H. T. Shen, and L. Shao, “Binary multi-
+view clustering,” IEEE transactions on pattern analysis and machine
+intelligence, vol. 41, no. 7, pp. 1774–1782, 2018.
+[33] W. Jin, S. Chaki, C. Cohen, A. Gurfinkel, J. Havrilla, C. Hines, and
+P. Narasimhan, “Binary function clustering using semantic hashes,” in
+2012 11th International Conference on Machine Learning and Applica-
+tions, vol. 1, 2012, pp. 386–391.
+[34] Z. Tian, H. Zhang, Y. Chen, and D. Zhang, “Unsupervised hashing based
+on the recovery of subspace structures,” Pattern Recognition, vol. 103,
+p. 107261, 2020.
+[35] G. Liu, Z. Lin, and Y. Yu, “Robust subspace segmentation by low-rank
+representation,” in Proceedings of the 27th International Conference on
+International Conference on Machine Learning, 2010, pp. 663–670.
+[36] D. Wang, Q. Wang, and X. Gao, “Robust and flexible discrete hashing
+for cross-modal similarity search,” IEEE Transactions on Circuits and
+Systems for Video Technology, vol. 28, no. 10, pp. 2703–2715, 2017.
+[37] W. Wang, Y. Shen, H. Zhang, Y. Yao, and L. Liu, “Set and rebase: de-
+termining the semantic graph connectivity for unsupervised cross-modal
+hashing,” in Proceedings of the Twenty-Ninth International Conference
+on International Joint Conferences on Artificial Intelligence, 2021, pp.
+853–859.
+[38] L. Wang, J. Yang, M. Zareapoor, and Z. Zheng, “Cluster-wise unsuper-
+vised hashing for cross-modal similarity search,” Pattern Recognition,
+vol. 111, p. 107732, 2021.
+[39] Z. Zhang, L. Liu, J. Qin, F. Zhu, F. Shen, Y. Xu, L. Shao, and H. T. Shen,
+“Highly-economized multi-view binary compression for scalable image
+clustering,” in Proceedings of the European Conference on Computer
+Vision (ECCV), 2018, pp. 717–732.
+[40] F. Nie, J. Li, X. Li et al., “Parameter-free auto-weighted multiple graph
+learning: a framework for multiview clustering and semi-supervised
+classification.” in IJCAI, 2016, pp. 1881–1887.
+[41] C. Hou, F. Nie, H. Tao, and D. Yi, “Multi-view unsupervised feature
+selection with adaptive similarity and view weight,” IEEE Transactions
+on Knowledge and Data Engineering, vol. 29, no. 9, pp. 1998–2011,
+2017.
+[42] J. A. Hartigan and M. A. Wong, “Algorithm as 136: A k-means
+clustering algorithm,” Journal of the royal statistical society. series c
+(applied statistics), vol. 28, no. 1, pp. 100–108, 1979.
+[43] K. Zhan, C. Niu, C. Chen, F. Nie, C. Zhang, and Y. Yang, “Graph struc-
+ture fusion for multiview clustering,” IEEE Transactions on Knowledge
+and Data Engineering, vol. 31, no. 10, pp. 1984–1993, 2018.
+[44] S. Shi, F. Nie, R. Wang, and X. Li, “Auto-weighted multi-view clustering
+via spectral embedding,” Neurocomputing, vol. 399, pp. 369–379, 2020.
+[45] L. Jin, K. Li, H. Hu, G.-J. Qi, and J. Tang, “Semantic neighbor
+graph hashing for multimodal retrieval,” IEEE Transactions on Image
+Processing, vol. 27, no. 3, pp. 1405–1417, 2018.
+[46] J. Guan, Y. Li, J. Sun, X. Wang, H. Zhao, J. Zhang, Z. Liu, and S. Qi,
+“Graph-based supervised discrete image hashing,” Journal of Visual
+Communication and Image Representation, vol. 58, pp. 675–687, 2019.
+[47] L. Xiang, X. Shen, J. Qin, and W. Hao, “Discrete multi-graph hashing
+for large-scale visual search,” Neural Processing Letters, vol. 49, no. 3,
+pp. 1055–1069, 2019.
+[48] Y. Fang, H. Zhang, and Y. Ren, “Unsupervised cross-modal retrieval
+via multi-modal graph regularized smooth matrix factorization hashing,”
+Knowledge-Based Systems, vol. 171, pp. 69–80, 2019.
+[49] S. Li and Y. Fu, “Learning robust and discriminative subspace with low-
+rank constraints,” IEEE transactions on neural networks and learning
+systems, vol. 27, no. 11, pp. 2160–2173, 2015.
+[50] D. Hu, F. Nie, and X. Li, “Discrete spectral hashing for efficient
+similarity retrieval,” IEEE Transactions on Image Processing, vol. 28,
+no. 3, pp. 1080–1091, 2018.
+[51] F. Shen, X. Zhou, Y. Yang, J. Song, H. T. Shen, and D. Tao, “A fast op-
+timization method for general binary code learning,” IEEE Transactions
+on Image Processing, vol. 25, no. 12, pp. 5610–5621, 2016.
+[52] C. Mallah, J. Cope, J. Orwell et al., “Plant leaf classification using
+probabilistic integration of shape, texture and margin features,” Signal
+Processing, Pattern Recognition and Applications, vol. 5, no. 1, pp. 45–
+54, 2013.
+[53] S. Wang, X. Liu, X. Zhu, P. Zhang, Y. Zhang, F. Gao, and E. Zhu, “Fast
+parameter-free multi-view subspace clustering with consensus anchor
+guidance,” IEEE Transactions on Image Processing, vol. 31, pp. 556–
+568, 2021.
+[54] Z. Jin, C. Li, Y. Lin, and D. Cai, “Density sensitive hashing,” IEEE
+transactions on cybernetics, vol. 44, no. 8, pp. 1362–1371, 2013.
+[55] Y. Xia, K. He, P. Kohli, and J. Sun, “Sparse projections for high-
+dimensional binary codes,” in Proceedings of the IEEE conference on
+computer vision and pattern recognition, 2015, pp. 3332–3339.
+[56] Y. Gong, S. Lazebnik, A. Gordo, and F. Perronnin, “Iterative quantiza-
+tion: A procrustean approach to learning binary codes for large-scale
+image retrieval,” IEEE transactions on pattern analysis and machine
+intelligence, vol. 35, no. 12, pp. 2916–2929, 2012.
+[57] A. Ng, M. Jordan, and Y. Weiss, “On spectral clustering: Analysis and an
+algorithm,” Advances in neural information processing systems, vol. 14,
+2001.
+[58] J. Liu, C. Wang, J. Gao, and J. Han, “Multi-view clustering via joint
+nonnegative matrix factorization,” in Proceedings of the 2013 SIAM
+international conference on data mining.
+SIAM, 2013, pp. 252–260.
+[59] F. Nie, G. Cai, and X. Li, “Multi-view clustering and semi-supervised
+classification with adaptive neighbours,” in Thirty-first AAAI conference
+on artificial intelligence, 2017.
+[60] Z. Kang, Z. Guo, S. Huang, S. Wang, W. Chen, Y. Su, and Z. Xu, “Mul-
+tiple partitions aligned clustering,” arXiv preprint arXiv:1909.06008,
+2019.
+[61] S. Shi, F. Nie, R. Wang, and X. Li, “Multi-view clustering via nonnega-
+tive and orthogonal graph reconstruction,” IEEE Transactions on Neural
+Networks and Learning Systems, 2021.
+
diff --git a/T9E0T4oBgHgl3EQflQGf/content/tmp_files/load_file.txt b/T9E0T4oBgHgl3EQflQGf/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..96c690061303038bc432216118c30b5a581d0d46
--- /dev/null
+++ b/T9E0T4oBgHgl3EQflQGf/content/tmp_files/load_file.txt
@@ -0,0 +1,2288 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf,len=2287
+page_content='1 Graph-Collaborated Auto-Encoder Hashing for Multi-view Binary Clustering Huibing Wang, Mingze Yao, Guangqi Jiang, Zetian Mi, Xianping Fu Abstract—Unsupervised hashing methods have attracted widespread attention with the explosive growth of large-scale data, which can greatly reduce storage and computation by learning compact binary codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Existing unsupervised hashing methods attempt to exploit the valuable information from sam- ples, which fails to take the local geometric structure of unlabeled samples into consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Moreover, hashing based on auto- encoders aims to minimize the reconstruction loss between the input data and binary codes, which ignores the potential consis- tency and complementarity of multiple sources data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' To address the above issues, we propose a hashing algorithm based on auto- encoders for multi-view binary clustering, which dynamically learns affinity graphs with low-rank constraints and adopts col- laboratively learning between auto-encoders and affinity graphs to learn a unified binary code, called Graph-Collaborated Auto- Encoder Hashing for Multi-view Binary Clustering (GCAE).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we propose a multi-view affinity graphs learning model with low-rank constraint, which can mine the underlying geometric information from multi-view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Then, we design an encoder-decoder paradigm to collaborate the multiple affinity graphs, which can learn a unified binary code effectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Notably, we impose the decorrelation and code balance constraints on binary codes to reduce the quantization errors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Finally, we utilize an alternating iterative optimization scheme to obtain the multi-view clustering results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Extensive experimental results on 5 public datasets are provided to reveal the effectiveness of the algorithm and its superior performance over other state-of-the- art alternatives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Index Terms—Graph-collaborated, Auto-encoder, Multi-view clustering, Binary code I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' INTRODUCTION W ITH the development of information digitization [1], [2], [3] and computer technology, researchers have pro- posed a large number of feature extraction methods to extract features from multiple views of the same sample [4], [5], [6], [7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For example, an image can be extracted as different feature representations by multiple descriptors, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', LBP [8], Gabor [9], HOG [10] and SIFT [11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' However, these multi-view data extracted from different feature descriptors have properties that are large-scale and heterogeneous, which cry out for reliable mining methods to explore the discriminative information from multiple views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In order to effectively process the large- scale data, most existing researches introduce hash methods H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yao, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Ze and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fu are with College of Information Science and Technology, Dalian Maritime University, Liaoning, 116026, China, e-mail: (huibing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='wang@dlmu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' ymz0284@dlmu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' mize- tian@dlmu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' fxp@dlmu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cn).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Both Huibing Wang and Mingze Yao are first authors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jiang is with School of Computer Science and Artifical Intelli- gence, Changzhou University, Jiangsu, 213164, China, e-mail: (guangqi- jiang@cczu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cn).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Mingze Yao, Guangqi Jiang and Xianping Fu are corresponding authors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' due to its fast running speed and economical storage cost.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, hash methods encode the large-scale data by a set of compact binary codes in a low-dimensional Hamming space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, existing hash algorithms have been widely applied to various large-scale visual application tasks, such as cross-modal retrieval [12], object re-identification [13], image detection [14] and multi-view learning [15], [16], [17], [18] etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Considering the effectiveness of binary codes for various vision tasks with large-scale data, several methods have been proposed to explore the more discriminative binary code representation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Over the past few decades, several supervised hashing methods have been proposed, such as Supervised Discrete Hashing (SDH) [19], Strongly Constrained Discrete Hashing (SCDH) [20] and Fast Discriminative Discrete Hash- ing (FDDH) [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Note that while these aforementioned ap- proaches have achieved great performance with hashing, most of them deeply depend on the manual labels, which is time- consuming and less effective process the large-scale unlabeled data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, some unsupervised hashing methods have been proposed to deal with the unlabeled problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The typical unsupervised hashing is Locality-Sensitive Hashing (LSH) [22] which adopts random projections to generate discrete binary codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Based on LSH, Spectral Hashing (SH) [23], Discrete Graph Hashing (DGH) [24] and Scalable Graph Hashing (SGH) [25] have been proposed to explore similar information from the large-scale data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Even though the above methods have effectively learned compact binary codes in an unsupervised manner, most existing hashing methods usually utilize the data from single source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For multi-view data, these hashing methods are difficult to uncover the multi- view information holistically and ignore the consistent and complementary information from different views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Compared with the data from a single source, multi-view data usually contain more compatible and complementary information hidden in different views, which are extracted from same samples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, multi-view clustering methods have been proposed to explore the latent structure of different views and integrate complementary information from multi- view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Kumar et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [26] introduced a co-regularized model to complete spectral clustering with a centroid-based algo- rithm and pairwise algorithm which can mine the underlying structure from original data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [27] proposed a graph-learning method with the rank constraint to integrate different graphs into a global graph for multi-view clustering tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [28] proposed a multi-graph laplacian regularized LRR model, which can separately impose a low- rank constraint on each graph to achieve agreeable results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='02484v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='CV] 6 Jan 2023 2 Besides, Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [29] performed reinforcement learning on the graph of each view and the unified graph of all views by considering the weights of different views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xiao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [30] proposed a graph-based multi-view clustering framework with knowledge elements, which can combine knowledge and language for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Moreover, Shi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [31] proposed a common joint graph learning strategy, which utilizes non- negative constraint to fully explore the structure information from multi-view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' This strategy aims to directly obtain cluster results and avoid post-processing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The above methods mostly measure the distance between features in Euclidean space, while they still need a high computational cost and low efficiency for processing large-scale data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Some researchers proposed multi-view hash methods to learn compact binary codes and utilize efficient XOR operation [32], which can improve the speed and accuracy of the algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [33] proposed a binary function clustering scheme that captures the function semantics as semantic hash- ing to quickly cluster the high degree of similarity samples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tian et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [34] provided a variant of the LRR [35] model to recover the latent structure of original data, which can effectively learn similarity graphs for binary code learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [36] utilized l2,1-norm to learn compact binary codes, which can improve the robustness of the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [37] constructed a novel semantic-rebased model, which adopted a sparse graph setting and rebased the similarity graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Notably, most related hashing works focus on re- trieval tasks, which ignore the complementary information and underlying cluster structure from multi-view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Recently, several hashing algorithms have been proposed to solve large- scale image clustering problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [38] provided a cluster-wise unsupervised hashing framework, which projects the multi-view original data into latent low-dimensional space to learn cluster centroid for searching.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [39] explored a highly-economized algorithm for image clustering, which jointly learned binary representation and binary cluster results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Even though the above methods can process large- scale data effectively and achieve great performance, most of them heavily rely on affinity graphs from original data directly and fail to mine the local structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Meanwhile, some studies simplified the optimization problem by relaxing binary constraints, which may cause quantization errors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, it is essential to compose an effective graph collaboration framework to explore the local geometric information from multiple views and utilize suitable binary constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' To address the above limitations, this paper proposes a novel method, termed as Graph-Collaborated Auto-Encoder Hashing for Multi-view Binary Clustering (GCAE).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' GCAE constructs auto-encoders to learn binary codes for processing multi-view data, which emphasizes collaboratively learning between affinity graphs and auto-encoders to learn a uni- fied binary codes for multi-view clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Firstly, GCAE constructs affinity graphs from each view by imposing a low-rank constraint on the original data, which can preserve essential information and the latent structure from the multi- view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Secondly, to effectively explore the compatible and complementary information from multi-view data, GCAE adopts auto-encoders to collaborate multiple affinity graphs, which aim to learn unified binary codes for clustering and preserve the discrete binary constraint.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Subsequently, GCAE utilizes the matrix factorization strategy to directly obtain cluster results without post-processing, which can avoid error accumulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Finally, an alternating iterative optimization strategy is adopted to update each variable of the objective function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The whole model of GCAE has been shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The major contributions of the proposed method are summarized as follows: We propose Graph-Collaborated Auto-Encoder Hashing for Multi-view Binary Clustering (GCAE), which utilizes affinity graphs and auto-encoders collaboratively to learn compact binary codes for multi-view clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In particular, GCAE imposes the low-rank constraint on graphs to mine essential information effectively and utilizes auto-encoders to collaborate multiple graphs for learning unified binary codes, which can explore comple- mentary information from multi-view data and guide the learning of binary codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, our proposed GCAE directly obtain cluster results to avoid the accumulation of errors caused by post-processing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The remainder of the paper is outlined as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Section 2 introduces the related work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Section 3 presents the proposed GCAE model and the optimization process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Extensive experi- ments including complexity analysis and convergence analysis are conducted to verify our proposed model in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Finally, Section 5 concludes this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' RELATED WORK In this section, we briefly review the related studies about graph-based multi-view clustering and multi-view hashing methods with graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Graph-based multi-view clustering methods mostly aim to integrate information from multiple views and calculate simi- larity graphs in Euclidean distance for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For example, Nie et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [40] proposed a framework based on standard spectral learning which learns weights for multiple graphs automatically without introducing additive parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hou et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [41] presented an automatic method to learn a common similarity graph to characterize the structures across different views and tune balance weights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' However, the above methods require an additional clustering step to obtains the final clusters by utilizing K-means [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In order to avoid the impact of post-processing for obtain the cluster results, Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [29] proposed a model which can produce clusters directly without post-processing for clustering and construct each view graph and fusion graph simultaneously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [43] utilize Hadamard product to integrate multiple graphs into a global graph which can recover the graph structure effectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [44] proposed a unified framework for jointly learn- ing multiple similarity graphs and spectral embedding, which can obtain cluster results in a unified framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Although the above methods achieved great clustering results, they directly build graphs by original data, which may make the learned similarity graph inaccurate and ignore the underlying cluster structure of multi-view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Meanwhile, the real-value multi- view data requires high computational costs for generating 3 Multi-view Data View 1 View 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' View M Kernelized Representation of Multiple Views Clustering Result × Nonlinear Kernel Mapping 1 Clustering Centroids Indicator Vectors Affinity Graphs Expend to Multiple Views Z1 Z2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Projection Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Feature ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Encoder ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Feature ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Encoder ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Inverse Projection Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Feature ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Decoder ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Feature ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Decoder ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Projection Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Feature ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Encoder ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Inverse Projection Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Feature ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Decoder ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Common Binary Codes Matrix ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Collaborative Learning ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Collaborative Learning ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Binary Codes Learning by Auto-encoders ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='One Step Clustering ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Binary Matrix Factorization ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='View 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='View 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Kernelized Representation ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='ZM ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1: The whole model of Graph-Collaborated Auto-Encoder Hashing for Multi-view Binary Clustering (GCAE), which learns affinity graphs by low-rank constraint to integrate specific information from multiple views and adopts auto-encoder to generate unified binary code for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' graphs with large-scale multi-view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Some researchers explore hash methods which utilize binary codes rather than real-value features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, multi-view hashing methods are widely used to integrate large-scale data with the recent advances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Existing multi-view hashing methods with graphs can be roughly divided into supervised and unsupervised methods based on the usage of semantic presentation (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', class labels).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For supervised methods, Jin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [45] construct a semantic graph by jointly taking the semantic presentation and the local similarity structure into consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Guan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [46] proposed a supervised learning model to construct similarity graphs that capture the intrinsic manifold structure from semantic supervision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Despite these supervised methods can achieve great performance, it is not practical to label infor- mation manually for large-scale data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [24] utilized anchor graphs to capture the latent structure inherent in a given massive dataset and calculated hash codes in Hamming space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xiang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [47] proposed a novel hashing method that adopted a quantization regularization term to reduce the distortion error during constructing similarity graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, Fang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [48] jointly learned the intra-modal similarity graph and reconstructed the similarity graph by symmetric nonnegative matrix factorization and then utilized binary code inner product to learn binary codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The above approaches have obtained good results, but relaxing the discrete constraint on binary codes will cause quantization errors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Meanwhile, for the multi-view clustering tasks, existing multi-view hashing methods have not effectively mined the underlying cluster structure from different views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' GRAPH-COLLABORATED AUTO-ENCODER HASHING In this paper, boldface lowercase letters and boldface up- percase letters are used to denote the vectors and matrices, receptively, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', x and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For any matrix X, xij means Table I: The Descriptions of Some Important Formula Sym- bols.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Scalar N N ∈ R Vector a a ∈ RN ai i-th row av i i-th row with v-th view 1 all-1 vector 0 all-0 vector ||a||2 l2 norm as �� N i=1a2 i Martix Av Av ∈ RN×dv Ai,j (i, j)-th element I the identity matrix tr(A) the trace of matrix A AT the transpose of matrix A σ σ = diag(S) ||A||F Frobenius norm as �� i,jA2 i,j, ∥σ∥2 ||A||∗ nuclear norm as tr( √ ATA) φ(Av) kernelized representation of (Av) the (i, j)-element of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides some important mathematic notations are listed in Table I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For given multi-data sets, X = {X1, X2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', Xv} where Xv ∈ RN×dv is the v-th view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' dv denotes the feature dimension and N is the number of sample.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We propose a novel model of graph-collaborated auto- encoder hashing for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we jointly learned the affinity graphs from each view and constructed auto- encoders for the unified hash code learning which can min- imize reconstruction loss between affinity graphs and hash code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, our method utilized matrix factorization strat- egy to obtain cluster results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' r G me4 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Multi-view Affinity Graph Learning For given data matrix X = {X1, X2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', Xv}, where Xv presents the v-th view data matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Different from existing methods that directly utilize original multi-view data for gen- erating multiple graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' GCAE firstly adopts nonlinear kernel mapping for each view, which aims to unify the dimension of multi-view data so that we can learn the underlying geometric structure from data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Inspired by [32], we employ the nonlinear RBF kernel mapping method for each view as follows: φ(Xv) = [exp(−∥xv 1 − av 1∥2/η), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', exp(−∥xv N − av t ∥2/η)]T (1) where η is the kernel width, φ(Xv) ∈ RN×t indicates the t- dimensional nonlinear mapping from the v-th view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, to ensure that the original data structure is not destroyed after nonlinear projection, we randomly select t anchor samples av t from the v-th view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' After φ(Xv) yielded by Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1, the affinity graph Zv is constructed by processing each view of φ(Xv).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Above this, we propose a new variant of the LRR [49] model that can make the learned affinity graph retain the important information of the original data instead of noise and redundant information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Then, we give the following model: min Zv M � v=1 ∥φ(Xv) − Zvφ(Xv)∥2 F + ∥Zv∥∗ (2) where Zv ∈ RN×N represents the learned affinity graph for v-th view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' ||Zv||∗ indicates the nuclear norm which calculates the sum of singular values for low-rank constraint.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Minimizing the objective function in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2 aims to preserve important information while exploring the low-rank structure from the original data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Graph-collaborated Auto-Encoders Hashing for Clustering In order to encourage collaborative learning of affinity graphs from different views, we proposed an auto-encoders hashing by the multi-graphs model which can effectively mine the consistency and complementarity information and generate binary code for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Different from the existing hash methods, the auto-encoders hashing model preserves the discrete constraint and the uncorrelated constraint on binary bits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' min Wv,B,Zv M � v=1 (∥WvZv − B∥2 F + ∥Zv − WvT B∥2 F ) s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI (3) Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3 aims to minimize the difference between low-rank affinity graphs mapping and discrete hash code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' b represents binary bits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The decorrelation and code balance constraints B ∈ {−1, +1}b×N, BBT = NI are imposed to generate mutually independent binary codes and reduce quantization errors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wv presents the projection matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For simplicity, we utilize the transpose matrix of Wv to replace the inverse map- ping matrix in auto-encoder, and add a constraint WvWvT = I on projection matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Meanwhile, the regularization term ||Wv||2 F is the constant due to ||Wv||2 F = tr(WvWvT ) = tr(I) = const.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Finally, we construct a matrix factorization model with the generated unified binary codes, which can avoid the suboptimal results caused by the two-step clustering methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Above all, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3 can be reformulated as follows: min Zv,Wv,B,Q,H M � v=1 (∥WvZv − B∥2 F + ∥Zv − WvT B∥2 F )) + λ ∥B − QH∥2 F s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI, QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N, c � i=1 his = 1 (4) where λ is the regularization parameter and c is the number of clusters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Q and H represent the clustering centroids and cluster indicator matrix, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Meanwhile, in order to generate the efficient binary code and maximize the information of each bits for clustering, we add the balance constraint on clustering centroids Q, which can produce efficient code and make our model to adapt binary clustering task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Overall Objective Function of GCAE In order to collaborating affinity graphs and auto-encoders to learn a unified binary code for clustering, we summarize the above graph-collaborated auto-encoders hashing and the low-rank affinity graphs learning, which are both crucial to multi-view clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The above consideration can be fulfilled as follows: min L(Zv, Wv, B, Q, H, pv) = M � v=1 ∥φ(Xv) − Zvφ(Xv)∥2 F + ∥Zv∥∗ � �� � Multi−view Affinity Graph Learning + M � v=1 (pv)k(∥WvZv−B∥2 F +∥Zv−WvTB∥2 F )+λ∥B−QH∥2 F � �� � Auto−Encoders Hashing by Multi−Graphs for Clustering s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI, QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N, M � v=1 pv = 1, pv > 0, c � i=1 his = 1 (5) where pv indicates the normalized weighting coefficient, which aims to balance the affinity graphs from each view ac- cording to the contribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' we also add the constraint pv > 0 to ensure nonnegative vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Above all, we summarize multi-view affinity graph learning and auto-encoders hashing by multi-graphs in a unified framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The proposed GCAE model learns the low-rank affinity graph from each view, which can effectively preserve important information from original data to improve the quality of affinity graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And then, the proposed model utilizes binary matrix factorization model for the unified binary codes, which can effectively gen- erate cluster results in a one-step model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' By jointly optimizing 5 the above equation, we obtain the graph-collaborated binary code representation to obtain the cluster results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We propose a novel optimization algorithm for the objective function Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5 in the following section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Optimization Process for GCAE In this section, we describe the optimization process for GCAE in detail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It is obvious that Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5 is a nonconvex optimization problem that can not directly obtain the holistic optimal solution in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We develop an auxiliary matrices strategy, which separates the problem into two sub-problems include low-rank affinity graphs learning and auto-encoders hash for obtaining optimized cluster results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Based on the aux- iliary matrices strategy, we introduce two auxiliary matrices to relax the nuclear norm and control the rank of affinity graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And then the proposed auto-encoders utilize the affinity graphs from each view to generate binary code for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Finally, we develop an iterative alternative strategy, which alternatively updates each variable when fixing others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The proposed auxiliary matrices strategy introduces the multiplication of two auxiliary matrices Fv and Gv (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', Zv = FvGvT ) to replace the low-rank affinity graph Zv, and then problem Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2 can be converted as: min Fv,Gv ���φ(Xv) − FvGvT φ(Xv) ��� 2 F (6) where Fv ∈ RN×r and Gv ∈ RN×r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' r is a parameter that we use to relax nuclear norm and control the rank of Zv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And this strategy is based on the fact that rank(Zv) = rank(FvGvT ) ≤ min(rank(Fv), rank(Gv)) ≤ r (generally r ≪ N).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The updating process of Fv and Gv are shown as follows: Updating Fv: Based on the operational rules of matrix trace, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 6 can be unfolded as: O(Fv) = tr((φ(Xv) − FvGvT φ(Xv))(φ(Xv) − FvGvT φ(Xv))T ) (7) then we calculate Fv iteratively by setting the derivation ∂O(Fv) ∂Fv = 0: ∂O(Fv) ∂Fv =2FvGvT φ(Xv)φ(Xv)T Gv − 2φ(Xv)φ(Xv)T Gv = 0 (8) Obviously, the close solution of Fv is written as: Fv = φ(Xv)φ(Xv)T Gv(GvT φ(Xv)φ(Xv)T Gv)−1 (9) Updating Gv: With Fv fixed, we can solve the opti- mization problem of Gv which is similar with above method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The solution of Gv can be easily obtained: Gv = Fv(FvT Fv)−1 (10) Based on the above solution of Fv and Gv, we sum- marize the above iteratively method in Algorithm 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' No- tably, in order to avoid obtaining the singular value, we add a small smooth item in optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' That is, Fv = φ(Xv)φ(Xv)T Gv(GvT φ(Xv)φ(Xv)T Gv +θI)−1 and Gv = Algorithm 1 Low-Rank Affinity Graphs Learning Algorithm Input: Kernelized dataset φ(Xv);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' the parameter r, and the maximum number of iterations Iter Output: Fv, Gv 1: set t = 1 and randomly initialize F(0)v and G(0)v 2: while t < Iter do 3: Compute F(t)v according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 9 4: Compute G(t)v according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 10 5: t = t + 1 6: end while 7: return F(t)v and G(t)v Fv(FvT Fv + θI)−1 where θ is set in range of [1e−4, 1e−6] in our paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' After the above low-rank affinity graphs learning with auxiliary matrices strategy, we have rewritten the formula to auto-encoders hash for clustering as Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In order to obtain the optimal solution of Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 11, we propose an iterative optimization method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The proposed method can effectively maintain discrete constraints for binary code and get more efficient binary code for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' min Zv,B,Q,H,pv M � v=1 (∥FvGvT − Zv∥2 F + (pv)k(∥WvZv − B∥2 F + ∥Zv − WvT B∥2 F )) + λ ∥B − QH∥2 F s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' WvWvT = I, B ∈ {−1, +1}b×N, BBT = NI, QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N, M � v=1 pv = 1,pv > 0, c � i=1 his = 1 (11) Updating Zv: By fixing all variables but Zv, problem(11) reduces to: min Z M � v=1 (∥FvGvT − Zv∥2 F + (pv)k(∥WvZv − B∥2 F + ∥Zv − WvT B∥2 F )) (12) In order to minimize the above equation, we convert the Frobenius norm in the above equation to the trace form of the matrix which is convenient for derivation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And then we take the derivative of the trace of the matrix with regard to Zv as follows: ∂LZv ∂Zv = M � v=1 tr(ZvZvT − 2FvGvT ZvT +(pv)k(WvZvZvTWvT +ZvZvT −4WvTBZvT )) (13) whose solution can be easily achieved by setting the derivation to 0: Zv = ((pv)kWvT Wv + I + (lv)kI)−1(FvGvT + 2(lv)kWvT B) (14) 6 Updating Wv: In this stage, other variables are fixed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We take the constraint WvWvT = I in consider, the whole loss function related to Wv in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 11 can be rewritten as: LWv = min Wv M � v=1 (∥WvZv − B∥2 F + ∥Zv − WvT B∥2 F ) = max tr(WvZvBT ) s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' WvWvT = I (15) where we also need to consider the condition BBT = I which is used during the optimization process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we firstly convert Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 15 to trace of matrix form, and then we utilize the SVD algorithm [50] to solve the optimization problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wv = SDT (16) where S and D are the left and right singular vectors of the compact Singular Value Decomposition (SVD) of ZvBT .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Updating B: The common gradient method is not suit- able for solving the discrete binary codes B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We rewrite Eq11 related to B as follows: max B tr(BT (2 M � v=1 (pv)kWvZv + λQH)) s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' B ∈ {−1, +1}b×N, BBT = NI (17) We need to preserve constraints of B which can generate compact and effective binary code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Thus, the optimal binary code B can be obtained as follows, and sgn(·) means sym- bolic function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' B = sgn( M � v=1 (2(pv)kWvZv) + λQH) (18) Updating Q and H: In this part, we iteratively op- timize the binary clustering model by matrix factorization in Hamming space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' By removing the irrelevant terms, the problem can be rewritten as: min Q,H ∥B − QH∥2 F s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' QT 1 = 0, Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N, c � i=1 his = 1, B ∈ {−1, +1}b×N (19) We simply reformulate Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 19 to: min Q,H ∥B − QH∥2 F + ρ ���QT 1 ��� 2 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Q ∈ {−1, +1}b×c, H ∈ {0, 1}c×N, c � i=1 his = 1 (20) which is equal Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 19 with adaptively large ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The proposed matrix factorization strategy iteratively optimizes the cluster centroids and indicators as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Updating Q : Due to the discrete constraint on Q, we utilize the discrete proximal linearized minimization (DPLM) [51] method, which can effectively obtain high-quality binary solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' With H fixed, the optimal Q can be obtained as follow: minLQ = ∥B − QH∥2 F + ρ ���QT 1 ��� 2 = −2tr(BT QH) + ρ ���QT ��� 2 + con s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Q ∈ {−1, +1}l×c (21) According to the DPLM, in the t + 1-th iteration Q can be updated as: Qt+1 = sgn(Qt − 1 µ∇LQt) (22) where ∇LQt represents the gradient of LQ Updating H : We utilize vectors-based method to optimize the indicator matrix H, the solution to hi,j can be easily obtained by ht+1 i,j = � 1, j = arg minsD(bi, qt+1 s ) 0, otherwise (23) where D(bi, qt+1 s ) is the distance between i-th binary code bi and the s-th cluster centroid qs in Hamming space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Notably, we use binary code rather than real-value to calculate Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 23, which adopt Hamming distance that can save time compared to the Euclidean distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Updating pv: For simplicity, we let av = ∥WvZv − B∥2 F + ∥Zv − WvT B∥2 F .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Based on the attributes of different views, the weighting coefficient pv can be equivalent as following equation: min pv M � v=1 (pv)kav s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' M � v=1 pv = 1,pv > 0 (24) Due to the constraint on pv, we can solve this problem by the Lagrange multiplier method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' By setting the Lagrange multiplier Γ, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 24 can be rewritten as: min L(pv, Γ) = M � v=1 (pv)kav − Γ( M � v=1 pv − 1) (25) and then, we calculate the partial derivative of L(pv, Γ) about pv and Γ, we can get: � � � ∂L ∂pv = k(pv)k−1av − Γ ∂L ∂Γ = M � v=1 (pv)k − 1 (26) Therefore, we set partial derivative to zero which can get: pv = (av) 1 1−k M � v=1 (av) 1 1−k (27) We have presented the whole optimization process for Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And we summarize the whole process in Algorithm 2, which iteratively updates the variables until convergence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 7 Table II: The Comparison Result with Hash Method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Datasets Metrics SH DSH SP ITQ SGH RSSH RFDH HSIC BMVC GCAE 100leaves ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4713 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4494 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4750 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4569 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5088 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3631 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4513 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6563 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4981 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8888 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7214 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7320 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7240 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7405 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7579 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6203 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7161 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8245 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7291 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9426 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4950 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4988 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5138 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4944 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5338 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3888 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4838 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6788 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5331 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9031 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3471 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3219 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3312 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3324 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3996 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2209 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3234 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5431 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3057 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8366 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3184 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2653 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2942 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2528 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3626 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2094 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2708 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5128 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2547 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8166 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3404 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3141 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3240 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3240 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3933 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2131 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3158 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5385 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2978 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8350 Caltech-101 ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1747 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1610 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2077 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2472 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2256 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2860 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2196 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2429 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2930 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3005 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3252 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3600 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4034 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4404 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4349 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4846 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4424 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4451 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4900 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4711 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3089 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3439 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3878 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4231 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4051 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4729 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4228 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4125 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4907 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4421 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1486 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1250 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1738 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2311 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2089 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2533 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2081 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2055 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2465 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3023 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2604 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1975 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2897 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3544 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3085 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4071 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3419 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3430 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4147 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4229 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1347 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1091 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1596 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2167 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1935 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2406 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1943 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1919 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2336 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2880 Cifar-10 ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1708 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2189 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2275 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2245 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2205 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1960 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2252 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2153 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2350 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2498 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0282 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0938 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0979 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0987 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0995 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0659 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1011 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0920 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1016 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1020 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1727 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2271 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2285 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2297 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2218 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2046 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2333 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2236 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2368 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2549 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1137 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1462 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1342 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1266 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1458 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1316 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1560 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1441 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1590 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1597 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1129 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1520 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1495 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1544 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1408 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1284 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1457 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1420 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1533 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1563 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0145 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0624 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0584 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0610 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0600 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0325 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0582 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0476 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0618 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0642 SUNRGBD ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1164 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1577 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1925 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1895 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1823 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1613 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1738 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1616 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1379 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2423 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1319 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2198 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2180 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2108 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2172 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1980 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2032 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2202 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1545 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2207 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2441 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3271 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3418 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3366 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3433 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3279 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3332 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3524 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2803 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3435 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0650 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1033 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1223 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1274 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1184 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0976 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1145 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1059 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0822 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1541 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1193 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1861 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1722 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1802 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1822 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1869 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1901 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2013 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1550 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1909 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0309 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0699 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0895 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0951 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0859 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0661 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0823 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0744 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0496 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1076 Caltech256 ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0622 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0782 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0856 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0865 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0826 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0937 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0783 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0678 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0932 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1071 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2557 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2866 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2947 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2717 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2969 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3058 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2855 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2350 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3185 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2871 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1096 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1255 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1399 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1386 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1363 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1512 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1278 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1064 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1534 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1415 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0465 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0644 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0635 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0975 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0623 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0805 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0610 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0402 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0745 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1145 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0527 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0681 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0658 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0950 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0649 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0932 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0644 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0305 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0811 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1037 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0415 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0591 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0582 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0920 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0569 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0758 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0558 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0327 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0695 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1087 Algorithm 2 GCAE Algorithm Input: Dataset by kernelized φ(Xv);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' the parameter λ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' the auxiliary matrices Fv and Gv;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Output: Binary code B;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' cluster indicator matrix H;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1: Randomly initialize binary code B;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' affinity graph Zv;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' weights for different view pv;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' binary code length b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' project matrix Wv 2: repeat 3: Update Zv according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 12 4: Update Wv according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 16 5: Update B according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 18 6: Update Q and H according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 22 and Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 23 7: Update pv according to Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 27 8: until convergance IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' EXPERIMENTS In this section, we describe the datasets and comparison methods to verify the effectiveness of the proposed GCAE for clustering tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We evaluated the performance of GCAE by comparing several hash methods and multi-view clustering methods on 5 widely used datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Moreover, we analyze the parameter sensitivity of our model, which can affect the fluctuations of results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We also summarize the running times on all datasets and evaluate the convergence of our model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' All the experiments are conducted using Matlab 2020a on a Windows PC with Intel 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8-GHz CPU and 64-GB RAM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Experimental settings In this part, we describe the utilized datasets and the com- parison methods in detail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We also introduce some evaluation metrics which aim to verify the effectiveness of GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1) Datasets: To evaluate the clustering performance of the proposed model and comparison models, five widely-accepted multi-view datasets are selected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The details of the selected datasets are as follows: Caltech1011 contains 9144 samples of 101 objects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 6 pub- licly available features are utilized as multiple views, including Gabor feature with 48 dimension, LBP feature with 928 dimension, GIST feature with 512 dimension, CENTRIST feature with 254 dimension, wavelet moments(WM) with 40 dimension and HOG feature with 1984 dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Caltech2562 contains 30607 images by 4 kinds of features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It includes 256 classes with more than 80 samples per class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, each image is represented by color histogram with 729 dimension, Gist feature with 1024 dimension, HOG feature with 1152 dimension and features based on convolution network with 1440 dimension, which are four different types of presentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cifar-103 is composed by 60000 tiny color images in 10 kinds of classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we represent the features by 1http://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='vision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='caltech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu/ImageDatasets/Caltech101/ 2http://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='vision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='caltech.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu/Image-Datasets/Caltech256 3https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='toronto.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu/kriz/cifar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 8 DSD feature with 220 dimension, HOG feature with 512 dimension and Gist feature with 768 dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, a subset of this dataset is selected in the experiment that contains 10000 samples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 100leaves4 [52] contains 1600 samples from each of 100 plant species, which form the UCI repository with three different views, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', Shape, Texture and Margin features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' SUNRGBD5 contains 10335 indoor scene images by 3D cameras in 45 classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Similar with [53], we utilize two views with 4096 dimension and features extracted by different convolution network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2) Comparing Algorithms and Evaluation Metrics: In our experiments, we evaluate the performance of GCAE by com- paring several state-of-the-art multi-view algorithms and hash algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we utilize seven single-view hash methods and two multi-view hash clustering algorithms, in- cluding SH [23], DSH [54], SP [55], ITQ [56], SGH [25], RSSH [34], RFDH [36], HSIC [39], BMVC [32].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For single- view hash methods, we utilize K-means algorithm to process the generated binary code for clustering and we preserve the best result of each view clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, we adopt eight real-value multi-view clustering algorithms as comparing algorithms, including K-means [42], SC [57], Co-regularize [26], AMGL [40], Mul-NMF [58], MLAN [59], mPAC [60], GMC [?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We utilize the source code from the author’s public homepage for comparing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Notably, we set the length of binary code as 128-bits in our experiments, which can effectively preserve critical information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In order to generally evaluate the superiority of our method, we utilize six widely used evaluation metrics, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', clustering accuracy(ACC), Normalized Mutual Information(NMI), Pu- rity, F-score, Precision and ARI [?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' ], [61].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For all algorithms, better performance is improved by the higher value of evalu- ation metrics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Experimental Results and Analysis In this section, we conducted experiments on 5 multi-view datasets to show the superiority of the proposed GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The detailed clustering results are shown in Table II, III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We utilize the bold values to represent the best performance in each table corresponding to the dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, this section also introduced the analysis of parameter sensitivity, which can reflect clustering results with different parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Finally, we provide the complexity analysis and convergence analysis of the proposed model to verify the stability of GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1) Comparison with Hash Methods: In this section, we conduct experiments with hash methods on five multi-view datasets to verify the performance of our proposed model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We compare GCAE with seven single-view hash methods and two multi-view hash clustering methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In single-view hash methods, we generate cluster results by adopting k-means algorithm to finish cluster task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we adopt K-means algorithm with cluster binary code to obtain sample labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The results with different datasets are reported in Table II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 4https://archive.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='ics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='uci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu/ml/dataset 5http://rgbd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='princeton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='edu/ We summarized the clustering performance with compared methods on all multimedia datasets in Table II, and the best results are highlighted in bold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' As shown in Table II, we can obvious that GCAE obtains the best clustering accuracy on five multi-view datasets compared with hash methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we proposed GCAE method evidently outperforms other state- of-the-art methods for 100leaves and Cifar-10 datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For the 100leaves dataset, the performance of GCAE improves around 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2%, 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8%, and 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4% with terms of ACC, NMI, and Purity over the second-best method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Moreover, for the Cifar-10 dataset, the performance improvements over the second-best method are 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5%, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4%, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='81% with terms of ACC, NMI, and Purity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The experiments demonstrate that the multi-view hash clus- tering methods get better results than single-view methods, which proves multi-view methods can exploit complementary information from multi-view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For single-view methods, we perform clustering on each views and select the best results to report in the above table.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Generally speaking, multi-view methods take the complementary information and consis- tent information between multi-view data into consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, multi-view methods can get better performance and single-view methods can not obtain satisfactory performance in most situations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The results in Table II also verify the low- rank affinity graphs construction is a vital part of our proposed GCAE, which aims to effectively generate binary code with essential information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Overall, the clustering performance about our proposed model outperforms the other hash methods in most situations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The phenomenon denotes that GCAE can effectively preserve essential information by auto-encoders structure and keep the discrete constraint on binary code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, GCAE has the ability to integrate comprehensive information from multi- view data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2) Multi-view Methods Experimental Results and Analysis: In this section, we present the detailed comparison clustering results with multi-view clustering methods in Table III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' As shown in this table, we compare GCAE with eight real- value clustering methods and two hash clustering methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Moreover, for single-view methods i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', K-means and SC, we concatenates all multiple views into one view for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' As shown in Table III, it is clear that GCAE obtains the best clustering accuracy with the comparison results on all multi-view datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, the proposed GCAE model outperforms all other methods on ACC, NMI, Purity, F-score, Precision, and ARI in most situations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We can obviously find the performance improvements over the second-best method are 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9%, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4% and 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1% respectively in metrics of ACC, NMI, and Purity for 100leaves dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Moreover, for the Cifar- 10 dataset, our proposed model can improve 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5%, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4%, and 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='81% in the above metrics which is compared with the second-best method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We conducted the experiments with comparing real-value based multi-view clustering method and hash methods, which aims to verify the stability of clustering in Hamming space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It is clear that hash methods can obtain satisfactory perfor- mance in most situations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' This is because real-value methods utilize Euclidean distance to measure two samples, which have 9 Table III: The Clustering Results on Five Datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Datasets Metrics k-means SC Co-re-p Co-re-c AMGL Mul-NMF MLAN mPAC GMC HSIC BMVC GCAE 100leaves ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6200 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4894 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7253 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7939 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7631 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8694 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7356 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8238 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4338 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6563 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4981 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8888 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8284 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7649 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8835 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9257 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9065 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9288 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8848 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9292 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7014 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8245 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7291 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9426 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6550 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5575 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7565 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8272 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8063 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8981 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7625 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8506 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5350 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6788 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5331 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9031 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5233 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2144 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6595 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7558 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4513 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8236 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6583 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5042 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3145 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5431 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3057 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8366 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4689 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1307 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6107 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7011 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3086 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7832 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6082 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3521 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3744 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5128 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2547 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8166 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5183 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2021 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6560 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7533 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4437 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8219 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6548 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4974 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3070 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5385 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2978 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8350 Caltech-101 ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1331 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1751 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2611 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2587 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1476 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1908 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2274 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1950 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2672 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2429 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2940 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3005 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3078 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3207 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4752 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4912 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3757 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3519 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4564 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3446 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4408 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4451 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4900 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4711 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2907 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3107 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4622 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4664 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1681 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3184 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4401 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3012 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3497 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4125 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4907 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4421 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0985 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1362 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2202 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2226 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0338 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0470 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1930 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0496 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2658 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2055 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2465 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3023 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1351 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1440 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3954 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3999 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0175 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0248 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3185 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0261 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2337 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3430 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4147 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4229 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0796 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1126 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2078 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2103 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0155 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0068 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1790 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0042 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2475 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1919 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2336 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2880 Cifar-10 ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2036 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1712 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2169 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2084 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2232 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1217 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2304 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1038 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2312 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2153 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2350 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2498 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0891 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0773 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0944 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0903 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0845 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0204 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0919 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0066 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1014 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0920 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1016 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1020 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2055 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1751 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2214 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2180 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2276 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1235 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2522 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1043 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2462 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2236 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2368 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2549 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1529 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1443 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1593 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1477 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1587 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1582 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1570 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1511 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1499 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1441 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1590 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1597 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1344 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1080 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1453 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1393 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1393 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1005 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1537 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0999 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1495 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1420 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1533 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1563 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0443 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0155 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0561 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0467 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0594 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0012 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0612 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0603 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0607 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0476 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0618 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0642 SUNRGBD ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1859 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1060 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1824 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1829 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1010 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1346 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1898 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1277 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2155 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1616 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1379 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2423 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1866 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0084 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2109 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2161 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1883 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0976 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2101 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0728 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2204 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2202 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1545 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2207 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3022 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1087 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3274 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3360 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1119 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1583 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3257 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1415 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2303 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3524 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2803 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3435 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1293 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1213 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1172 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1221 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0637 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1200 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1209 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1215 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1387 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1059 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0822 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1541 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1057 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0646 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1849 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1748 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0364 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0645 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1822 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0650 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1007 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2013 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1550 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1909 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0976 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0323 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0865 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0916 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0262 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0251 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0891 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0008 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1026 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0744 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0496 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1076 Caltech-256 ACC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0845 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0924 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0854 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1025 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0467 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0612 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0761 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0904 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0723 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0678 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0932 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1071 NMI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2748 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2764 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2997 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2786 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1070 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1486 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2820 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2198 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1347 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2350 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3185 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2871 Purity 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1296 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1339 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1412 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1522 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0415 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1060 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1343 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1381 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1008 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1064 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1534 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1415 F-score 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0562 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0628 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0596 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0713 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0466 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0106 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0556 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0498 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0355 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0402 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0745 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1145 Precision 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0522 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0855 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0692 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0813 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0458 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0053 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0570 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0412 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0515 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0305 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0811 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1037 ARI 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0501 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0846 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0549 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0689 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0488 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0011 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0501 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0361 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0834 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0327 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0695 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1087 1e-9 1e-8 1e-7 1e-6 1e-5 1e-4 1e-3 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='35 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 ACC SUNRGBD Caltech-101 Cifar-10 (a) Clustering accuracy with λ 3 4 5 6 7 8 k 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='35 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 ACC SUNRGBD Caltech-101 Cifar-10 (b) Clustering accuracy with k 200 400 600 800 1000 1200 1400 t 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='35 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 ACC SUNRGBD Caltech-101 Cifar-10 (c) Clustering accuracy with t 1e-8 1e-7 1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='35 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 ACC SUNRGBD Caltech-101 Cifar-10 (d) Clustering accuracy with θ Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2: Clustering accuracy with parameter tuning for the Caltech-101, Cifar-10 and SUNRGBD datasets (a)λ (b)k (c)t (d)θ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' problems of low efficiency and high computational complexity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And hash methods learn binary code and obtain cluster results in Hamming space, which can improve calculation efficiency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The adoption of the affinity graph is more conductive to pre- serving vital information of mining original data, but utilizing hash methods only improves significantly in computational complexity, which can not significantly improve evaluation metrics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It is worth noting that the proposed GCAE obtain the best cluster accuracy and satisfactory performance on other evaluation metrics for five multi-view datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' This is because GCAE can explore consistency and complementary informa- tion in multi-view data structure by learning low-rank affinity graphs, which learn high-quality collaborative representation well and eliminates some redundant and noise information in the original real-valued features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, GCAE can also im- prove the computational complexity of algorithm significantly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Parameter Sensitivity and Complexity Analysis In this section, we discover that the fluctuation of several parameters has a significant influence on the experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In order to evaluate the sensitivity of the parameters in the proposed GCAE, we select 100leaves, Caltech-101, Cifar-10, and SUNRGBD as the benchmark datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For GCAE, there are six parameters to be tuned, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', λ, k, t, θ, b and r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The above parameters represent regularization parameter for binary clustering, the power of normalized weighting coefficient, the dimension of nonlinear kernel, the small smooth item, binary bits, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We firstly conducted experiments on the Caltech-101, Cifar-10, and SUNRGBD datasets to show the influence of parameters λ, k, t and θ change trend on the results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, we calculate the cluster accuracy values and display them in line charts, which can intuitively find the influence of parameters changing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2 shows the different change trends on three benchmark datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2(a), we can observe the different λ values have effected 10 8 16 32 64 128 Length of binary code 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9 1 Performance ACC NMI Purity (a) 100leaves 8 16 32 64 128 Length of binary code 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6 Performance ACC NMI Purity (b) Caltech-101 8 16 32 64 128 Length of binary code 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='35 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 Performance ACC NMI Purity (c) Cifar-10 Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3: From left to right, the performance with binary code length tuning for the 100leaves, Caltech-101 and Cifar-10 datasets, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' (a) 100leaves 04 。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' uuv b r (b) Caltech-101 (c) Cifar-10 Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 4: Parameter tuning with respect to r and b for the 100leaves, Caltech-101 and Cifar-10 datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' the cluster accuracy notably on three datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We conduct experiments with different parameters λ selected from the range of [1e − 9, 1e − 3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It is obvious that the change trend of GCAE on the three datasets and GCAE can achieve ideal performance when λ is set to an appropriate value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, it is obvious that the parameter k can effect cluster accuracy values selected from the range of [3, 8] in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2(b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The proposed GCAE can achieve excellent performance with a good selection of k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' As shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2(c) and Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2(d), the change trend of t and θ do not influence cluster results obviously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It is notable that GCAE can still achieve a relatively stable and good performance with the change in the dimension of nonlinear kernel t and the small smooth item θ, which are selected in range of [200, 1400] and [1e−8, 1e−1], respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' As shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3, we adopt comparison experiments of the length of binary code to verify our proposed GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, we change the binary code length and calculate ACC, NMI, and Purity on 100leaves, Caltech-101 and Cifar- 10 datasets to show the performance of GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Above the figure, we select 128-bits binary code in GCAE, which aims to fully preserve essential information for clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, in order to demonstrate the relation between the rank of affinity graphs and the length of binary code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We conducted a three- dimensional statistical graph of clustering results by adjusting the values of r and b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The statistical results of 100leaves, Caltech-101 and Cifar-10 datasets are shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 4, which clustering performance as ACC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We can obviously find GCAE can obtain excellent performance with a good selection of r and b, which are designed in the range of [20, 200] and [8, 128] respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In conclusion, we select 128-bits binary codes and r = 100 on these datasets, which achieves the best clustering performance with the proposed GCAE algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We summarized the overall algorithm of the proposed GCAE method in Algorithm 1 and Algorithm 2, which present low-rank affinity graphs learning and graph-collaborated auto- encoder hashing respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The computational burden of GCAE mainly consists of the above two sub-process, and we provide the analysis of GCAE in detail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Specifically, the two closed-form solutions, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 9 and Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 10 cost O(Ndvr + Nr2 + r3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Therefore, the whole computational complexity of Algorithm 1 is O(N × max(dv, r) × r × Iter), where we set the Iter to 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For Algorithm 2, we adopt the iterative optimization method to obtain the solution for each parameter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' When updating Zv, which is the most time-consuming part of GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It costs O(N 3 + N 2b + N 2r) because it needs to calculate the inverse of the matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And then, we need O(N 2b) for updating W v, which contains SVD operation and matrix multiplication.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' After that, in order to update B and 23, we need O(N 2b + Nbc) which includes Frobenius norm and ADPLM method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Besides, solving Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 22 and Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 23 needs O(Nc) and updating lv costs O(rN) for each iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Overall, the whole computational complexity of GCAE is O((N 3 +3N 2b+N 2r +Nbc)p+Nmax(dv, r)rIter), which p is the number of epoch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' For clarity, we summarized the running time of GCAE for each dataset in Table IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Table IV: Running time of GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Datasets 100leaves Caltech-101 Cifar-10 SUNRGBD Caltech-256 Times(s) 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='16 1158.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='71 1120.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 1488.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='33 47618 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Convergence Analysis In order to verify the proposed iteratively optimization al- gorithm for GCAE is convergent, we conduct the convergence 3 6- 120 4 203 2 200 0 180 160 128 140 120 64 100 32 80 门 16 60 40 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 200 0 180 160 128 140 64 120 32 100 80 I b 16 60 8 40 200.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='6 CC 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='200 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='180 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='160 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='128 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='140 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='64 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='120 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='100 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='80 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='I ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='b ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='2011 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Iteration step ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='181 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='182 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='183 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='184 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='185 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='186 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='187 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='188 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Objective value ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='(a) 100leaves ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Iteration step ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1785 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1790 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1795 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1800 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1805 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='1810 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Objective value ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='(b) Caltech-101 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Iteration step ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='939 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='940 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='941 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='942 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='943 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='944 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='945 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='946 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='947 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='948 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='949 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='950 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Objective value ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='(c) Cifar-10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Iteration step ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='656 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='658 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='660 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='662 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='664 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='666 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='668 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='670 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Objective value ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='(d) SUNRGBD ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5: Convergence results for the 100leaves, Caltech-101, Cifar-10 and SUNRGBD datasets, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' analysis of the proposed GCAE in this section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5 shows the convergence curves of GCAE on the datasets 100leaves, Caltech-101 and Cifar-10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In Fig .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='5, the x-axis and y-axis present the number of iterations and the value of the objective function, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It can obviously find that the values of the objective function in GCAE decrease monotonically with the increased iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Furthermore, the objective values converge very quickly after 5-10 iterations which verifies the effectiveness of our proposed optimized solution for each sub- process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' CONCLUSION In this paper, we proposed a novel binary multi-view clustering method termed as Graph-Collaborated Auto-encoder Hashing for Multi-view Binary Clustering (GCAE), which can effectively construct low-rank affinity graphs from each view and jointly learn binary code by auto-encoders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' GCAE constructs graphs by utilizing auxiliary matrices to control the low-rank constraint, which can reasonably preserve essential information from original data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' And then GCAE adopts auto- encoders to collaborate multiple graphs for learning unified binary codes and obtaining cluster results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' With the pro- posed optimization algorithm, the objective function converges quickly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' The extensive experiments demonstrated the supe- riority of GCAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' It can be obviously seen that different from real-value multi-view clustering methods, GCAE can effectively obtain cluster results in Hamming space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Compared with hash methods, GCAE can reasonably utilize multi-view consistency and complementarity information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' We conducted our experiments on five multi-view datasets for experimental examination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' In the future work, we intend to further improve the speed of the algorithm, which is the deficiency of the proposed method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' ACKNOWLEDGMENT This work was supported in part by the National Nat- ural Science Foundation of China Grant 62002041 and 62176037, Liaoning Fundamental Research Funds for Uni- versities Grant LJKQZ2021010, Liaoning Doctoral Research Startup Fund Project Grant 2021-BS-075, Liaoning Province Applied Basic Research Project 22JH2/101300264 and Dalian Science and Technology Innovation Fund 2022JJ12GX019, 2021JJ12GX028 and 2022JJ12GX016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' REFERENCES [1] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Geng, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Han, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fu, and Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hu, “Autoencoder in autoencoder networks,” IEEE transactions on neural networks and learning systems, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [2] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yan, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, “Dual-constrained deep semi-supervised coupled factorization network with enriched prior,” International Journal of Computer Vision, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 129, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 12, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3233–3254, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [3] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lin, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gao, “Multiview spectral clustering via structured low-rank matrix factorization,” IEEE transactions on neural networks and learning systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 29, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 10, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 4833–4843, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [4] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Qian, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hong, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, “Switchable online knowledge distillation,” in European Conference on Computer Vision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Springer, 2022, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 449–466.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [5] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, “Survey on deep multi-modal data analytics: Collaboration, rivalry, and fusion,” ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 17, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1s, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1–25, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [6] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Peng, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Feng, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhou, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lei, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yan, “Deep subspace clustering,” IEEE transactions on neural networks and learning systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 31, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 12, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5509–5521, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [7] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lin, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wu, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Huang, “Robust subspace clustering for multi-view data by exploiting correlation con- sensus,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 24, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 11, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3939–3949, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [8] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Guo, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, “A completed modeling of local binary pattern operator for texture classification,” IEEE transactions on image processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 19, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 6, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1657–1663, 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [9] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Movellan, “Tutorial on gabor filters,” Open Source Document, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 40, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1–23, 2002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [10] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Dalal and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Triggs, “Histograms of oriented gradients for human detection,” in 2005 IEEE computer society conference on computer vision and pattern recognition (CVPR’05), vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Ieee, 2005, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 886–893.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [11] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lowe, “Distinctive image features from scale-invariant keypoints,” International journal of computer vision, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 60, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 91–110, 2004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [12] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yao, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yang, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, “Clustering- driven unsupervised deep hashing for image retrieval,” Neurocomputing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 368, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 114–123, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [13] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Peng, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, “Progressive learning with multi-scale attention network for cross-domain vehicle re-identification,” Science China Information Sciences, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 65, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 6, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1–15, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [14] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fan, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xing, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Sun, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cai, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Maybank, “Deep constrained siamese hash coding network and load-balanced locality- sensitive hashing for near duplicate image detection,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 27, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 9, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 4452–4464, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [15] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jiang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Peng, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Deng, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fu, “Towards adaptive consensus graph: Multi-view clustering via graph collaboration,” IEEE Transactions on Multimedia, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [16] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jiang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Peng, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Mi, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fu, “Tensorial multi-view clustering via low-rank constrained high-order graph learning,” IEEE Transactions on Circuits and Systems for Video Technology, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [17] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fu, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhuo, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xu, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, “Kernelized multiview subspace analysis by self-weighted learning,” IEEE Transactions on Multimedia, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 23, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3828–3840, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [18] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wu, “Beyond low-rank representations: Orthogonal clustering basis reconstruction with optimized graph structure for multi- view spectral clustering,” Neural Networks, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 103, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1–8, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [19] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tao Shen, “Supervised discrete hashing,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 37–45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [20] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tian, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, “Strongly constrained discrete hashing,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 29, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3596–3611, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 12 [21] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='-m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cheung, “Fddh: Fast discriminative discrete hashing for large-scale cross-modal retrieval,” IEEE Transactions on Neural Networks and Learning Systems, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [22] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Indyk and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Motwani, “Approximate nearest neighbors: towards removing the curse of dimensionality,” in Proceedings of the thirtieth annual ACM symposium on Theory of computing, 1998, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 604–613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [23] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Weiss, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Torralba, and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fergus, “Spectral hashing,” Advances in neural information processing systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 21, 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [24] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Mu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Kumar, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chang, “Discrete graph hashing,” Advances in neural information processing systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 27, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [25] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jiang and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, “Scalable graph hashing with feature transfor- mation,” in Twenty-Fourth International Joint Conference on Artificial Intelligence, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [26] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Kumar, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Rai, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Daume, “Co-regularized multi-view spectral clustering,” Advances in neural information processing systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 24, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [27] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhan, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Guan, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, “Graph learning for multiview clustering,” IEEE Transactions on Cybernetics, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 48, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 10, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2887–2895, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [28] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fang, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Pan, “Iterative views agreement: an iterative low-rank based structured optimization method to multi-view spectral clustering,” in Proceedings of the Twenty- Fifth International Joint Conference on Artificial Intelligence, 2016, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2153–2159.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [29] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yang, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, “Gmc: Graph-based multi-view cluster- ing,” IEEE Transactions on Knowledge and Data Engineering, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 32, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 6, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1116–1129, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [30] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xiao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chen, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shi, “Knowledge graph embedding based on multi-view clustering framework,” IEEE Transactions on Knowledge and Data Engineering, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 33, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 585–596, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [31] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shi, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, “Multi-view clustering via nonnega- tive and orthogonal graph reconstruction,” IEEE Transactions on Neural Networks and Learning Systems, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1–14, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [32] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shao, “Binary multi- view clustering,” IEEE transactions on pattern analysis and machine intelligence, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 41, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 7, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1774–1782, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [33] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jin, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chaki, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cohen, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gurfinkel, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Havrilla, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hines, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Narasimhan, “Binary function clustering using semantic hashes,” in 2012 11th International Conference on Machine Learning and Applica- tions, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1, 2012, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 386–391.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [34] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tian, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chen, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, “Unsupervised hashing based on the recovery of subspace structures,” Pattern Recognition, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 103, p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 107261, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [35] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lin, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yu, “Robust subspace segmentation by low-rank representation,” in Proceedings of the 27th International Conference on International Conference on Machine Learning, 2010, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 663–670.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [36] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gao, “Robust and flexible discrete hashing for cross-modal similarity search,” IEEE Transactions on Circuits and Systems for Video Technology, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 28, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 10, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2703–2715, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [37] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yao, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, “Set and rebase: de- termining the semantic graph connectivity for unsupervised cross-modal hashing,” in Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence, 2021, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 853–859.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [38] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zareapoor, and Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zheng, “Cluster-wise unsuper- vised hashing for cross-modal similarity search,” Pattern Recognition, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 111, p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 107732, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [39] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Qin, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhu, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xu, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shao, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, “Highly-economized multi-view binary compression for scalable image clustering,” in Proceedings of the European Conference on Computer Vision (ECCV), 2018, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 717–732.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [40] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', “Parameter-free auto-weighted multiple graph learning: a framework for multiview clustering and semi-supervised classification.” in IJCAI, 2016, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1881–1887.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [41] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hou, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tao, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yi, “Multi-view unsupervised feature selection with adaptive similarity and view weight,” IEEE Transactions on Knowledge and Data Engineering, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 29, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 9, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1998–2011, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [42] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hartigan and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wong, “Algorithm as 136: A k-means clustering algorithm,” Journal of the royal statistical society.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' series c (applied statistics), vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 28, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 100–108, 1979.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [43] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhan, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Niu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chen, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yang, “Graph struc- ture fusion for multiview clustering,” IEEE Transactions on Knowledge and Data Engineering, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 31, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 10, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1984–1993, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [44] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shi, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, “Auto-weighted multi-view clustering via spectral embedding,” Neurocomputing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 399, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 369–379, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [45] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jin, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hu, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Qi, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tang, “Semantic neighbor graph hashing for multimodal retrieval,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 27, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1405–1417, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [46] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Guan, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Sun, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Qi, “Graph-based supervised discrete image hashing,” Journal of Visual Communication and Image Representation, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 58, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 675–687, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [47] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xiang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Qin, and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hao, “Discrete multi-graph hashing for large-scale visual search,” Neural Processing Letters, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 49, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1055–1069, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [48] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Ren, “Unsupervised cross-modal retrieval via multi-modal graph regularized smooth matrix factorization hashing,” Knowledge-Based Systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 171, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 69–80, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [49] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Fu, “Learning robust and discriminative subspace with low- rank constraints,” IEEE transactions on neural networks and learning systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 27, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 11, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2160–2173, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [50] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Hu, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, “Discrete spectral hashing for efficient similarity retrieval,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 28, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1080–1091, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [51] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhou, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Yang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Song, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shen, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Tao, “A fast op- timization method for general binary code learning,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 25, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 12, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5610–5621, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [52] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Mallah, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cope, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Orwell et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=', “Plant leaf classification using probabilistic integration of shape, texture and margin features,” Signal Processing, Pattern Recognition and Applications, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 5, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 45– 54, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [53] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhu, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhang, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gao, and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Zhu, “Fast parameter-free multi-view subspace clustering with consensus anchor guidance,” IEEE Transactions on Image Processing, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 31, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 556– 568, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [54] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jin, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lin, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cai, “Density sensitive hashing,” IEEE transactions on cybernetics, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 44, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 8, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 1362–1371, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [55] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xia, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' He, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Kohli, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Sun, “Sparse projections for high- dimensional binary codes,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 3332–3339.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [56] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gong, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Lazebnik, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gordo, and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Perronnin, “Iterative quantiza- tion: A procrustean approach to learning binary codes for large-scale image retrieval,” IEEE transactions on pattern analysis and machine intelligence, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 35, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 12, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 2916–2929, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [57] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Ng, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Jordan, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Weiss, “On spectral clustering: Analysis and an algorithm,” Advances in neural information processing systems, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 14, 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [58] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Liu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Gao, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Han, “Multi-view clustering via joint nonnegative matrix factorization,” in Proceedings of the 2013 SIAM international conference on data mining.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' SIAM, 2013, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' 252–260.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [59] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Cai, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, “Multi-view clustering and semi-supervised classification with adaptive neighbours,” in Thirty-first AAAI conference on artificial intelligence, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [60] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Kang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Guo, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Huang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Chen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Su, and Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Xu, “Mul- tiple partitions aligned clustering,” arXiv preprint arXiv:1909.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content='06008, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' [61] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Shi, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Nie, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Wang, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
+page_content=' Li, “Multi-view clustering via nonnega- tive and orthogonal graph reconstruction,” IEEE Transactions on Neural Networks and Learning Systems, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/T9E0T4oBgHgl3EQflQGf/content/2301.02484v1.pdf'}
diff --git a/U9FJT4oBgHgl3EQfNyw_/vector_store/index.faiss b/U9FJT4oBgHgl3EQfNyw_/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..0801ae90b1413fba281ed0b73f07b3f23dfe9839
--- /dev/null
+++ b/U9FJT4oBgHgl3EQfNyw_/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c5f60d60db66c0dc6c0ff2f6f766ccf1e3ec117df810aaf39727ff934607ed9
+size 17498157
diff --git a/UNE2T4oBgHgl3EQfXQdn/content/2301.03842v1.pdf b/UNE2T4oBgHgl3EQfXQdn/content/2301.03842v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..7399336a23f368a5498f4c53833526ef23eb836b
--- /dev/null
+++ b/UNE2T4oBgHgl3EQfXQdn/content/2301.03842v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef30774f03fa49171f9d3fd7fcafd02ccd0d9245da6d19c91e8583cece3a1de4
+size 358276
diff --git a/UNE2T4oBgHgl3EQfXQdn/vector_store/index.faiss b/UNE2T4oBgHgl3EQfXQdn/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..c644fccdfeffeec4f2a66c753954985867564664
--- /dev/null
+++ b/UNE2T4oBgHgl3EQfXQdn/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:892c1b8b7efbc1aad2c14882ff8d2ac4dea7063ae7e1dc2d43b278cb124c62b7
+size 1179693
diff --git a/VdFIT4oBgHgl3EQfgyug/vector_store/index.faiss b/VdFIT4oBgHgl3EQfgyug/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..4d04d0ffcb9f32df324b6a7f1b1164df7737cb4f
--- /dev/null
+++ b/VdFIT4oBgHgl3EQfgyug/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0498c94e3811ec2c2e909805a944dd2170c395f5f25507c40be1848398ba65ea
+size 6029357
diff --git a/W9E2T4oBgHgl3EQfDwYa/content/tmp_files/2301.03627v1.pdf.txt b/W9E2T4oBgHgl3EQfDwYa/content/tmp_files/2301.03627v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..adde7492287c65c9219ffba74f1e8b698389f024
--- /dev/null
+++ b/W9E2T4oBgHgl3EQfDwYa/content/tmp_files/2301.03627v1.pdf.txt
@@ -0,0 +1,2401 @@
+Quantifying the structural stability of simplicial homology
+Nicola Guglielmi∗, Anton Savostianov† , and Francesco Tudisco†
+Abstract. The homology groups of a simplicial complex reveal fundamental properties of the topology of the
+data or the system and the notion of topological stability naturally poses an important yet not
+fully investigated question.
+In the current work, we study the stability in terms of the smallest
+perturbation sufficient to change the dimensionality of the corresponding homology group. Such
+definition requires an appropriate weighting and normalizing procedure for the boundary operators
+acting on the Hodge algebra’s homology groups. Using the resulting boundary operators, we then
+formulate the question of structural stability as a spectral matrix nearness problem for the corre-
+sponding higher-order graph Laplacian. We develop a bilevel optimization procedure suitable for
+the formulated matrix nearness problem and illustrate the method’s performance on a variety of
+synthetic quasi-triangulation datasets and transportation networks.
+Key words. simplicial complexes, homology groups, graph Laplacian, Hodge Laplacian, matrix nearness prob-
+lems, matrix ODEs, spectral optimization, constrained gradient system
+MSC codes. 05C50, 65F45, 65K10, 57M15, 62R40
+1. Introduction. Models based on graphs and networks are ubiquitous in the sciences and
+engineering; for example, they have been successfully applied to model chemical reactions,
+traffic and electric flows, social interactions, and to describe abstract datasets in machine
+learning pipelines. Graph properties can be used to determine important nodes [18, 37, 13],
+reveal modular structure of a system [16, 36, 27], model collective network dynamics such as
+synchronization [31] and opinion formation [19]. However, models based on graphs are limited
+to descriptions based on pairwise node-node relationships.
+While graph-based models are widely used and successful, many complex systems and
+datasets are better described by higher-order relations that go beyond pairwise interactions
+[4, 7, 9]. Relational data is full of interactions that happen in groups. For example, friendship
+relations often involve groups that are larger than two individuals. In fact, monophily and
+triadic closure principles from the social sciences suggest that motifs, such as triangles, are
+important building blocks of relational data [1, 3, 26]. Also in the presence of point-cloud data,
+directly modeling higher-order data interactions has led to improvements in numerous data
+mining settings, including clustering [8, 17, 33], link prediction [3, 6], and ranking [5, 32, 35].
+Simplicial complexes are standard higher-order network models, where simplicies of dif-
+ferent order can connect a larger number of nodes. Higher-order Laplacians are key algebraic
+tools that naturally correspond to a simplicial complex. Formally, these are a sequence of lin-
+ear operators that generalizes the better-known graph Laplacian, obtained when only pairwise
+edge relations are considered. Very useful topological properties about the data are revealed
+by the kernels of these operators which, by the Fundamental Lemma of Homology, define a
+homology of the data and reveal fundamental properties such as connected components, holes,
+and voids.
+∗Gran
+Sasso
+Science
+Institute,
+L’Aquila,
+Italy
+(nicola.guglielmi@gssi.it,
+anton.savostianov@gssi.it,
+francesco.tudisco@gssi.it).
+1
+arXiv:2301.03627v1 [math.NA] 9 Jan 2023
+
+2
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+In this work, we are concerned with quantifying the stability of such homological proper-
+ties. More precisely, given an initial simplicial complex, we develop a numerical method based
+on a suitable matrix differential equation that allows us to compute the closest simplicial
+complex with different homology. The precise meaning of being “close” and being “different”
+will be based on the structure of the specific higher-order Laplacian matrix. While a great
+effort has been devoted in recent years to measure the presence and persistence of simplicial
+homology [12, 28], very little work is available about the stability of the homology classes with
+respect to data perturbation.
+In order to make a sound mathematical formulation of the problem we aim to solve,
+and of the numerical model we design for its solution, in Section 2 we review in detail the
+notion of homology simplicial complexes and the corresponding higher-order Laplacians. In
+Subsection 2.2 and Subsection 3.1 we discuss how these operators may be extended to account
+for weighted higher-order node relations and formulate the corresponding stability problem in
+Section 3 and Section 4. Then, in Section 5 and Section 6 we present our numerical method
+based on a two-level constrained matrix gradient flow approach. Finally, we devote Section 7
+to illustrate the performance of the proposed numerical scheme on several example datasets.
+2. Simplicial complexes and higher-order relations. A graph G is a pair of sets (V, E),
+where V = {1, . . . , n} is the set of vertices and E ⊂ V×V is a set of unordered pairs representing
+the undirected edges of G. We let m denote the number of edges E = {e1, . . . , em} and we
+assume them ordered lexicographically, with the convention that i < j for any {i, j} ∈ E. For
+brevity, we often write ij in place of {i, j} to denote the edge joining i and j. Moreover, we
+assume no self-loops, i.e., ii /∈ E for all i ∈ V.
+A graph only considers pairwise relations between the vertices. A simplicial complex K is
+a generalization of a graph that allows us to model connections involving an arbitrary number
+of nodes by means of higher-order simplicies. Formally, a k-th order simplex (or k-simplex,
+briefly) is a set of k + 1 vertices {i0, i1, . . . , ik} with the property that every subset of k nodes
+itself is a (k − 1)-simplex. Any (k − 1)-simplex of a k-simplex is called a face. The collection
+of all such simplices forms a simplicial complex K, which therefore essentially consists of
+a collection of sets of vertices such that every subset of the set in the collection is in the
+collection itself. Thus, a graph G can be thought of as the collection of 0- and 1- simplices:
+the 0-simplices form the nodes set of G, while 1-simplices form its edges. To emphasize this
+analogy, in the sequel we often specify that K is a simplicial complex on the vertex set V.
+Just like the edges of a graph, to any simplicial complex, we can associate an orientation
+(or ordering). To underline that an ordering has been fixed, we denote an ordered k-simplex
+σ using square brackets σ = [i0 . . . ik]. In particular, as for the case of edges, in this work, we
+always assume the lexicographical ordering, unless specified otherwise. That is, we assume
+that:
+1. any k-simplex [i0 . . . ik] in K is such that i0 < · · · < ik;
+2. the k-simplices σ1, σ2, . . . of K are ordered so that σi ≺ σi+1 for all i, where [i0 . . . ik] ≺
+[i′
+0 . . . i′
+k] if and only if there exists h such that 0 ≤ h ≤ k, ij = i′
+j for j = 0, . . . , h and
+ih < i′
+h.
+As for the edges, we often write i0 . . . ik in place of [i0 . . . ik] in this case.
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+3
+2.1. Homology, boundary operators and higher-order Laplacians. Topological proper-
+ties of a simplicial complex can be studied by considering boundary operators, higher-order
+Laplacians, and the associated homology. Here we recall these concepts trying to emphasize
+their matrix-theoretic flavor. To this end, we first fix some further notation and recall the
+notion of a real k-chain.
+Definition 2.1. Assume K is a simplicial complex on the vertex set V. For k ≥ 0, we denote
+the set of all the oriented k-th order simplices in K as Vk(K) or simply Vk. Thus, V0 = V and
+V1 = E form the underlying graph of K, which we denote by GK = (V, E) = (V0, V1).
+Definition 2.2. The formal real vector space spanned by all the elements of Vk with real
+coefficients is denoted by Ck(K). Any element of Ck(K), the formal linear combinations of
+simplices in Vk, is called a k-chain.
+We remark that, in the graph-theoretic terminology, C0(K) is usually called the space of
+vertices’ states, while C1(K) is usually called the space of flows in the graph.
+The chain spaces are finite vector spaces generated by the set of k-simplicies. The bound-
+ary and co-boundary operators are particular linear mappings between Ck and Ck−1, which
+in a way are the discrete analogous of high-order differential operators (and their adjoints) on
+continuous manifolds. The boundary operator ∂k maps a k-simplex to an alternating sum of
+its (k − 1)-dimensional faces obtained by omitting one vertex at a time. Its precise definition
+is recalled below, while Figure 1 provides an illustrative example of its action.
+Definition 2.3. Let k ≥ 1. Given a simplicial complex K over the set V, the boundary op-
+erator ∂k : Ck(K) �→ Ck−1(K) maps every ordered k-simplex [i0 . . . ik] ∈ Ck(K) to the following
+alternated sum of its faces:
+∂k[i0 . . . ik] =
+k
+�
+j=0
+(−1)j[i0 . . . ij−1ij+1 . . . ik] ∈ Ck−1(K)
+As we assume the k-simplicies in Vk are ordered lexicographically, we can fix a canonical
+basis for Ck(K) and we can represent ∂k as a matrix Bk with respect to such basis. In fact,
+once the ordering is fixed, Ck(K) is isomorphic to RVk, the space of functions from Vk to R
+or, equivalently, the space of real vectors with |Vk| entries. Thus, Bk is a |Vk−1| × |Vk| matrix
+and ∂∗
+k coincides with B⊤
+k . We shall always assume the canonical basis for Ck(K) is fixed in
+this way and we will deal exclusively with the matrix representation Bk from now on. An
+example of Bk for k = 1 and k = 2 is shown in Figure 1.
+A direct computation shows that the following fundamental identity holds (see e.g. [24,
+Thm. 5.7])
+(2.1)
+BkBk+1 = 0
+for any k. This identity is also known in the continuous case as the Fundamental Lemma of
+Homology and it allows us to define a homology group associated with each k-chain. In fact,
+(2.1) implies in particular that im Bk+1 ⊂ ker Bk, so that the k-th homology group is correctly
+defined:
+Hk := ker Bk⧸im Bk+1
+
+4
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+1
+2
+3
+4
+5
+6
+7
+B1 =
+�
+�
+�
+�
+� 1
+2
+� � 1
+3
+� � 2
+3
+� � 2
+4
+� � 3
+5
+� � 4
+5
+� � 4
+6
+� � 4
+7
+� � 5
+6
+� � 6
+7
+�
+[1] −1
+−1
+0
+0
+0
+0
+0
+0
+0
+0
+[2]
+1
+0
+−1
+−1
+0
+0
+0
+0
+0
+0
+[3]
+0
+1
+1
+0
+−1
+0
+0
+0
+0
+0
+[4]
+0
+0
+0
+1
+0
+−1
+−1
+−1
+0
+0
+[5]
+0
+0
+0
+0
+1
+1
+0
+0
+−1
+0
+[6]
+0
+0
+0
+0
+0
+0
+1
+0
+1
+−1
+[7]
+0
+0
+0
+0
+0
+0
+0
+1
+0
+1
+�
+�
+�
+�
+B2 =
+�
+�
+�
+�
+�
+�
+�
+�
+�
+� 1
+2
+3
+� � 4
+5
+6
+� � 4
+6
+7
+�
+[1, 2]
+1
+0
+0
+[1, 3] −1
+0
+0
+[2, 3]
+1
+0
+0
+[2, 4]
+0
+0
+0
+[3, 5]
+0
+0
+0
+[4, 5]
+0
+1
+0
+[4, 6]
+0
+−1
+1
+[4, 7]
+0
+0
+−1
+[5, 6]
+0
+1
+0
+[6, 7]
+0
+0
+1
+�
+�
+�
+�
+�
+�
+�
+�
+�
+∂2([1, 2, 3]) = [1, 2] − [1, 3] + [2, 3]
+Figure 1. Left-hand side panel: example of simplicial complex K on 7 nodes, and of the action of ∂2 on the
+2-simplex [1, 2, 3]; 2-simplices included in the complex are shown in red, arrows correspond to the orientation.
+Panels on the right: matrix forms B1 and B2 of boundary operators ∂1 and ∂2 respectively.
+The dimensionality of the k-th homology group is known as k-th Betti’s number βk =
+dim Hk, while the elements of Hk correspond to so-called k-dimensional holes in the simplicial
+complex.
+For example, H0, H1 and H2 describe connected components, holes and three-
+dimensional voids respectively. By standard algebraic passages one sees that Hk is isomorphic
+to ker
+�
+B⊤
+k Bk + Bk+1B⊤
+k+1
+�
+.
+Thus, the number of k-dimensional holes corresponds to the
+dimension of the kernel of a linear operator, which is known as k-th order Laplacian or
+higher-order Laplacian of the simplicial complex K.
+Definition 2.4. Given a simplicial complex K and the boundary operators Bk and Bk+1, the
+k-th order Laplacian Lk of K is the |Vk| × |Vk| matrix defined as:
+(2.2)
+Lk = B⊤
+k Bk + Bk+1B⊤
+k+1
+In particular, we remark that:
+• the 0-order Laplacian is the standard combinatorial graph Laplacian L0 = B1B⊤
+1 ∈ Rn×n,
+whose diagonal entries consist of the degrees of the corresponding vertices (i.e. the number
+of 1-simplices each vertex belongs to), while the off-diagonal (L0)ij is equal to −1 if either
+ij is a 1-simplex, and it is zero otherwise;
+• the 1-order Laplacian is known as Hodge Laplacian L1 = B⊤
+1 B1+B2B⊤
+2 ∈ Rm×m. Similarly
+to the 0-order case L0, one can describe the entries of L1 in terms of the structure of the
+simplicial complex, see e.g. [25].
+2.1.1. Connected components and holes. The boundary operators Bk on K are directly
+connected with discrete notions of differential operators on the graph.
+In particular, B1,
+B⊤
+1 , and B⊤
+2 are the graph’s divergence, gradient, and curl operators, respectively. We refer
+to [24] for more details. As Hk is isomorphic to ker Lk, we have that the following Hodge
+decomposition of RVk holds
+RVk = im Bk+1 ⊕ im B⊤
+k ⊕ ker Lk = im Bk+1 ⊕ im B
+⊤
+k ⊕ ker Lk .
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+5
+Figure 2. Continuous and analogous discrete manifolds with one 1-dimensional hole (dim H1 = 1). Left
+pane: the continuous manifold; center pane: the discretization with mesh vertices; right pane: a simplicial
+complex built upon the mesh. Triangles in the simplicial complex K are colored gray (right).
+Thus, the space of vertex states RV0 can be decomposed as RV0 = im B1 ⊕ ker L0, the sum of
+divergence-free vectors and harmonic vectors, which correspond to the connected components
+of the graph. In particular, for a connected graph, ker L0 is one-dimensional and consists of
+entry-wise constant vectors. Thus, for a connected graph, im B1 is the set of vectors whose
+entries sum up to zero.
+Similarly, the space of flows on graph’s edges RV1 can be decomposed as RV1 = im B2 ⊕
+im B⊤
+1 ⊕ ker L1.
+Thus, each flow can be decomposed into its gradient part im B⊤
+1 , which
+consists of flows with zero cycle sum, its curl part im B2, which consists of circulations around
+order-2 simplicies in K, and its harmonic part ker L1, which represents 1-dimensional holes
+defined as global circulations modulo the curl flows.
+While 0-dimensional holes are easily understood as the connected components of the graph
+GK, a notion of “holes in the graph” GK corresponds to 1-dimensional holes in K.
+This
+terminology comes from the analogy with the continuous case. In fact, if the graph is obtained
+as a discretization of a continuous manifold, harmonic functions in the homology group H1
+would correspond to the holes in the manifold, as illustrated in Figure 2.
+Moreover, the
+Hodge Laplacian of a simplicial complex built on N randomly sampled points in the manifold
+converges in the thermodynamic limit to its continuous counterpart, as N → ∞, [11].
+2.2. Normalized and weighted higher-order Laplacians. In the classical graph setting,
+a normalized and weighted version of the Laplacian matrix is very often employed in ap-
+plications. From a matrix theoretic point of view, having a weighted graph corresponds to
+allowing arbitrary nonnegative entries in the adjacency matrix defining the graph. In terms
+of boundary operators, this coincides with a positive diagonal rescaling. Analogously, the
+normalized Laplacian is defined by applying a diagonal congruence transformation to the
+standard Laplacian using the node weights. We briefly review these two constructions below.
+Let G = (V, E) be a graph with positive node and edge weight functions w0 : V → R++
+and w1 : E → R++, respectively. Define the |V| × |V| and |E| × |E| diagonal matrices W0 =
+Diag{w0(vi)1/2}n
+i=1 and W1 = Diag{w1(ei)1/2}m
+i=1. Then B1 = W −1
+0 B1W1 is the normalized
+and weighted boundary operator of G and we have that
+(2.3)
+L0 = B1B
+⊤
+1 = W −1
+0 B1W 2
+1 B⊤
+1 W −1
+0
+is the normalized weighted graph Laplacian of G.
+In particular, note that, as for L0, the entries of L0 uniquely characterize the graph
+
+6
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+topology, in fact we have
+(L0)ii = deg(i)
+w0(i) ,
+(L0)ij =
+�
+�
+�
+−
+w1(ij)
+√
+w0(i)w0(j)
+ij ∈ E
+0
+otherwise
+, for i ̸= j
+where deg(i) denotes the (weighted) degree of node i, i.e. deg(i) = �
+e∈E:i∈e w1(e).
+While the definition of k-th order Laplacian is well-established for the case of unweighted
+edges and simplices, a notion of weighted and normalized k-th order Laplacian is not univer-
+sally available and it might depend on the application one has at hand. For example, different
+definitions of weighted Hodge Laplacian are considered in [10, 22, 24, 29].
+At the same time, we notice that the notation used in (2.3) directly generalizes to higher
+orders. Thus, we propose the following notion of normalized and weighted k-th Laplacian
+Definition 2.5. Let K be a simplicial complex and let wk : Vk → R++ be a positive-valued
+weight function on the k-simplicies of K. Define the diagonal matrix Wk = Diag
+�
+wk(σi)1/2�|Vk|
+i=1.
+Then, Bk = W −1
+k−1BkWk is the normalized and weighted k-th boundary operator, to which cor-
+responds the normalized and weighted k-th Laplacian
+Lk = B
+⊤
+k Bk + Bk+1B
+⊤
+k+1
+(2.4)
+= WkB⊤
+k W −2
+k−1BkWk + W −1
+k Bk+1W 2
+k+1B⊤
+k+1W −1
+k .
+Note that, from the definition Bk = W −1
+k−1BkWk and (2.1), we immediately have that
+BkBk+1 = 0. Thus, the group Hk = ker Bk/ im Bk+1 is well defined for any choice of positive
+weights wk and is isomorphic to ker Lk. While the homology group may depend on the weights,
+we observe below that its dimension does not. Precisely, we have
+Proposition 2.6. The dimension of the homology groups of K is not affected by the weights
+of its k-simplicies. Precisely, if Wk are positive diagonal matrices, we have
+(2.5)
+dim ker Bk = dim ker Bk,
+dim ker B
+⊤
+k = dim ker B⊤
+k ,
+dim Hk = dim Hk .
+Moreover, ker Bk = Wk ker Bk and ker B⊤
+k = W −1
+k−1 ker B
+⊤
+k .
+Proof. Since Wk is an invertible diagonal matrix,
+Bkx = 0 ⇐⇒ W −1
+k−1BkWkx = 0 ⇐⇒ BkWkx = 0.
+Hence, if x ∈ ker Bk, then Wkx ∈ ker Bk, and, since Wk is bijective, dim ker Bk = dim ker Bk.
+Similarly, one observes that dim ker B
+⊤
+k = dim ker B⊤
+k .
+Moreover, since BkBk+1 = 0, then im Bk+1 ⊆ ker Bk and im B
+⊤
+k ⊆ ker B
+⊤
+k+1. This yields
+ker Bk ∪ ker B
+⊤
+k+1 = RVk = ker Bk ∪ ker B⊤
+k+1. Thus, for the homology group it holds:
+dim Hk = dim
+�
+ker Bk ∩ ker B
+⊤
+k+1
+�
+=
+= dim ker Bk + dim ker B
+⊤
+k+1 − dim
+�
+ker Bk ∪ ker B
+⊤
+k+1
+�
+=
+= dim ker Bk + dim ker B⊤
+k+1 − dim
+�
+ker Bk ∪ ker B⊤
+k+1
+�
+= dim Hk
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+7
+2.3. Principal spectral inheritance. Before moving on to the next section, we recall here
+a relatively direct but important spectral property that connects the spectra of the k-th and
+(k + 1)-th order Laplacians.
+Theorem 2.7 (HOL’s spectral inheritance).
+Let Lk and Lk+1 be higher-order Laplacians
+for the same simplicial complex K. Let Lk = L
+down
+k
++ L
+up
+k , where L
+down
+k
+= B
+⊤
+k Bk and L
+up
+k =
+Bk+1B
+⊤
+k+1. Then:
+1. σ+(L
+up
+k ) = σ+(L
+down
+k+1 ), where σ+(·) denotes the positive part of the spectrum;
+2. if 0 ̸= µ ∈ σ+(L
+up
+k ) = σ+(L
+down
+k+1 ), then the eigenvectors are related as follows:
+(a) if x is and eigenvector for L
+up
+k
+with the eigenvalue µ, then y =
+1
+õB
+⊤
+k+1x is an
+eigenvector for L
+down
+k+1
+with the same eigenvalue;
+(b) if u is and eigenvector for L
+down
+k+1
+with the eigenvalue µ and u /∈ ker Bk+1, then v =
+1
+õBk+1u is an eigenvector for L
+up
+k
+with the same eigenvalue;
+3. for each Laplacian Lk: if v /∈ ker L
+down
+k
+is the eigenvector for L
+down
+k
+, then v ∈ ker L
+up
+k ; vice
+versa, if u /∈ ker L
+up
+k
+is the eigenvector for L
+up
+k , then v ∈ ker L
+down
+k
+;
+4. consequently, there exist µ ∈ σ+(Lk) with an eigenvector u ∈ ker L
+up
+k , and ν ∈ σ+(Lk+1)
+with an eigenvector u ∈ ker L
+down
+k+1 , such that:
+B
+⊤
+k Bkv = µv,
+Bk+2B
+⊤
+k+2u = νu .
+Proof. For (2a) it is sufficient to note that L
+down
+k+1 y = B
+⊤
+k+1Bk+1 1
+õB
+⊤
+k+1x =
+1
+õB
+⊤
+k+1L
+up
+k x =
+õB
+⊤
+k+1x = µy. Similarly, for (2b): L
+up
+k v = Bk+1B
+⊤
+k+1
+1
+õBk+1u =
+1
+õBk+1L
+down
+k+1 u = µv;
+joint 2(a) and 2(b) yield (1). Hodge decomposition immediately yields the strict separation of
+eigenvectors between L
+up
+k
+and L
+down
+k
+, (3); given (3), all the inherited eigenvectors from (2a)
+fall into the ker L
+down
+k+1 , thus resulting into (4).
+In other words, the variation of the spectrum of the k-th Laplacian when moving from one
+order to the next one works as follows: the down-term L
+down
+k+1 inherits the positive part of the
+spectrum from the up-term of L
+up
+k ; the eigenvectors corresponding to the inherited positive
+part of the spectrum lie in the kernel of L
+up
+k+1; at the same time, the “new” up-term L
+up
+k+1 has
+a new, non-inherited, part of the positive spectrum (which, in turn, lies in the kernel of the
+(k + 2)-th down-term).
+In particular, we notice that for k = 0, since B0 = 0 and L0 = L
+up
+0 , the theorem yields
+σ+(L0) = σ+(L1
+down) ⊆ σ+(L1). In other terms, the positive spectrum of the L0 is inher-
+ited by the spectrum of L1 and the remaining (non-inherited) part of σ+(L1) coincides with
+σ+(L
+up
+1 ). Figure 3 provides an illustration of the statement of Theorem 2.7 for k = 0.
+3. Problem setting: Nearest complex with different homology. Suppose we are given
+a simplicial complex K on the vertex set V, with simplex weight functions w0, w1, . . . , and
+let βk = dim Hk = dim Hk the dimension of its k-homology. We aim at finding the closest
+simplex on the same vertex set V, with a strictly larger number of holes. As it is the most
+frequent higher-order Laplacian appearing in applications and since this provides already a
+
+8
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+0 0 · · · 0
+λ1
+λ2
+λ3
+λ4
+λ5
+λ6
+λ7
+λ8
+λ9 λ10
+←
+σ(L1)
+0 0 · · · 0
+λ1
+λ2 0
+λ4 0 0
+λ7
+λ8 0
+λ10
+←
+σ(B
+T
+1 B1)
+0 0 · · · 0
+0 0
+λ3 0
+λ5
+λ6 0 0
+λ9 0
+←
+σ(B2B
+T
+2 )
+holes
+µ
+Figure 3.
+Illustration for the principal spectrum inheritance (Theorem 2.7) in case k = 0: spectra of
+L1, L
+down
+1
+and L
+up
+1
+are shown.
+Colors signify the splitting of the spectrum, λi > 0 ∈ σ(L1) ; all yellow
+eigenvalues are inherited from σ+(L0); red eigenvalues belong to the non-inherited part.
+Dashed barrier µ
+signifies the penalization threshold (see the target functional in Section 4) preventing homological pollution (see
+Subsection 3.1).
+large number of numerical complications, we focus here only on the Hodge Laplacian case:
+given the simplicial complex K = (V0, V1, V2, . . . ), we look for the smallest perturbation of the
+edges V1 which increases the dimension of H1. More precisely:
+Problem 3.1. Let K be a simplicial complex of order at least 2 with edge weights w1 and
+corresponding diagonal weight matrix W1, and let β1(W1) be the dimension of the homology
+group corresponding to the weights W1. For ε > 0, let
+Ω(ε) =
+�
+diagonal matrices W such that ∥W∥ = ε
+�
+,
+Π(W1) =
+�
+diagonal matrices W such that W1 + W ≥ 0
+�
+.
+In other words, Ω(ε) is an ε-sphere and Π(W1) allows only non-negative simplex weights.
+We look for the smallest perturbation ε such that there exists a weight modification δW1 ∈
+Ω(ε) ∩ Π(W1) such that β1(W1) < β1(W1 + δW1).
+Here, and throughout the paper, ∥X∥ denotes either the Frobenius norm if X is a matrix,
+or the Euclidean norm if X is a vector. Note that, as we are looking for the smallest ε, the
+equality ∥W∥ = ε is an obvious choice, as opposed to ∥W∥ ≤ ε.
+As the dimension β1 coincides with the kernel of L1, we approach this problem through
+the minimization of a functional based on the spectrum of the 0-th and 1-st Laplacian of the
+simplicial complex. In order to define such functional, we first make a number of considera-
+tions.
+Note that, due to Proposition 2.6, the dimension of the first homology group does not
+change when the edge weights are perturbed, as long as all the weights remain positive. Thus,
+in order to find the desired perturbation δW1, we need to set some of the initial weights to
+zero. This operation creates several potential issues we need to carefully address, as discussed
+next.
+First, setting an edge to zero implies that one is formally removing that edge from the
+simplicial complex. As the simplicial complex structure needs to be maintained, when doing
+so we need to set to zero also the weight of any 2-simplex that contains that edge. For this
+reason, if �w1(e) = w1(e) + δw1(e) is the new edge weight function, we require the weight
+function of the 2-simplices to change into �w2, defined as
+�w2(i1i2i3) = f
+�δw1(i1i2)
+w1(i1i2) , δw1(i2i3)
+w1(i2i3) , δw1(i1i3)
+w1(i1i3)
+�
+· w2(i1i2i3)
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+9
+where f(u1, u2, u3) is a function such that f(0, 0, 0) = 1 and that monotonically decreases to
+zero as ui → −1, for any i = 1, 2, 3. An example of such f is
+f(u1, u2, u3) = 1 − min{u1, u2, u3} .
+(3.1)
+Second, when setting the weight of an edge to zero we may disconnect the underlying
+graph and create an all-zero column and row in the Hodge Laplacian. This gives rise to the
+phenomenon that we call “homological pollution”, which we will discuss in detail in the next
+subsection.
+3.1. Homological pollution: inherited almost disconnectedness. As the Hodge homol-
+ogy β1 corresponds to the number of zero eigenvalues in ker L1, the intuition suggests that if
+L1 has some eigenvalue that is close to zero, then the simplicial complex is “close to” having
+at least one more 1-dimensional hole. There are a number of problems with this intuitive
+consideration.
+By Theorem 2.7 for k = 0, σ+(L1) inherits σ+(L0). Hence, if the weights in W1 vary
+continuously so that a positive eigenvalue in σ+(L0) approaches 0, the same happens to
+σ+(L1). Assuming the initial graph GK is connected, an eigenvalue that approaches zero in
+σ(L0) would imply that GK is approaching disconnectedness. This leads to a sort of pollution of
+the kernel of L1 as an almost-zero eigenvalue which corresponds to an “almost” 0-dimensional
+hole (disconnected component) from L0 is inherited into the spectrum of L1, but this small
+eigenvalue of L1 does not correspond to the creation of a new 1-dimensional hole in the reduced
+complex.
+To better explain the problem of homological pollution, let us consider the following
+illustrative example.
+Example 3.2. Consider the simplicial complex of order 2 depicted in Figure 4a. In this ex-
+ample we have V0 = {1, . . . , 7}, V1 = {[1, 2], [1, 3], [2, 3], [2, 4], [3, 5], [4, 5], [4, 6], [5, 6], [5, 7], [6, 7]}
+and V2 = {[1, 2, 3], [4, 5, 6], [5, 6, 7]}, all with weight equal to one: wk ≡ 1 for k = 0, 1, 2. The
+only existing 1-dimensional hole is shown in red and thus the corresponding Hodge homology
+is β = 1. Now, consider perturbing the weight of edges [2, 4] and [3, 5] by setting their weights
+to ε > 0 Figure 4b. For small ε, this perturbation implies that the smallest nonzero eigenvalue
+µ2 in σ+(L0) is scaled by ε. As σ+(L0) ⊆ σ+(L1), we have that dim ker L1 = 1 and σ+(L1)
+has an arbitrary small eigenvalue, approaching 0 with ε → 0. At the same time, when ε → 0,
+the reduced complex obtained by removing the zero edges as in Figure 4c does not have any
+1-dimensional hole, i.e. β1 = 0. Thus, in this case, the presence of a very small eigenvalue
+µ2 ∈ σ+(L1) does not imply that the simplicial complex is close to a simplicial complex with a
+larger Hodge homology.
+To mitigate the phenomenon of homological pollution, in our spectral-based functional for
+Problem 3.1 we include a term that penalizes the spectrum of L0 from approaching zero. To
+this end, we observe below that a careful choice of the vertex weights is required.
+The smallest non-zero eigenvalue of the Laplacian µ2 ∈ σ(L0) is directly related to the
+connectedness of the graph GK. This relation is well-known and dates back to the pioneering
+work of Fiedler [14]. In particular, as µ2 is a function of node and edge weights, the following
+
+10
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+1
+2
+3
+4
+5
+6
+7
+(a) Connected
+1
+1
+1
+ε
+ε
+1
+1
+1
+1
+1
+1
+2
+3
+4
+5
+6
+7
+(b) Close to discon-
+nected
+1
+1
+1
+1
+1
+1
+1
+1
+1
+2
+3
+4
+5
+6
+7
+(c) Disconnected
+Figure 4. Example of the homological pollution, Example 3.2, for the simplicial complex K on 7 vertices;
+the existing hole is shown in red (left and center pane), all 3 cliques are included in the simplicial complex. The
+left pane demonstrates the initial setup with 1 hole; the center pane retains the hole exhibiting spectral pollution;
+the continuous transition to the eliminated edges with β1 = 0 (no holes) is shown on the right pane.
+generalized version of the Cheeger inequality holds (see e.g. [34])
+(3.2)
+1
+2µ2 ≤ h(GK) ≤
+�
+2 µ2 max
+i∈V0
+deg(i)
+w0(i)
+�1/2
+,
+where
+h(GK) = min
+S⊂V0
+w1(S, V0\S)
+min{w0(S), w0(V0\S)} ,
+with
+w1(S, V0\S) =
+�
+ij∈V1:i∈S,j /∈S
+w1(ij),
+deg(i) =
+�
+j:ij∈V1
+wi(ij),
+w0(S) =
+�
+i∈S
+w0(i).
+We immediately see from (3.2) that when the graph GK is disconnected, then h(GK) = 0
+and µ2 = 0 as well. Vice-versa, if µ2 goes to zero, then h(GK) decreases to zero too. While
+this happens independently of the choice of w0 and w1, if w0 is a function of w1 then it might
+fail to capture the presence of edges whose weight is decreasing and is about to disconnect
+the graph.
+To see this, consider the example choice w0(i) = deg(i), the degree of node i in GK. Note
+that this is a very common choice in the graph literature, with several useful properties,
+including the fact that no other graph-dependent constant appears in the Cheeger inequality
+(3.2) other than µ2. For this weight choice, consider the case of a leaf node, a node i ∈ V0
+that has only one edge ij0 ∈ V1 connecting i to the rest of the (connected) graph GK via the
+node j0. If we set w1(ij0) = ε and we let ε decrease to zero, the graph GK is approaching
+disconnectedness and we would expect h(GK) and µ2 to decrease as well. However, one easily
+verifies that both µ2 and h(GK) are constant with respect to ε in this case, as long as ε ̸= 0.
+In order to avoid such a discontinuity, in our weight perturbation strategy for the simplex
+K, if w0 is a function of w1, we perturb it by a constant shift. Precisely, if w0 is the initial
+vertex weight of K, we set �w0(i) = w0(i) + ϱ, with ϱ > 0. So, for example, if w0 = deg and
+the new edge weight function �w1(e) = w1(e) + δw1(e) is formed after the addition of δW1, we
+set �w0(i) = �
+j [w1(ij) + δw1(ij)] + ϱ.
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+11
+4. Spectral functional for 1-st homological stability. We are now in the position to
+formulate our proposed spectral-based functional, whose minimization leads to the desired
+small perturbation that changes the first homology of K. In the notation of Problem 3.1,
+we are interested in the smallest perturbation ε and the corresponding modification δW1 ∈
+Ω(ε) ∩ Π(W1) that increases β1.
+As ∥δW1∥ = ε, for convenience we indicate δW1 = εE with ∥E∥ = 1 so E ∈ Ω(1)∩Πε(W1),
+where Πε(W1) = {W | εW ∈ Π(W1)}. For the sake of simplicity, we will omit the dependencies
+and write Ω and Πε, when there is no danger of ambiguity. Finally, let us denote by β1(ε, E) the
+first Betti number corresponding to the simplicial complex perturbed via the edge modification
+εE. With this notation, we can reformulate Problem 3.1 as follows:
+Problem 4.1. Find the smallest ε > 0, such that there exists an admissible perturbation
+E ∈ Ω ∩ Πε with an increased number of holes, i.e.
+min
+�
+ε > 0 : β1(ε, E) ≥ β1 + 1 for some E ∈ Ω ∩ Πε
+�
+(4.1)
+where β1 = β1(0, ·) is the first Betti number of the original simplicial complex.
+In order to approach Problem 4.1, we introduce a target functional F(ε, E), based on the
+spectrum of the 1-Laplacian L1(ε, E) and the 0-Laplacian L0(ε, E), where the dependence on
+ε and E is to emphasize the corresponding weight perturbation is of the form W1 �→ W1 +εE.
+Our goal is to move a positive entry in σ+(L1(ε, E)) into the kernel. At the same time,
+assuming the initial graph GK is connected, one should avoid the inherited almost discon-
+nectedness with small positive entries of σ+(L0(ε, E)).
+As, by Theorem 2.7 for k = 0,
+σ+(L0(ε, E)) = σ+(L
+down
+1
+(ε, E)), the only eigenvalue of L1(ε, E) that can be continuously
+driven to 0 comes from L
+up
+1 (ε, E). For this reason, let us denote the first non-zero eigenvalue
+of the up-Laplacian L
+up
+1 (ε, E) by λ+(ε, E). The proposed target functional F(ε, E) is defined
+as:
+F(ε, E) = λ+(ε, E)2
+2
++ α
+2 max
+�
+0, 1 − µ2(ε, E)
+µ
+�2
+(4.2)
+where α and µ are positive parameters, and µ2(ε, E) is the first nonzero eigenvalue of L0(ε, E).
+As recalled in Subsection 3.1, µ2(ε, E) is an algebraic measure of the connectedness of the
+perturbed graph, thus the whole second term in (4.2) “activates” when such algebraic con-
+nectedness falls below the threshold µ.
+By design, F(ε, E) is non-negative and is equal to 0 iff λ+(ε, E) reaches 0, increasing the
+dimension of H1. Using this functional, we recast the Problem 4.1 as
+min {ε > 0 : F(ε, E) = 0 for some E ∈ Ωε}
+(4.3)
+Remark 4.2. When GK is connected, dim ker L0 = 1 and by Theorem 2.7 dim ker L
+up
+1
+=
+dim ker L1+(n−dim ker L0) = n+β1−1, so the first nonzero eigenvalue of L
+up
+1 is the (n+β1)-
+th. While (n + β1) can be a large number in practice, we will discuss in Subsection 6.1 an
+efficient method that allows us to compute λ+(ε, E) without computing any of the previous
+(n + β1 − 1) eigenvalues.
+
+12
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+5. A two-level optimization procedure. We propose to approach (4.3) by means of a two-
+level iterative method, which is based on the successive minimization of the target functional
+F(ε, E) and a subsequent tuning of the parameter ε. More precisely, we propose the following
+two-level scheme. A similar procedure was used in the context of graph spectral nearness in
+[2] and in other matrix nearness problems [21].
+1. Inner level: for fixed ε > 0, solve the minimization problem
+E(ε) = arg min
+E∈Ω∩Πε
+F(ε, E)
+by a constrained gradient flow which we formulate below, where we denote the computed
+minimizer by E(ε).
+2. Outer level: given the function ε �→ E(ε), we consider the optimization problem:
+(5.1)
+F(ε, E(ε)) = 0
+and look for the smallest value ε∗ > 0 that solves (5.1).
+Details on the implementation of both levels are given below.
+By construction, the resulting algorithm converges to a minimum of F(ε, E). Although
+global convergence to the global optimum is not guaranteed, in our experiments we always
+observe the method reaches the expected global solution. However, we point out that there
+exists one pathological configuration of the simplicial complex K which would prevent global
+optimality. This happens when there are several holes and simplices from V2(K) being adjacent
+to each other only through a common edge, which has a small weight.
+In that case, the
+algorithm would always eliminate that common edge, whilst the correctness of this answer
+would depend on the balance between the number of holes and adjacent simplices. These
+configurations are rarely present in real-life systems (we did not encounter them in any of our
+tests); nevertheless, one can tackle such unsuccessful runs through manual weight modification
+of the undesired edge.
+5.1. Inner Level Iteration. Here we consider the minimization problem with respect to
+E and fixed scalar parameters: the perturbation norm ε is inherited from the outer level, and
+the connectedness parameters α and µ are assumed to be given (we will discuss the choice of
+µ and α later).
+We solve the resulting minimization problem min{F(ε, E) : E ∈ Ω ∩ Πε} by solving the
+associated constrained gradient system
+E
+•
+(t) = −PΩ∩ΠεG(ε, E(t))
+(5.2)
+where G(ε, E) = ∇EFk(ε, E) and PΩ∩Πε is a projector onto the admissible set Ω∩Πε (where ε
+is fixed). Since the system integrates the anti-gradient, a minimizer (at least local) of F(ε, E)
+is obtained at t → ∞. Equation (5.2) introduces a dummy time dependence outlining the
+difference between the discrete gradient descent and the gradient flow. In the current work, we
+use the latter which benefits from a variety of known integrators and simpler implementation
+of constraints.
+We devote the next two Subsection 5.2 and Subsection 5.3 to computing the projected
+gradient in (5.2). The idea is to express the derivative of F in terms of the derivative of the
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+13
+perturbation E
+•
+, and to identify the constrained gradient of the functional. To this end, we
+first compute the free gradient and then we discuss how to deal with the projection onto the
+admissible set. Then, in Subsection 5.2 we will discuss the free gradient transition phase that
+characterizes the outer iteration level for (5.1).
+5.2. The free gradient. We compute here the free gradient of F with respect to E, given
+a fixed ε. In order to proceed, we need a few preliminary results.
+The following is a standard perturbation result for eigenvalues; see e.g. [23], where we
+denote by ⟨X, Y ⟩ = �
+i,j xijyij = Tr(X⊤Y ) the inner product in Rn×n that induces the
+Frobenius norm ∥X∥ = ⟨X, X⟩1/2.
+Theorem 5.1 (Derivative of simple eigenvalues). Consider a continuously differentiable path
+of square symmetric matrices A(t) for t in an open interval I. Let λ(t), t ∈ I, be a continuous
+path of simple eigenvalues of A(t). Let x(t) be the eigenvector associated to the eigenvalue
+λ(t) and assume ∥x(t)∥ = 1 for all t. Then λ is continuously differentiable on I with the
+derivative (denoted by a dot)
+(5.3)
+λ
+•
+= x⊤A
+•
+x = ⟨xx⊤, A
+•
+⟩ .
+Moreover, “continuously differentiable” can be replaced with “analytic” in the assumption and
+the conclusion.
+Let us denote the perturbed weight matrix by �
+W1(t) = W1 +εE(t), and the corresponding
+�
+W0(t) = W0(�
+W1(t)) and �
+W2(t) = W2(�
+W1(t)), defined accordingly as discussed in Section 3;
+we omit the time dependence for the perturbed matrices to simplify the notation. Since �
+W0,
+�
+W1 and �
+W2 are necessarily diagonal, by the chain rule we have �
+W
+•
+i(t) = ε diag
+�
+Ji
+1E
+•
+1
+�
+, where
+1 is the vector of all ones, diag(v) is the diagonal matrix with diagonal entries the vector v,
+and Ji
+1 is the Jacobian matrix of the i-th weight matrix with respect to �
+W1, which for any
+u1 ∈ V1 and u2 ∈ Vi, has entries
+[Ji
+1]u1,u2 =
+∂
+∂ �w1(u1) �wi(u2) .
+Next, in the following two lemmas, we express the time derivative of the Laplacian L0
+and L
+up
+1
+as functions of E(t). The proofs of these results are straightforward and omitted
+for brevity. In what follows, Sym[A] denotes the symmetric part of the matrix A, namely
+Sym[A] = (A + A⊤)/2.
+Lemma 5.2 (Derivative of L0). For the simplicial complex K with the initial edges’ weight
+matrix W1 and fixed perturbation norm ε, let E(t) be a smooth path and �
+W0, �
+W1, �
+W2 be cor-
+responding perturbed weight matrices. Then,
+1
+2ε
+d
+dtL0(t) = �
+W −1
+0 B1�
+W1E
+•
+B⊤
+1 �
+W −1
+0
+− Sym
+�
+�
+W −1
+0
+diag
+�
+J0
+1E
+•
+1
+�
+L0
+�
+.
+(5.4)
+
+14
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+Lemma 5.3 (Derivative of L
+up
+1 ). For the simplicial complex K with the initial edges’ weight
+matrix W1 and fixed perturbation norm ε, let E(t) be a smooth path and �
+W0, �
+W1, �
+W2 be cor-
+responding perturbed weight matrices. Then,
+1
+2ε
+d
+dtL
+up
+1 (t) = − Sym
+�
+�
+W −1
+1 B2�
+W 2
+2 B⊤
+2 �
+W −1
+1 E
+• �
+W −1
+1
+�
++
+(5.5)
++ �
+W −1
+1 B2�
+W2 diag
+�
+J0
+1E
+•
+1
+�
+B⊤
+2 �
+W −1
+1
+(5.6)
+Combining Theorem 5.1 with Lemma 5.2 and Lemma 5.3 we obtain the following expres-
+sion for the gradient of the functional.
+Theorem 5.4 (The free gradient of F(ε, E)).
+Assume the initial weight matrices W0, W1
+and W2, as well as the parameters ε > 0, α > 0 and µ > 0, are given. Additionally assume
+that E(t) is a differentiable matrix-valued function such that the first non-zero eigenvalue
+λ+(ε, E) of L
+up
+1 (ε, E) and the second smallest eigenvalue µ2(ε, E) of L0(ε, E) are simple. Let
+�
+W0, �
+W1, �
+W2 be corresponding perturbed weight matrices; then:
+1
+ε∇EF(ε, E)(t) = λ+(ε, E)
+�
+Sym
+�
+−�
+W −1
+1 B2�
+W 2
+2 B⊤
+2 �
+W −1
+1 x+x⊤
++�
+W −1
+1
+�
++ diag
+�
+J2
+1
+⊤ diagvec
+�
+B⊤
+2 �
+W −1
+1 x+x⊤
++�
+W −1
+1 B2�
+W2
+�� �
+−
+− α
+µ max
+�
+0, 1 − µ2(ε, E)
+µ
+� �
+B⊤
+1 �
+W −1
+0 y2y⊤
+2 �
+W −1
+0 B1�
+W1−
+− diag
+�
+J0
+1
+⊤ diagvec
+�
+Sym[�
+W −1
+0 y2y⊤
+2 L0]
+�� �
+where x+ is a unit eigenvector of L
+up
+1
+corresponding to λ+, y2 is a unit eigenvector of L0
+corresponding to µ2, and the operator diagvec(X) returns the main diagonal of X as a vector.
+Proof. To derive the expression for the gradient ∇EF, we exploit the chain rule for the
+time derivative: λ
+•
+= ⟨ d
+dtA(E(t)), xx⊤⟩ = ⟨∇Eλ, E
+•
+⟩. Then it is sufficient to apply the cyclic
+perturbation for the scalar products of Lemma 5.2 and Lemma 5.3 with x+x⊤
++ and y2y⊤
+2
+respectively. The final transition requires the formula:
+⟨A, diag(BE1))⟩ =
+�
+diag
+�
+B⊤(diagvec A)
+�
+, E
+�
+Remark 5.5. The derivation above assumes the simplicity of both µ2(ε, E) and λ+(ε, E).
+This assumption is not too restrictive as simplicity for these extremal eigenvalues is a generic
+property. We observe simplicity in all our numerical tests.
+5.3. The constrained gradient system and its stationary points. In this section we are
+deriving from the free gradient determined in Theorem 5.4 the constrained gradient of the
+considered functional, that is the projected gradient (with respect to the Frobenius inner
+product) onto the manifold Ω∩Πε, composed of perturbations E which preserve the structure
+of W and the unit norm constraint of E.
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+15
+In order to obtain the constrained gradient system, we need to project the unconstrained
+gradient given by Theorem 5.4 onto the feasible set and also to normalize E to preserve its
+unit norm. On a time interval where the set of 0-weight edges remains unchanged, the norm-
+unconstrained gradient is given by (a formal derivation can be established through the KKT
+conditions):
+P+G(ε, E)
+(5.7)
+where P+ is the non-negativity projection such that
+[P+X]ij =
+�
+Xij,
+[W1 + εE]ij > 0
+0,
+otherwise
+.
+Further, in order to comply with the constraint ∥E(t)∥2 = 1, we must have
+(5.8)
+0 = 1
+2
+d
+dt∥E(t)∥2 = ⟨E(t), E
+•
+(t)⟩.
+We are thus led to the following constrained optimization problem for the admissible direction
+of the steepest descent.
+Lemma 5.6 (Direction of steepest admissible descent).
+Let E, G ∈ Rn,n with G given by
+(5.7) and ∥E∥F = 1. On a time interval where the set of 0-weight edges remains unchanged,
+the gradient system reads
+E
+•
+(t) = −P+G(ε, E(t)) + κP+E(t),
+where
+κ = ⟨ε, G(E(t)), P+E(t)⟩
+∥P+E(t)∥2
+.
+(5.9)
+Proof. We need to orthogonalize E
+•
+(t) with respect to E(t). As usual, this can be obtained
+by the introduction of a linear orthogonality correction. The gradient system reads
+(5.10)
+E
+•
+= P+(−G − κE),
+where κ is determined from the constraint ⟨E, E
+•
+⟩ = 0. We then have
+0 = ⟨E, E
+•
+⟩ = ⟨E, P+(−G − κE)⟩ = −⟨P+E, G⟩ − κ⟨P+E, P+E⟩,
+and the result follows.
+Equation (5.9) suggests that the systems goes “primarily” in the direction of the antigra-
+dient −G(E, ε), thus the functional is expected to decrease along it.
+Lemma 5.7 (Monotonicity). Let E(t) of unit Frobenius norm satisfy the differential equa-
+tion (5.10) with G given by (5.7). Then, the functional Fk(ε, E)(t) decreases monotonically
+with t.
+
+16
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+Proof. We consider first the simpler case where the non-negativity projection does nt apply
+so that G = G(E, ε) (without P+). Then
+d
+dtF(ε, E)(t) =
+�
+∇EFk(ε, E), E
+• �
+= ⟨εG(ε, E(t)), −G(ε, E(t)) + κE(t)⟩ =
+= −ε∥G(ε, E)∥2 + ε⟨G(ε, E), E⟩
+⟨E, E⟩
+⟨G(ε, E), E⟩ =
+= ε
+�
+−∥G(ε, E)∥2 + |⟨G(ε, E), E⟩|2
+∥E∥2
+�
+≤ 0
+(5.11)
+where the final estimate is given by the Cauchy-Bunyakovsky-Schwarz inequality. The derived
+inequality holds on the time interval without the change in the support of P+ (so that no new
+edges are prohibited by the non-negativity projection).
+As a direct consequence, we observe that the stationary points of the differential equation
+(5.10) are characterized as follows.
+Theorem 5.8 (Stationary points). Let E⋆ be an admissible perturbation with ∥E⋆∥F = 1 be
+such that
+(i) The eigenvalue λ+(ε, E) is simple at E = E⋆ and depends continuously on E in a neigh-
+borhood of E⋆.
+(ii) Penalization is not active, i.e. µ2(ε, E) > µ.
+(iii) The gradient G(ε, E⋆) is nonzero.
+Let E(t) be the solution of (5.10) passing through E⋆. Then the following are equivalent:
+1.
+d
+dtF (ε, E(t)) = 0.
+2. E
+•
+= 0.
+3. E⋆ is a real multiple of G(ε, E⋆).
+Proof. It is immediate to see that 3. implies 2., which implies 1. The proof is concluded
+noting that (5.11) shows that 1. implies 3 by the strict form of Cauchy-Bunyakovsky-Schwarz
+inequality.
+5.4. Free Gradient Transition in the Outer Level. The optimization problem in the inner
+level is non-convex due to the non-convexity of the functional F(ε, E). Hence, for a given ε,
+the computed minimizer E(ε) may depend on the initial guess E0 = E0(ε).
+The effects of the initial choice are particularly important upon the transition �ε → ε =
+�ε + ∆ε between constrained inner levels: given reasonably small ∆ε, one should expect rela-
+tively close optimizers E(�ε) and E(ε), and, hence, the initial guess E0(ε) being close to and
+dependent on E(ε).
+This choice, which seems very natural, determines however a discontinuity
+F(�ε, E(�ε)) ̸= F(ε, E(�ε)),
+which may prevent the expected monotonicity property with respect to ε in the (likely unusual
+case) where F(�ε, E(�ε)) < F(ε, E(�ε)). This may happen in particular when ∆ε is not taken
+small; since the choice of ∆ε is driven by a Newton-like iteration we are interested in finding
+a way to prevent this situation and making the whole iterative method more robust. The goal
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+17
+E0(ε0)
+E(ε0)
+constrain. E(t)
+E0(ε1)
+E(ε1)
+constrain. E(t)
+free flow, �E(t)
+ε0 → ε0 + ∆ε0
+. . .
+E0(εv)
+E(εv)
+constrain. E(t)
+F(εv, E(εv)) = 0
+Figure 5. The scheme of alternating constrained (blue) and free gradient (red) flows. Each stage inherits
+the final iteration of the previous stage as initial E0(εi) or �E0(εi) respectively. The scheme alternates until the
+target functional vanishes (F(ε, E) = 0).
+is that of guaranteeing monotonicity of the functional both with respect to time and with
+respect to ε.
+When in the outer iteration we increase ε from a previous value �ε < ε, we have the problem
+of choosing a suitable initial value for the constrained gradient system (5.9), such that at the
+stationary point E(�ε) we have F(�ε, E(�ε)) < F(ε, E(ε)) (which we assume both positive, that
+is on the left of the value ε⋆ which identifies the closest zero of the functional).
+In order to maintain monotonicity with respect to time and also with respect to ε, it is
+convenient to start to look at the optimization problem with value ε, with the initial datum
+δW1 = �εE(�ε) of norm �ε < ε.
+This means we have essentially to optimize with respect to the inequality constraint
+∥δW1∥F ≤ ε, or equivalently solve the problem (now with inequality constrain on ∥E∥F ):
+E(ε) =
+arg min
+E∈Ω,∥E∥F ≤1
+F(ε, E)
+The situation changes only slightly from the one discussed above. If ∥E∥F < 1, every
+direction is admissible, and the direction of the steepest descent is given by the negative
+gradient. So we choose the free gradient flow
+(5.12)
+E
+•
+= −P+G(ε, E(t))
+as long as ∥E(t)∥F < 1.
+When ∥E(t)∥F = 1, then there are two possible cases. If ⟨P+G(ε, E), E⟩ ≥ 0, then the solution
+of (5.12) has
+d
+dt∥E(t)∥2
+F = 2 ⟨E
+•
+, E⟩ = −2 ⟨P+G(ε, E(t)), E⟩ ≤ 0,
+and hence the solution of (5.12) remains of Frobenius norm at most 1.
+Otherwise, if ⟨P+G(ε, E), E⟩ < 0, the admissible direction of steepest descent is given by
+the right-hand side of (5.9), i.e. −P+G(ε, E) + κE, κ = ⟨G(ε,E), P+E⟩
+∥P+E∥2
+and so we choose that
+differential equation to evolve E. The situation can be summarized as taking, if ∥E(t)∥F = 1,
+(5.13)
+E
+•
+= −P+G(ε, E) + µE
+with µ = min
+�
+0, κ
+�
+with κ =
+⟨G(ε,E), P+E⟩
+∥P+E∥2
+.
+Along the solutions of (5.13), the functional F decays monotoni-
+cally, and stationary points of (5.13) with P+G(ε, E(t)) ̸= 0 are characterized, by the same
+
+18
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+Algorithm 6.1 Pseudo-code of the complete constrained- and free-gradient flow.
+Require: initial edge perturbation guess E0; initial ε0 > 0; ε-stepsize ∆ε > 0; bounds α∗, α∗
+for the α-phase;
+1: α, E ← AlphaPhase(E0, ε0, α∗, α∗)
+▷ for details see Section 7
+2: while |F(ε, E)| < 10−6 do
+3:
+ε ← ε + ∆ε
+4:
+E ←
+ε
+ε+∆εE
+▷ before the free gradient ∥E∥ < 1
+5:
+Ei ← FreeGradientFlow(E, ∆ε, ε)
+▷ see Subsection 5.4
+6:
+E ← ConstrainedGradientFlow(E, ε)
+▷ see Section 6
+7: end while
+arguments used before, as
+(5.14)
+E is a negative real multiple of P+G(ε, E(t)).
+If it can be excluded that the gradient P+G(ε, E(t)) vanishes at an optimizer, it can thus be
+concluded that the optimizer of the problem with inequality constraints is a stationary point
+of the gradient flow (5.9) for the problem with equality constraints.
+Remark 5.9. As a result, F(ε, E(t)) monotonically decreases both with respect to time t
+and to the value of the norm ε, when ε ≤ ε⋆.
+6. Algorithm details. In Algorithm 6.1 we provide the pseudo-code of the whole bi-level
+procedure. The initial “α-phase” is used to choose an appropriate value for the regularization
+parameter α. In order to avoid the case in which the penalizing term on the right-hand side of
+(4.2) dominates the loss F(ε, E(t)) in the early stages of the descent flow, we select α by first
+running such an initial phase, prior to the main alternated constrained/free gradient loop. In
+this phase, we fix a small ε = ε0 and run the constrained gradient integration for an initial
+α = α∗. After the computation of a local optimum E∗, we then increase α and rerun for the
+same ε0 with E∗ as starting point. We iterate until no change in E∗ is observed or until α
+reaches an upper bound α∗.
+The resulting value of α and E∗ are then used in the main loop where we first increase
+ε by the chosen step size, we rescale Ei by 0 < ε/(ε + ∆ε) < 1, and then we perform the
+free gradient integration described in Subsection 5.4 until we reach a new point Ei on the
+unit sphere ∥Ei∥ = 1. Then, we perform the inner constrained gradient step by integrating
+Equation (5.9), iterating the following two-step norm-corrected Euler scheme:
+�
+Ei+1/2 = Ei − hi (P+G(Ei, ε) − κiP+Ei) .
+Ei+1 = PΠεEi+1/2/∥PΠεEi+1/2∥
+(6.1)
+where the second step is necessary to numerically guarantee the Euler integration remains in
+the set of admissible flows since the discretization does not conserve the norm and larger steps
+hi may violate the non-negativity of the weights.
+In both the free and constrained integration phases, since we aim to obtain the solution
+at t → ∞ instead of the exact trajectory, we favor larger steps hi given that the established
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+19
+monotonicity is conserved. Specifically, if F(ε, Ei+1) < F(ε, Ei), then the step is accepted and
+we set hi+1 = βhi with β > 1; otherwise, the step is rejected and repeated with a smaller step
+hi ← hi/β.
+Remark 6.1. The step acceleration strategy described above, where βhi is immediately
+increased after one accepted step, may lead to “oscillations” between accepted and rejected
+steps in the event the method would prefer to maintain the current step size hi. To avoid this
+potential issue, in our experiments we actually increase the step length after two consecutive
+accepted steps.
+Alternative step-length selection strategies are also possible, for example,
+based on Armijo’s rule or non-monotone stabilization techniques [20].
+Remark 6.2. When the weight of an edge e is moved to zero, we are formally reducing the
+initial complex K to a smaller �K with V1( �K) = �
+V1 = V1\{e}. While the Hodge Laplacian of the
+new �K should have a smaller dimension than the initial one, in our perturbative approach, we
+want to maintain the dimension of L1 unchanged so as to be able to explore the set of possible
+perturbations Ω(ε) ∩ Π(W1) in a continuous way. While this may create a complication in
+the general unweighted setting, as it might give rise to an almost-zero row in B1 and thus a
+degenerate almost-zero entry in σ(Lup
+1 ) which does not correspond to a different homology, we
+emphasize that in our weighted set-up, this sort of faux edges are automatically ignored. In
+fact, in our model, the weights of the 2-simplicies W2 evolve as a function of the edge weights
+W1 as in (3.1). Thus, when an entry of W1 approaches zero, the corresponding entries of W2
+approach 0 with the same order, and the weight normalization in the definition of B1 prevents
+the formation of zero rows.
+6.1. Computational costs. Each step of either the free or the constrained flows requires
+one step of explicit Euler integration along the anti-gradient −∇EF(ε, E(t)). As discussed
+in Section 5, the construction of such a gradient requires several sparse and diagonal matrix-
+vector multiplications as well as the computation of the smallest nonzero eigenvalue of both
+L
+up
+1 (ε, E) and L0(ε, E). The latter two represent the major computational requirements of
+the numerical procedure. Fortunately, as both matrices are of the form A⊤A, with A being
+either of the two boundary or co-boundary operators B2 and B
+⊤
+1 , we have that both the two
+eigenvalue problems boil down to a problem of the form
+min
+x⊥ ker A
+∥Ax∥
+∥x∥
+i.e., the computation of the smallest singular value of the sparse matrix A. This problem
+can be approached by a sparse singular value solver based on a Krylov subspace scheme for
+the pseudo inverse of A⊤A. In practice, we implement the pseudo inversion by solving the
+corresponding least squares problems
+min
+x ∥L
+up
+1 (ε, E)x − b∥,
+min
+x ∥L0(ε, E)x − b∥ ,
+which, in our experiments, we solved using the least square minimal-residual method (LSMR)
+from [15]. This approach allows us to use a preconditioner for the normal equation correspond-
+ing to the least square problem. For simplicity, in our tests we used a constant preconditioner
+computed by means of an incomplete Cholesky factorization of the initial unperturbed L
+up
+1 , or
+
+20
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+L0. Possibly, much better performance can be achieved with a tailored preconditioner that is
+efficiently updated throughout the matrix flow. We also note that the eigenvalue problem for
+the graph Laplacian L0(ε, E) may be alternatively approached by a combinatorial multigrid
+strategy [30]. However, designing a suitable preconditioning strategy goes beyond the scope
+of this work and will be the subject of future investigations.
+7. Numerical experiments. In this section, we provide several synthetic and real-world
+example applications of the proposed stability algorithms. The code for all the examples is
+available at https://github.com/COMPiLELab/HOLaGraF. All experiments are run until the
+global stopping criterion |F(ε, E(t))| < 10−6 is met and the parameters µ and α are chosen
+as follows. Concerning µ, since the effect of weight perturbations on µ2(ε, E) diminishes with
+the growth of the network, F(ε, E) becomes less sensitive to it. Thus, in the computations
+we set µ = 0.75µ2, where µ2 is the smallest nonzero eigenvalue of the initial Laplacian L0.
+As for α, we run the α-phase described in Section 6 with parameters ε0 = 10−3, α∗ = 1 and
+α∗ = 100.
+7.1. Illustrative Example. We consider here a small example of a simplicial complex K of
+order 2 consisting of eight 0-simplicies (vertices), twelve 1-simplicies (edges), four 2-simplicies
+V2 = {[1, 2, 3], [1, 2, 8], [4, 5, 6], [5, 6, 7]} and one corresponding hole [2, 3, 4, 5], hence, β1 = 1.
+By design, the dimensionality of the homology group H1 can be increased only by eliminating
+edges [1, 2] or [5, 6]; for the chosen weight profile w1([1, 2]) > w1([5, 6]), hence, the method
+should converge to the minimal perturbation norm ε = w1([5, 6]) by eliminating the edge [5, 6].
+The exemplary run of the optimization framework in time is shown on Figure 6. The top
+panel of Figure 6 provides the continued flow of the target functional F(ε, E(t)) consisting
+of the initial α-phase (in green) and alternated constrained (in blue) and free gradient (in
+orange) stages. As stated above, F(ε, E(t)) is strictly monotonic along the flow since the
+support of P+ does not change. Since the initial setup is not pathological with respect to the
+connectivity, the initial α-phase essentially reduces to a single constrained gradient flow and
+terminates after one run with α = α∗. The constrained gradient stages are characterized by a
+slow changing E(t), which is essentially due to the flow performing small adjustments to find
+the correct rotation on the unit sphere, whereas the free gradient stage quickly decreases the
+target functional.
+The second panel shows the behaviour of first non-zero eigenvalue λ+(ε, E(t)) (solid line)
+of L
+up
+1 (ε, E(t)) dropping through the ranks of σ(L1(ε, E(t))) (semi-transparent); similar to
+the case of the target functional F(ε, E(t)), λ+(ε, E(t)) monotonically decreases. The rest of
+the eigenvalues exhibit only minor changes, and the rapidly changing λ+ successfully passes
+through the connectivity threshold µ (dotted line).
+The third and the fourth panels show the evolution of the norm of the perturbation
+∥E(t)∥ and the perturbation E(t) itself, respectively. The norm ∥E(t)∥ is conserved during
+the constrained-gradient and the α- stages; these stages correspond to the optimization of
+the perturbation shape, as shown by the small positive values at the beginning of the bottom
+panel which eventually vanish. During the free gradient integration the norm ∥E(t)∥ increases,
+but the relative change of the norm declines with the growth of εi to avoid jumping over
+the smallest possible ε. Finally, due to the simplicity of the complex, the edge we want to
+eliminate, 56, dominates the flow from the very beginning (see bottom panel); such a clear
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+21
+0
+100
+200
+300
+400
+0.000
+0.002
+0.004
+0.006
+0.008
+0.010
+0.012
+functional F(ε, E(t))
+α−stage
+free gradient
+constrained gradient
+0
+100
+200
+300
+400
+0.0
+0.1
+0.2
+0.3
+spectrum σ(L1)
+target λ+
+other λi
+threshold µ
+0
+100
+200
+300
+400
+1.00
+1.25
+1.50
+1.75
+2.00
+norm ∥E(t)∥
+0
+100
+200
+300
+400
+[1 2]
+[1 3]
+[1 8]
+[2 3]
+[2 4]
+[2 8]
+[3 5]
+[4 5]
+[4 6]
+[5 6]
+[5 7]
+[6 7]
+perturbation E(t)
+−1
+−0.5
+0
+0.5
+1
+1
+2
+3
+4
+5
+6
+7
+8
+ε = 0.025
+1
+2
+3
+4
+5
+6
+7
+8
+ε = 0.075
+1
+2
+3
+4
+5
+6
+7
+8
+ε = 0.125
+1
+2
+3
+4
+5
+6
+7
+8
+ε = 0.31
+Figure 6. Illustrative run of the framework determining the topological stability: the top pane — the flow
+of the functional F(ε, E(t)); the second pane — the flow of σ(L1), λ+ is highlighted; third pane — the change
+of the perturbation norm ∥E(t)∥; the bottom pane — the heatmap of the perturbation profile E(t).
+pattern persists only in small examples, whereas for large networks the perturbation profile
+is initially spread out among all the edges.
+7.2. Triangulation Benchmark. To provide more insight into the computational behavior
+of the method, we synthesize here an almost planar graph dataset. Namely, we assume N
+uniformly sampled vertices on the unit square with a network built by the Delaunay triangula-
+tion; then, edges are randomly added or erased to obtain the sparsity ν (so that the graph has
+1
+2νN(N − 1) edges overall). An order-2 simplicial complex K = (V0, V1, V2) is then formed by
+letting V0 be the generated vertices, V1 the edges, and V2 every 3-clique of the graph; edges’
+weights are sampled uniformly between 1/4 and 3/4, namely w1(ei) ∼ U[ 1
+4, 3
+4].
+An example of such triangulation is shown in Figure 7a; here N = 8 and edges [6, 8] and
+[2, 7] were eliminated to achieve the desired sparsity.
+We sample networks with a varying number of vertices N = 10, 16, 22, 28 and vary-
+ing sparsity pattern ν = 0.35, 0.5 which determine the number of edges in the output as
+m = ν N(N−1)
+2
+. Due to the highly randomized procedure, topological structures of a sam-
+pled graph with a fixed pair of parameters may differ substantially, so 10 networks with the
+same (N, ν) pair are generated. For each network, the working time (without considering the
+sampling itself) and the resulted perturbation norm ε, and are reported in Figure 7b and
+Figure 7c, respectively. As anticipated in Subsection 6.1, we show the performance of two
+implementations of the method, one based on LSMR and one based on LSMR preconditioned
+
+22
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+1
+2
+3
+4
+5
+6
+7
+8
+(a)
+Example
+of
+Triangulation
+and
+Holes
+102
+103
+10
+102
+execution time, s
+ν = 0.35, LSMR
+LSMR, ichol
+102
+103
+10
+102
+103
+number of edges, m
+execution time, s
+ν = 0.5, LSMR
+LSMR, ichol
+(b) Time (in seconds)
+102
+103
+0.5
+1.0
+pertubation norm
+ν = 0.35, LSMR
+102
+103
+0.5
+1.0
+number of edges, m
+perturbation norm
+ν = 0.5, LSMR
+(c) Perturbation norm, ε
+Figure 7. Benchmarking Results on the Synthetic Triangulation Dataset: varying sparsities ν = 0.35, 0.5
+and N = 16, 22, 28, 34, 40; each network is sampled 10 times. Shapes correspond to the number of eliminated
+edges in the final perturbation: 1 : �, 2 : □, 3 : �, 4 : △. For each pair (ν, N), the un-preconditioned and
+Cholesky-preconditioned execution times are shown.
+by using the incomplete Cholesky factorization of the initial matrices. We observe that,
+• the computational cost of the whole procedure lies between O(m2) and O(m3)
+• denser structures, with a higher number of vertices, result in the higher number of edges
+being eliminated; at the same time, even most dense cases still can exhibit structures
+requiring the elimination of a single edge, showing that the flow does not necessarily favor
+multi-edge optima;
+• the required perturbation norm ε is growing with the size of the graph, Figure 7c, but not
+too fast: it is expected that denser networks would require larger ε to create a new hole;
+at the same time if the perturbation were to grow drastically with the sparsity ν, it would
+imply that the method tries to eliminate sufficiently more edges, a behavior that resembles
+convergence to a sub-optimal perturbation;
+• preconditioning with a constant incomplete Cholesky multiplier, computed for the initial
+Laplacians, provides a visible execution time gain for medium and large networks. Since
+the quality of the preconditioning deteriorates as the flow approaches the minimizer (as a
+non-zero eigenvalue becomes 0), it is worth investigating the design of a preconditioner for
+the up-Laplacian that can be efficiently updated.
+7.3. Transportation Networks. Finally, we provide an application to real-world examples
+based on city transportation networks. We consider networks for Bologna, Anaheim, Berlin
+Mitte, and Berlin Tiergarten; each network consists of nodes — intersections/public transport
+stops — connected by edges (roads) and subdivided into zones; for each road the free flow time,
+length, speed limit are known; moreover, the travel demand for each pair of nodes is provided
+through the dataset of recorded trips. All the datasets used here are publicly available at
+https://github.com/bstabler/TransportationNetworks; Bologna network is provided by the
+Physic Department of the University of Bologna (enriched through the Google Maps API
+https://developers.google.com/maps).
+The regularity of city maps naturally lacks 3-cliques, hence forming the simplicial complex
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+23
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+24
+25
+26
+27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+Bologna: Regional Network
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+24
+25
+26
+27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+1st original eigenflow
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+24
+25
+26
+27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+2nd original eigenflow
+1
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+24
+25
+26
+27
+28
+29
+30
+31
+32
+33
+34
+35
+36
+37
+38
+39
+40
+41
+42
+43
+44
+45
+46
+47
+48
+49
+50
+51
+52
+53
+54
+55
+56
+57
+58
+59
+60
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+−∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+∥v∥
+created eigenflow
+Figure 8. Example of the Transportation Network for Bologna. Left pane: original zone graph where the
+width of edges corresponds to the weight, to-be-eliminated edge is colored in red. Right pane: eigenflows, original
+and created; color and width correspond to the magnitude of entries.
+based on triangulations as done before frequently leads to trivial outcomes. Instead, here
+we “lift” the network to city zones, thus more effectively grouping the nodes in the graph.
+Specifically:
+1. we consider the completely connected graph where the nodes are zones in the city/region;
+2. the free flow time between two zones is temporarily assigned as a weight of each edge: the
+time is as the shortest path between the zones (by the classic Dijkstra algorithm) on the
+initial graph;
+3. similarly to what is done in the filtration used for persistent homology, we filter out exces-
+sively distant nodes; additionally, we exclude the longest edges in each triangle in case it
+is equal to the sum of two other edges (so the triangle is degenerate and the trip by the
+longest edge is always performed through to others);
+4. finally, we use the travel demand as an actual weight of the edges in the final network;
+travel demands are scaled logarithmically via the transformation wi �→ log10
+�
+wi
+0.95 min wi
+�
+;
+see the example on the left panel of Figure 8.
+Given the definition of weights in the network, high instability (corresponding to small per-
+turbation norm ε) implies structural phenomena around the “almost-hole”, where the faster
+and shorter route is sufficiently less demanded.
+In the case of Bologna, Figure 8, the algorithm eliminates the edge [11, 47] (Casalecchio
+
+24
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+Table 1
+Topological instability of the transportation networks: filtered zone networks with the corresponding pertur-
+bation norm ε and its percentile among w1(·) profile. For each simplicial complex the number of nodes, edges
+and triangles in V2(K) are provided alongside the initial number of holes β1. The results of the algorithm consist
+of the perturbation norm, ε, computation time, and approximate percentile p.
+network
+β1
+logarithmic weights
+n
+m
+∆
+time
+ε
+p
+Bologna
+60
+175
+171
+2
+2.43s
+0.65
+0.003
+[11, 47] (4th smallest)
+Anaheim
+38
+159
+221
+1
+5.39s
+0.57
+0.003
+[10, 29] (11th smallest)
+Berlin-Tiergarten
+26
+63
+55
+0
+2.46s
+1.18
+0.015
+[6, 16] (20th smallest)
+Berlin-Mitte
+98
+456
+900
+1
+127s
+0.887
+0.0016
+[57, 87] (6th), [58, 87], (17th)
+di Reno – Pianoro) creating a new hole 6 − 11 − 57 − 47. We also provide examples of the
+eigenflows in the kernel of the Hodge Laplacian (original and additional perturbed): original
+eigenvectors correspond to the circulations around holes 7−26−12−20 and 8−21−20−16−37
+non-locally spread in the neighborhood [29].
+The results for four different networks are summarized in the Table 1; p mimics the
+percentile, ε/ �
+e∈V1 wi(e), showing the overall small perturbation norm contextually.
+At
+the same time, we emphasize that except Bologna (which is influenced by the geographical
+topology of the land), the algorithm does not choose the smallest weight possible; indeed,
+given our interpretation of the topological instability, the complex for Berlin-Tiergarten is
+stable and the transportation network is effectively constructed.
+8. Discussion. In the current work, we formulate the notion of k-th order topological
+stability of a simplicial complex K as the smallest data perturbation required to create one
+additional k-th order hole in K. By introducing an appropriate weighting and normaliza-
+tion, the stability is reduced to a matrix nearness problem solved by a bi-level optimization
+procedure. Despite the highly nonconvex landscape, our proposed procedure alternating con-
+strained and free gradient steps yields a monotonic descending scheme. Our experiments show
+that this approach is generally successful in computing the minimal perturbation increasing
+β1(ε, E), even for potentially difficult cases, as the one proposed in Subsection 7.1.
+For simplicity, here we limit our attention to the smallest perturbation that introduces
+only one new hole. However, a simple modification may be employed to address the case
+of the introduction of m new holes: include the sum of m nonzero eigenvalues of Lup
+1 (ε, E)
+rather than just the first one in the spectral functional (4.2). We also remark that, due to
+the spectral inheritance principle Theorem 2.7, the proposed framework for H1 can be in
+principle extended to a general Hk; however, this extension requires nontrivial considerations
+on the data modification procedure and on the numerical linear algebra tools, as a nontrivial
+topology of higher-order requires a much denser network.
+Different improvements are possible in terms of numerical implementation, including in-
+vestigating the use of more sophisticated (e.g.
+implicit) integrators for the gradient flow
+system (Equation (5.9)), which would additionally require the use of higher-order derivatives
+
+QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY
+25
+of λ+(ε, E). Moreover, as already mentioned in Subsection 6.1, the numerical method for the
+computation of the small singular values would benefit from the use of an efficient precondi-
+tioner that can be effectively updated throughout the flow. Investigations in this direction are
+in progress and will be the subject of future work.
+REFERENCES
+[1] K. M. Altenburger and J. Ugander, Monophily in social networks introduces similarity among
+friends-of-friends, Nature human behaviour, 2 (2018), pp. 284–290.
+[2] E. Andreotti, D. Edelmann, N. Guglielmi, and C. Lubich, Graph partitioning using matrix differ-
+ential equations, 2017, https://doi.org/10.48550/ARXIV.1712.05993.
+[3] F. Arrigo, D. J. Higham, and F. Tudisco, A framework for second-order eigenvector centralities and
+clustering coefficients, Proceedings of the Royal Society A, 476 (2020), p. 20190724.
+[4] F. Battiston, G. Cencetti, I. Iacopini, V. Latora, M. Lucas, A. Patania, J.-G. Young, and
+G. Petri, Networks beyond pairwise interactions: Structure and dynamics, Physics Reports, 874
+(2020), pp. 1–92, https://doi.org/10.1016/j.physrep.2020.05.004.
+[5] A. R. Benson, Three hypergraph eigenvector centralities, SIAM Journal on Mathematics of Data Science,
+1 (2019), pp. 293–312.
+[6] A. R. Benson, R. Abebe, M. T. Schaub, A. Jadbabaie, and J. Kleinberg, Simplicial closure
+and higher-order link prediction, Proceedings of the National Academy of Sciences, 115 (2018),
+pp. E11221–E11230.
+[7] A. R. Benson, D. F. Gleich, and J. Leskovec, Higher-order organization of complex networks, Science,
+353 (2016), pp. 163–166, https://doi.org/10.1126/science.aad9029.
+[8] A. R. Benson, D. F. Gleich, and J. Leskovec, Higher-order organization of complex networks, Science,
+353 (2016), pp. 163–166.
+[9] C. Bick, E. Gross, H. A. Harrington, and M. T. Schaub, What are higher-order networks?, CoRR,
+abs/2104.11329 (2021), https://arxiv.org/abs/2104.11329.
+[10] Y.-C. Chen and M. Meila, The decomposition of the higher-order homology embedding constructed from
+the k-Laplacian, Advances in Neural Information Processing Systems, 34 (2021), pp. 15695–15709.
+[11] Y.-C. Chen, M. Meil˘a, and I. G. Kevrekidis, Helmholtzian eigenmap: Topological feature discovery
+and edge flow learning from point cloud data, rXiv:2103.07626, (2021), https://doi.org/10.48550/
+ARXIV.2103.07626.
+[12] S. Ebli and G. Spreemann, A notion of harmonic clustering in simplicial complexes, in 2019 18th IEEE
+International Conference On Machine Learning And Applications (ICMLA), IEEE, 2019, pp. 1083–
+1090.
+[13] E. Estrada and D. J. Higham, Network properties revealed through matrix functions, SIAM review, 52
+(2010), pp. 696–714.
+[14] M. Fiedler, Laplacian of graphs and algebraic connectivity, Banach Center Publications, 25 (1989),
+pp. 57–70.
+[15] D. C.-L. Fong and M. Saunders, Lsmr: An iterative algorithm for sparse least-squares problems, SIAM
+Journal on Scientific Computing, 33 (2011), pp. 2950–2971.
+[16] S. Fortunato and D. Hric, Community detection in networks: A user guide, Physics reports, 659
+(2016), pp. 1–44.
+[17] K. Fountoulakis, P. Li, and S. Yang, Local hyper-flow diffusion, Advances in Neural Information
+Processing Systems, 34 (2021), pp. 27683–27694.
+[18] N. E. Friedkin, Theoretical foundations for centrality measures, American journal of Sociology, 96 (1991),
+pp. 1478–1504.
+[19] J. Ghaderi and R. Srikant, Opinion dynamics in social networks with stubborn agents: Equilibrium
+and convergence rate, Automatica, 50 (2014), pp. 3209–3215.
+[20] L. Grippo, F. Lampariello, and S. Lucidi, A class of nonmonotone stabilization methods in uncon-
+strained optimization, Numerische Mathematik, 59 (1991), pp. 779–805.
+[21] N. Guglielmi and C. Lubich, Matrix nearness problems and eigenvalue optimization, in preparation,
+
+26
+N. GUGLIELMI, A. SAVOSTIANOV, AND F. TUDISCO
+2022.
+[22] D. Horak and J. Jost, Spectra of combinatorial Laplace operators on simplicial complexes, Advances
+in Mathematics, 244 (2013), pp. 303–336.
+[23] R. A. Horn and C. R. Johnson, Matrix Analysis, Cambridge University Press, 1990.
+[24] L.-H. Lim, Hodge Laplacians on graphs, SIAM Review, (2015), pp. 685–715.
+[25] A. Muhammad and M. Egerstedt, Control using higher order laplacians in network topologies, in Proc.
+of 17th International Symposium on Mathematical Theory of Networks and Systems, Citeseer, 2006,
+pp. 1024–1038.
+[26] B. Nettasinghe, V. Krishnamurthy, and K. Lerman, Diffusion in social networks:
+Effects of
+monophilic contagion, friendship paradox and reactive networks, IEEE Transactions on Network Sci-
+ence and Engineering, (2019).
+[27] M. E. Newman, Finding community structure in networks using the eigenvectors of matrices, Physical
+review E, 74 (2006), p. 036104.
+[28] N. Otter, M. A. Porter, U. Tillmann, P. Grindrod, and H. A. Harrington, A roadmap for the
+computation of persistent homology, EPJ Data Science, 6 (2017), pp. 1–38.
+[29] M. T. Schaub, A. R. Benson, P. Horn, G. Lippner, and A. Jadbabaie, Random walks on simplicial
+complexes and the normalized Hodge 1-Laplacian, SIAM Review, 62 (2020), pp. 353–391.
+[30] D. A. Spielman and S.-H. Teng, Nearly linear time algorithms for preconditioning and solving sym-
+metric, diagonally dominant linear systems, SIAM Journal on Matrix Analysis and Applications, 35
+(2014), pp. 835–885.
+[31] S. H. Strogatz, From kuramoto to crawford: exploring the onset of synchronization in populations of
+coupled oscillators, Physica D: Nonlinear Phenomena, 143 (2000), pp. 1–20.
+[32] F. Tudisco, F. Arrigo, and A. Gautier, Node and layer eigenvector centralities for multiplex networks,
+SIAM Journal on Applied Mathematics, 78 (2018), pp. 853–876.
+[33] F. Tudisco, A. R. Benson, and K. Prokopchik, Nonlinear higher-order label spreading, in Proceedings
+of the Web Conference 2021, 2021, pp. 2402–2413.
+[34] F. Tudisco and M. Hein, A nodal domain theorem and a higher-order cheeger inequality for the graph
+p-laplacian, (2016), https://arxiv.org/abs/1602.05567.
+[35] F. Tudisco and D. J. Higham, Core-periphery detection in hypergraphs, SIAM Journal on Mathematics
+of Data Science, (to appear).
+[36] F. Tudisco, P. Mercado, and M. Hein, Community detection in networks via nonlinear modularity
+eigenvectors, SIAM Journal on Applied Mathematics, 78 (2018), pp. 2393–2419.
+[37] S. Vigna, Spectral ranking, Network Science, 4 (2016), pp. 433–445.
+
diff --git a/W9E2T4oBgHgl3EQfDwYa/content/tmp_files/load_file.txt b/W9E2T4oBgHgl3EQfDwYa/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..87e160b578c7a3ede7e5988d62fa050ff35694b2
--- /dev/null
+++ b/W9E2T4oBgHgl3EQfDwYa/content/tmp_files/load_file.txt
@@ -0,0 +1,1436 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf,len=1435
+page_content='Quantifying the structural stability of simplicial homology Nicola Guglielmi∗, Anton Savostianov† , and Francesco Tudisco† Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The homology groups of a simplicial complex reveal fundamental properties of the topology of the data or the system and the notion of topological stability naturally poses an important yet not fully investigated question.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In the current work, we study the stability in terms of the smallest perturbation sufficient to change the dimensionality of the corresponding homology group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Such definition requires an appropriate weighting and normalizing procedure for the boundary operators acting on the Hodge algebra’s homology groups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Using the resulting boundary operators, we then formulate the question of structural stability as a spectral matrix nearness problem for the corre- sponding higher-order graph Laplacian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We develop a bilevel optimization procedure suitable for the formulated matrix nearness problem and illustrate the method’s performance on a variety of synthetic quasi-triangulation datasets and transportation networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Key words.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' simplicial complexes, homology groups, graph Laplacian, Hodge Laplacian, matrix nearness prob- lems, matrix ODEs, spectral optimization, constrained gradient system MSC codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 05C50, 65F45, 65K10, 57M15, 62R40 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Introduction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Models based on graphs and networks are ubiquitous in the sciences and engineering;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' for example, they have been successfully applied to model chemical reactions, traffic and electric flows, social interactions, and to describe abstract datasets in machine learning pipelines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Graph properties can be used to determine important nodes [18, 37, 13], reveal modular structure of a system [16, 36, 27], model collective network dynamics such as synchronization [31] and opinion formation [19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' However, models based on graphs are limited to descriptions based on pairwise node-node relationships.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While graph-based models are widely used and successful, many complex systems and datasets are better described by higher-order relations that go beyond pairwise interactions [4, 7, 9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Relational data is full of interactions that happen in groups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For example, friendship relations often involve groups that are larger than two individuals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In fact, monophily and triadic closure principles from the social sciences suggest that motifs, such as triangles, are important building blocks of relational data [1, 3, 26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Also in the presence of point-cloud data, directly modeling higher-order data interactions has led to improvements in numerous data mining settings, including clustering [8, 17, 33], link prediction [3, 6], and ranking [5, 32, 35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Simplicial complexes are standard higher-order network models, where simplicies of dif- ferent order can connect a larger number of nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Higher-order Laplacians are key algebraic tools that naturally correspond to a simplicial complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Formally, these are a sequence of lin- ear operators that generalizes the better-known graph Laplacian, obtained when only pairwise edge relations are considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Very useful topological properties about the data are revealed by the kernels of these operators which, by the Fundamental Lemma of Homology, define a homology of the data and reveal fundamental properties such as connected components, holes, and voids.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ∗Gran Sasso Science Institute, L’Aquila, Italy (nicola.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='guglielmi@gssi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='it, anton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='savostianov@gssi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='it, francesco.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='tudisco@gssi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='it).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='03627v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='NA] 9 Jan 2023 2 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO In this work, we are concerned with quantifying the stability of such homological proper- ties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' More precisely, given an initial simplicial complex, we develop a numerical method based on a suitable matrix differential equation that allows us to compute the closest simplicial complex with different homology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The precise meaning of being “close” and being “different” will be based on the structure of the specific higher-order Laplacian matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While a great effort has been devoted in recent years to measure the presence and persistence of simplicial homology [12, 28], very little work is available about the stability of the homology classes with respect to data perturbation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to make a sound mathematical formulation of the problem we aim to solve, and of the numerical model we design for its solution, in Section 2 we review in detail the notion of homology simplicial complexes and the corresponding higher-order Laplacians.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In Subsection 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 and Subsection 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 we discuss how these operators may be extended to account for weighted higher-order node relations and formulate the corresponding stability problem in Section 3 and Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, in Section 5 and Section 6 we present our numerical method based on a two-level constrained matrix gradient flow approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Finally, we devote Section 7 to illustrate the performance of the proposed numerical scheme on several example datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Simplicial complexes and higher-order relations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A graph G is a pair of sets (V, E), where V = {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' , n} is the set of vertices and E ⊂ V×V is a set of unordered pairs representing the undirected edges of G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We let m denote the number of edges E = {e1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' , em} and we assume them ordered lexicographically, with the convention that i < j for any {i, j} ∈ E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For brevity, we often write ij in place of {i, j} to denote the edge joining i and j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Moreover, we assume no self-loops, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=', ii /∈ E for all i ∈ V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A graph only considers pairwise relations between the vertices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A simplicial complex K is a generalization of a graph that allows us to model connections involving an arbitrary number of nodes by means of higher-order simplicies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Formally, a k-th order simplex (or k-simplex, briefly) is a set of k + 1 vertices {i0, i1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' , ik} with the property that every subset of k nodes itself is a (k − 1)-simplex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Any (k − 1)-simplex of a k-simplex is called a face.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The collection of all such simplices forms a simplicial complex K, which therefore essentially consists of a collection of sets of vertices such that every subset of the set in the collection is in the collection itself.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, a graph G can be thought of as the collection of 0- and 1- simplices: the 0-simplices form the nodes set of G, while 1-simplices form its edges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To emphasize this analogy, in the sequel we often specify that K is a simplicial complex on the vertex set V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Just like the edges of a graph, to any simplicial complex, we can associate an orientation (or ordering).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To underline that an ordering has been fixed, we denote an ordered k-simplex σ using square brackets σ = [i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In particular, as for the case of edges, in this work, we always assume the lexicographical ordering, unless specified otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' That is, we assume that: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' any k-simplex [i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik] in K is such that i0 < · · · < ik;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the k-simplices σ1, σ2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' of K are ordered so that σi ≺ σi+1 for all i, where [i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik] ≺ [i′ 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' i′ k] if and only if there exists h such that 0 ≤ h ≤ k, ij = i′ j for j = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' , h and ih < i′ h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As for the edges, we often write i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik in place of [i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik] in this case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 3 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Homology, boundary operators and higher-order Laplacians.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Topological proper- ties of a simplicial complex can be studied by considering boundary operators, higher-order Laplacians, and the associated homology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Here we recall these concepts trying to emphasize their matrix-theoretic flavor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To this end, we first fix some further notation and recall the notion of a real k-chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Assume K is a simplicial complex on the vertex set V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For k ≥ 0, we denote the set of all the oriented k-th order simplices in K as Vk(K) or simply Vk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, V0 = V and V1 = E form the underlying graph of K, which we denote by GK = (V, E) = (V0, V1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The formal real vector space spanned by all the elements of Vk with real coefficients is denoted by Ck(K).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Any element of Ck(K), the formal linear combinations of simplices in Vk, is called a k-chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We remark that, in the graph-theoretic terminology, C0(K) is usually called the space of vertices’ states, while C1(K) is usually called the space of flows in the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The chain spaces are finite vector spaces generated by the set of k-simplicies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The bound- ary and co-boundary operators are particular linear mappings between Ck and Ck−1, which in a way are the discrete analogous of high-order differential operators (and their adjoints) on continuous manifolds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The boundary operator ∂k maps a k-simplex to an alternating sum of its (k − 1)-dimensional faces obtained by omitting one vertex at a time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Its precise definition is recalled below, while Figure 1 provides an illustrative example of its action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let k ≥ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Given a simplicial complex K over the set V, the boundary op- erator ∂k : Ck(K) �→ Ck−1(K) maps every ordered k-simplex [i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik] ∈ Ck(K) to the following alternated sum of its faces: ∂k[i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik] = k � j=0 (−1)j[i0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ij−1ij+1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ik] ∈ Ck−1(K) As we assume the k-simplicies in Vk are ordered lexicographically, we can fix a canonical basis for Ck(K) and we can represent ∂k as a matrix Bk with respect to such basis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In fact, once the ordering is fixed, Ck(K) is isomorphic to RVk, the space of functions from Vk to R or, equivalently, the space of real vectors with |Vk| entries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, Bk is a |Vk−1| × |Vk| matrix and ∂∗ k coincides with B⊤ k .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We shall always assume the canonical basis for Ck(K) is fixed in this way and we will deal exclusively with the matrix representation Bk from now on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' An example of Bk for k = 1 and k = 2 is shown in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A direct computation shows that the following fundamental identity holds (see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [24, Thm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7]) (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1) BkBk+1 = 0 for any k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This identity is also known in the continuous case as the Fundamental Lemma of Homology and it allows us to define a homology group associated with each k-chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In fact, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1) implies in particular that im Bk+1 ⊂ ker Bk, so that the k-th homology group is correctly defined: Hk := ker Bk⧸im Bk+1 4 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO 1 2 3 4 5 6 7 B1 = � � � � � 1 2 � � 1 3 � � 2 3 � � 2 4 � � 3 5 � � 4 5 � � 4 6 � � 4 7 � � 5 6 � � 6 7 � [1] −1 −1 0 0 0 0 0 0 0 0 [2] 1 0 −1 −1 0 0 0 0 0 0 [3] 0 1 1 0 −1 0 0 0 0 0 [4] 0 0 0 1 0 −1 −1 −1 0 0 [5] 0 0 0 0 1 1 0 0 −1 0 [6] 0 0 0 0 0 0 1 0 1 −1 [7] 0 0 0 0 0 0 0 1 0 1 � � � � B2 = � � � � � � � � � � 1 2 3 � � 4 5 6 � � 4 6 7 � [1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2] 1 0 0 [1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3] −1 0 0 [2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3] 1 0 0 [2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 4] 0 0 0 [3,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5] 0 0 0 [4,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5] 0 1 0 [4,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 6] 0 −1 1 [4,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 7] 0 0 −1 [5,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 6] 0 1 0 [6,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 7] 0 0 1 � � � � � � � � � ∂2([1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3]) = [1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2] − [1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3] + [2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3] Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Left-hand side panel: example of simplicial complex K on 7 nodes, and of the action of ∂2 on the 2-simplex [1, 2, 3];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2-simplices included in the complex are shown in red, arrows correspond to the orientation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Panels on the right: matrix forms B1 and B2 of boundary operators ∂1 and ∂2 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The dimensionality of the k-th homology group is known as k-th Betti’s number βk = dim Hk, while the elements of Hk correspond to so-called k-dimensional holes in the simplicial complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For example, H0, H1 and H2 describe connected components, holes and three- dimensional voids respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' By standard algebraic passages one sees that Hk is isomorphic to ker � B⊤ k Bk + Bk+1B⊤ k+1 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, the number of k-dimensional holes corresponds to the dimension of the kernel of a linear operator, which is known as k-th order Laplacian or higher-order Laplacian of the simplicial complex K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Given a simplicial complex K and the boundary operators Bk and Bk+1, the k-th order Laplacian Lk of K is the |Vk| × |Vk| matrix defined as: (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) Lk = B⊤ k Bk + Bk+1B⊤ k+1 In particular, we remark that: the 0-order Laplacian is the standard combinatorial graph Laplacian L0 = B1B⊤ 1 ∈ Rn×n, whose diagonal entries consist of the degrees of the corresponding vertices (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the number of 1-simplices each vertex belongs to), while the off-diagonal (L0)ij is equal to −1 if either ij is a 1-simplex, and it is zero otherwise;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the 1-order Laplacian is known as Hodge Laplacian L1 = B⊤ 1 B1+B2B⊤ 2 ∈ Rm×m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Similarly to the 0-order case L0, one can describe the entries of L1 in terms of the structure of the simplicial complex, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [25].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Connected components and holes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The boundary operators Bk on K are directly connected with discrete notions of differential operators on the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In particular, B1, B⊤ 1 , and B⊤ 2 are the graph’s divergence, gradient, and curl operators, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We refer to [24] for more details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As Hk is isomorphic to ker Lk, we have that the following Hodge decomposition of RVk holds RVk = im Bk+1 ⊕ im B⊤ k ⊕ ker Lk = im Bk+1 ⊕ im B ⊤ k ⊕ ker Lk .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 5 Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Continuous and analogous discrete manifolds with one 1-dimensional hole (dim H1 = 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Left pane: the continuous manifold;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' center pane: the discretization with mesh vertices;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' right pane: a simplicial complex built upon the mesh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Triangles in the simplicial complex K are colored gray (right).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, the space of vertex states RV0 can be decomposed as RV0 = im B1 ⊕ ker L0, the sum of divergence-free vectors and harmonic vectors, which correspond to the connected components of the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In particular, for a connected graph, ker L0 is one-dimensional and consists of entry-wise constant vectors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, for a connected graph, im B1 is the set of vectors whose entries sum up to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Similarly, the space of flows on graph’s edges RV1 can be decomposed as RV1 = im B2 ⊕ im B⊤ 1 ⊕ ker L1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, each flow can be decomposed into its gradient part im B⊤ 1 , which consists of flows with zero cycle sum, its curl part im B2, which consists of circulations around order-2 simplicies in K, and its harmonic part ker L1, which represents 1-dimensional holes defined as global circulations modulo the curl flows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While 0-dimensional holes are easily understood as the connected components of the graph GK, a notion of “holes in the graph” GK corresponds to 1-dimensional holes in K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This terminology comes from the analogy with the continuous case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In fact, if the graph is obtained as a discretization of a continuous manifold, harmonic functions in the homology group H1 would correspond to the holes in the manifold, as illustrated in Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Moreover, the Hodge Laplacian of a simplicial complex built on N randomly sampled points in the manifold converges in the thermodynamic limit to its continuous counterpart, as N → ∞, [11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Normalized and weighted higher-order Laplacians.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In the classical graph setting, a normalized and weighted version of the Laplacian matrix is very often employed in ap- plications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' From a matrix theoretic point of view, having a weighted graph corresponds to allowing arbitrary nonnegative entries in the adjacency matrix defining the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In terms of boundary operators, this coincides with a positive diagonal rescaling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Analogously, the normalized Laplacian is defined by applying a diagonal congruence transformation to the standard Laplacian using the node weights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We briefly review these two constructions below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let G = (V, E) be a graph with positive node and edge weight functions w0 : V → R++ and w1 : E → R++, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Define the |V| × |V| and |E| × |E| diagonal matrices W0 = Diag{w0(vi)1/2}n i=1 and W1 = Diag{w1(ei)1/2}m i=1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then B1 = W −1 0 B1W1 is the normalized and weighted boundary operator of G and we have that (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3) L0 = B1B ⊤ 1 = W −1 0 B1W 2 1 B⊤ 1 W −1 0 is the normalized weighted graph Laplacian of G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In particular, note that, as for L0, the entries of L0 uniquely characterize the graph 6 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO topology, in fact we have (L0)ii = deg(i) w0(i) , (L0)ij = � � � − w1(ij) √ w0(i)w0(j) ij ∈ E 0 otherwise , for i ̸= j where deg(i) denotes the (weighted) degree of node i, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' deg(i) = � e∈E:i∈e w1(e).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While the definition of k-th order Laplacian is well-established for the case of unweighted edges and simplices, a notion of weighted and normalized k-th order Laplacian is not univer- sally available and it might depend on the application one has at hand.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For example, different definitions of weighted Hodge Laplacian are considered in [10, 22, 24, 29].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' At the same time, we notice that the notation used in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3) directly generalizes to higher orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, we propose the following notion of normalized and weighted k-th Laplacian Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let K be a simplicial complex and let wk : Vk → R++ be a positive-valued weight function on the k-simplicies of K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Define the diagonal matrix Wk = Diag � wk(σi)1/2�|Vk| i=1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, Bk = W −1 k−1BkWk is the normalized and weighted k-th boundary operator, to which cor- responds the normalized and weighted k-th Laplacian Lk = B ⊤ k Bk + Bk+1B ⊤ k+1 (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4) = WkB⊤ k W −2 k−1BkWk + W −1 k Bk+1W 2 k+1B⊤ k+1W −1 k .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Note that, from the definition Bk = W −1 k−1BkWk and (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1), we immediately have that BkBk+1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, the group Hk = ker Bk/ im Bk+1 is well defined for any choice of positive weights wk and is isomorphic to ker Lk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While the homology group may depend on the weights, we observe below that its dimension does not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Precisely, we have Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The dimension of the homology groups of K is not affected by the weights of its k-simplicies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Precisely, if Wk are positive diagonal matrices, we have (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5) dim ker Bk = dim ker Bk, dim ker B ⊤ k = dim ker B⊤ k , dim Hk = dim Hk .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Moreover, ker Bk = Wk ker Bk and ker B⊤ k = W −1 k−1 ker B ⊤ k .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Since Wk is an invertible diagonal matrix, Bkx = 0 ⇐⇒ W −1 k−1BkWkx = 0 ⇐⇒ BkWkx = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hence, if x ∈ ker Bk, then Wkx ∈ ker Bk, and, since Wk is bijective, dim ker Bk = dim ker Bk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Similarly, one observes that dim ker B ⊤ k = dim ker B⊤ k .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Moreover, since BkBk+1 = 0, then im Bk+1 ⊆ ker Bk and im B ⊤ k ⊆ ker B ⊤ k+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This yields ker Bk ∪ ker B ⊤ k+1 = RVk = ker Bk ∪ ker B⊤ k+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, for the homology group it holds: dim Hk = dim � ker Bk ∩ ker B ⊤ k+1 � = = dim ker Bk + dim ker B ⊤ k+1 − dim � ker Bk ∪ ker B ⊤ k+1 � = = dim ker Bk + dim ker B⊤ k+1 − dim � ker Bk ∪ ker B⊤ k+1 � = dim Hk QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 7 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Principal spectral inheritance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Before moving on to the next section, we recall here a relatively direct but important spectral property that connects the spectra of the k-th and (k + 1)-th order Laplacians.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 (HOL’s spectral inheritance).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let Lk and Lk+1 be higher-order Laplacians for the same simplicial complex K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let Lk = L down k + L up k , where L down k = B ⊤ k Bk and L up k = Bk+1B ⊤ k+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' σ+(L up k ) = σ+(L down k+1 ), where σ+(·) denotes the positive part of the spectrum;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' if 0 ̸= µ ∈ σ+(L up k ) = σ+(L down k+1 ), then the eigenvectors are related as follows: (a) if x is and eigenvector for L up k with the eigenvalue µ, then y = 1 √µB ⊤ k+1x is an eigenvector for L down k+1 with the same eigenvalue;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' (b) if u is and eigenvector for L down k+1 with the eigenvalue µ and u /∈ ker Bk+1, then v = 1 √µBk+1u is an eigenvector for L up k with the same eigenvalue;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' for each Laplacian Lk: if v /∈ ker L down k is the eigenvector for L down k , then v ∈ ker L up k ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' vice versa, if u /∈ ker L up k is the eigenvector for L up k , then v ∈ ker L down k ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' consequently, there exist µ ∈ σ+(Lk) with an eigenvector u ∈ ker L up k , and ν ∈ σ+(Lk+1) with an eigenvector u ∈ ker L down k+1 , such that: B ⊤ k Bkv = µv, Bk+2B ⊤ k+2u = νu .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For (2a) it is sufficient to note that L down k+1 y = B ⊤ k+1Bk+1 1 √µB ⊤ k+1x = 1 √µB ⊤ k+1L up k x = √µB ⊤ k+1x = µy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Similarly, for (2b): L up k v = Bk+1B ⊤ k+1 1 √µBk+1u = 1 √µBk+1L down k+1 u = µv;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' joint 2(a) and 2(b) yield (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hodge decomposition immediately yields the strict separation of eigenvectors between L up k and L down k , (3);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' given (3), all the inherited eigenvectors from (2a) fall into the ker L down k+1 , thus resulting into (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In other words, the variation of the spectrum of the k-th Laplacian when moving from one order to the next one works as follows: the down-term L down k+1 inherits the positive part of the spectrum from the up-term of L up k ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the eigenvectors corresponding to the inherited positive part of the spectrum lie in the kernel of L up k+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' at the same time, the “new” up-term L up k+1 has a new, non-inherited, part of the positive spectrum (which, in turn, lies in the kernel of the (k + 2)-th down-term).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In particular, we notice that for k = 0, since B0 = 0 and L0 = L up 0 , the theorem yields σ+(L0) = σ+(L1 down) ⊆ σ+(L1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In other terms, the positive spectrum of the L0 is inher- ited by the spectrum of L1 and the remaining (non-inherited) part of σ+(L1) coincides with σ+(L up 1 ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Figure 3 provides an illustration of the statement of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 for k = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Problem setting: Nearest complex with different homology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Suppose we are given a simplicial complex K on the vertex set V, with simplex weight functions w0, w1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' , and let βk = dim Hk = dim Hk the dimension of its k-homology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We aim at finding the closest simplex on the same vertex set V, with a strictly larger number of holes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As it is the most frequent higher-order Laplacian appearing in applications and since this provides already a 8 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO 0 0 · · · 0 λ1 λ2 λ3 λ4 λ5 λ6 λ7 λ8 λ9 λ10 ← σ(L1) 0 0 · · · 0 λ1 λ2 0 λ4 0 0 λ7 λ8 0 λ10 ← σ(B T 1 B1) 0 0 · · · 0 0 0 λ3 0 λ5 λ6 0 0 λ9 0 ← σ(B2B T 2 ) holes µ Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Illustration for the principal spectrum inheritance (Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7) in case k = 0: spectra of L1, L down 1 and L up 1 are shown.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Colors signify the splitting of the spectrum, λi > 0 ∈ σ(L1) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' all yellow eigenvalues are inherited from σ+(L0);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' red eigenvalues belong to the non-inherited part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Dashed barrier µ signifies the penalization threshold (see the target functional in Section 4) preventing homological pollution (see Subsection 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' large number of numerical complications, we focus here only on the Hodge Laplacian case: given the simplicial complex K = (V0, V1, V2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ), we look for the smallest perturbation of the edges V1 which increases the dimension of H1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' More precisely: Problem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let K be a simplicial complex of order at least 2 with edge weights w1 and corresponding diagonal weight matrix W1, and let β1(W1) be the dimension of the homology group corresponding to the weights W1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For ε > 0, let Ω(ε) = � diagonal matrices W such that ∥W∥ = ε � , Π(W1) = � diagonal matrices W such that W1 + W ≥ 0 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In other words, Ω(ε) is an ε-sphere and Π(W1) allows only non-negative simplex weights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We look for the smallest perturbation ε such that there exists a weight modification δW1 ∈ Ω(ε) ∩ Π(W1) such that β1(W1) < β1(W1 + δW1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Here, and throughout the paper, ∥X∥ denotes either the Frobenius norm if X is a matrix, or the Euclidean norm if X is a vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Note that, as we are looking for the smallest ε, the equality ∥W∥ = ε is an obvious choice, as opposed to ∥W∥ ≤ ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As the dimension β1 coincides with the kernel of L1, we approach this problem through the minimization of a functional based on the spectrum of the 0-th and 1-st Laplacian of the simplicial complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to define such functional, we first make a number of considera- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Note that, due to Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6, the dimension of the first homology group does not change when the edge weights are perturbed, as long as all the weights remain positive.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, in order to find the desired perturbation δW1, we need to set some of the initial weights to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This operation creates several potential issues we need to carefully address, as discussed next.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' First, setting an edge to zero implies that one is formally removing that edge from the simplicial complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As the simplicial complex structure needs to be maintained, when doing so we need to set to zero also the weight of any 2-simplex that contains that edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For this reason, if �w1(e) = w1(e) + δw1(e) is the new edge weight function, we require the weight function of the 2-simplices to change into �w2, defined as �w2(i1i2i3) = f �δw1(i1i2) w1(i1i2) , δw1(i2i3) w1(i2i3) , δw1(i1i3) w1(i1i3) � w2(i1i2i3) QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 9 where f(u1, u2, u3) is a function such that f(0, 0, 0) = 1 and that monotonically decreases to zero as ui → −1, for any i = 1, 2, 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' An example of such f is f(u1, u2, u3) = 1 − min{u1, u2, u3} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1) Second, when setting the weight of an edge to zero we may disconnect the underlying graph and create an all-zero column and row in the Hodge Laplacian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This gives rise to the phenomenon that we call “homological pollution”, which we will discuss in detail in the next subsection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Homological pollution: inherited almost disconnectedness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As the Hodge homol- ogy β1 corresponds to the number of zero eigenvalues in ker L1, the intuition suggests that if L1 has some eigenvalue that is close to zero, then the simplicial complex is “close to” having at least one more 1-dimensional hole.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' There are a number of problems with this intuitive consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' By Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 for k = 0, σ+(L1) inherits σ+(L0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hence, if the weights in W1 vary continuously so that a positive eigenvalue in σ+(L0) approaches 0, the same happens to σ+(L1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Assuming the initial graph GK is connected, an eigenvalue that approaches zero in σ(L0) would imply that GK is approaching disconnectedness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This leads to a sort of pollution of the kernel of L1 as an almost-zero eigenvalue which corresponds to an “almost” 0-dimensional hole (disconnected component) from L0 is inherited into the spectrum of L1, but this small eigenvalue of L1 does not correspond to the creation of a new 1-dimensional hole in the reduced complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To better explain the problem of homological pollution, let us consider the following illustrative example.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Example 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Consider the simplicial complex of order 2 depicted in Figure 4a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In this ex- ample we have V0 = {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' , 7}, V1 = {[1, 2], [1, 3], [2, 3], [2, 4], [3, 5], [4, 5], [4, 6], [5, 6], [5, 7], [6, 7]} and V2 = {[1, 2, 3], [4, 5, 6], [5, 6, 7]}, all with weight equal to one: wk ≡ 1 for k = 0, 1, 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The only existing 1-dimensional hole is shown in red and thus the corresponding Hodge homology is β = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Now, consider perturbing the weight of edges [2, 4] and [3, 5] by setting their weights to ε > 0 Figure 4b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For small ε, this perturbation implies that the smallest nonzero eigenvalue µ2 in σ+(L0) is scaled by ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As σ+(L0) ⊆ σ+(L1), we have that dim ker L1 = 1 and σ+(L1) has an arbitrary small eigenvalue, approaching 0 with ε → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' At the same time, when ε → 0, the reduced complex obtained by removing the zero edges as in Figure 4c does not have any 1-dimensional hole, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' β1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, in this case, the presence of a very small eigenvalue µ2 ∈ σ+(L1) does not imply that the simplicial complex is close to a simplicial complex with a larger Hodge homology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To mitigate the phenomenon of homological pollution, in our spectral-based functional for Problem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 we include a term that penalizes the spectrum of L0 from approaching zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To this end, we observe below that a careful choice of the vertex weights is required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The smallest non-zero eigenvalue of the Laplacian µ2 ∈ σ(L0) is directly related to the connectedness of the graph GK.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This relation is well-known and dates back to the pioneering work of Fiedler [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In particular, as µ2 is a function of node and edge weights, the following 10 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO 1 1 1 1 1 1 1 1 1 1 1 2 3 4 5 6 7 (a) Connected 1 1 1 ε ε 1 1 1 1 1 1 2 3 4 5 6 7 (b) Close to discon- nected 1 1 1 1 1 1 1 1 1 2 3 4 5 6 7 (c) Disconnected Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Example of the homological pollution, Example 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2, for the simplicial complex K on 7 vertices;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the existing hole is shown in red (left and center pane), all 3 cliques are included in the simplicial complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The left pane demonstrates the initial setup with 1 hole;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the center pane retains the hole exhibiting spectral pollution;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the continuous transition to the eliminated edges with β1 = 0 (no holes) is shown on the right pane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' generalized version of the Cheeger inequality holds (see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [34]) (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) 1 2µ2 ≤ h(GK) ≤ � 2 µ2 max i∈V0 deg(i) w0(i) �1/2 , where h(GK) = min S⊂V0 w1(S, V0\\S) min{w0(S), w0(V0\\S)} , with w1(S, V0\\S) = � ij∈V1:i∈S,j /∈S w1(ij), deg(i) = � j:ij∈V1 wi(ij), w0(S) = � i∈S w0(i).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We immediately see from (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) that when the graph GK is disconnected, then h(GK) = 0 and µ2 = 0 as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Vice-versa, if µ2 goes to zero, then h(GK) decreases to zero too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While this happens independently of the choice of w0 and w1, if w0 is a function of w1 then it might fail to capture the presence of edges whose weight is decreasing and is about to disconnect the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To see this, consider the example choice w0(i) = deg(i), the degree of node i in GK.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Note that this is a very common choice in the graph literature, with several useful properties, including the fact that no other graph-dependent constant appears in the Cheeger inequality (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) other than µ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For this weight choice, consider the case of a leaf node, a node i ∈ V0 that has only one edge ij0 ∈ V1 connecting i to the rest of the (connected) graph GK via the node j0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' If we set w1(ij0) = ε and we let ε decrease to zero, the graph GK is approaching disconnectedness and we would expect h(GK) and µ2 to decrease as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' However, one easily verifies that both µ2 and h(GK) are constant with respect to ε in this case, as long as ε ̸= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to avoid such a discontinuity, in our weight perturbation strategy for the simplex K, if w0 is a function of w1, we perturb it by a constant shift.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Precisely, if w0 is the initial vertex weight of K, we set �w0(i) = w0(i) + ϱ, with ϱ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' So, for example, if w0 = deg and the new edge weight function �w1(e) = w1(e) + δw1(e) is formed after the addition of δW1, we set �w0(i) = � j [w1(ij) + δw1(ij)] + ϱ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 11 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Spectral functional for 1-st homological stability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We are now in the position to formulate our proposed spectral-based functional, whose minimization leads to the desired small perturbation that changes the first homology of K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In the notation of Problem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1, we are interested in the smallest perturbation ε and the corresponding modification δW1 ∈ Ω(ε) ∩ Π(W1) that increases β1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As ∥δW1∥ = ε, for convenience we indicate δW1 = εE with ∥E∥ = 1 so E ∈ Ω(1)∩Πε(W1), where Πε(W1) = {W | εW ∈ Π(W1)}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For the sake of simplicity, we will omit the dependencies and write Ω and Πε, when there is no danger of ambiguity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Finally, let us denote by β1(ε, E) the first Betti number corresponding to the simplicial complex perturbed via the edge modification εE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' With this notation, we can reformulate Problem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 as follows: Problem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Find the smallest ε > 0, such that there exists an admissible perturbation E ∈ Ω ∩ Πε with an increased number of holes, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' min � ε > 0 : β1(ε, E) ≥ β1 + 1 for some E ∈ Ω ∩ Πε � (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1) where β1 = β1(0, ·) is the first Betti number of the original simplicial complex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to approach Problem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1, we introduce a target functional F(ε, E), based on the spectrum of the 1-Laplacian L1(ε, E) and the 0-Laplacian L0(ε, E), where the dependence on ε and E is to emphasize the corresponding weight perturbation is of the form W1 �→ W1 +εE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Our goal is to move a positive entry in σ+(L1(ε, E)) into the kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' At the same time, assuming the initial graph GK is connected, one should avoid the inherited almost discon- nectedness with small positive entries of σ+(L0(ε, E)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As, by Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 for k = 0, σ+(L0(ε, E)) = σ+(L down 1 (ε, E)), the only eigenvalue of L1(ε, E) that can be continuously driven to 0 comes from L up 1 (ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For this reason, let us denote the first non-zero eigenvalue of the up-Laplacian L up 1 (ε, E) by λ+(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The proposed target functional F(ε, E) is defined as: F(ε, E) = λ+(ε, E)2 2 + α 2 max � 0, 1 − µ2(ε, E) µ �2 (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) where α and µ are positive parameters, and µ2(ε, E) is the first nonzero eigenvalue of L0(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As recalled in Subsection 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1, µ2(ε, E) is an algebraic measure of the connectedness of the perturbed graph, thus the whole second term in (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) “activates” when such algebraic con- nectedness falls below the threshold µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' By design, F(ε, E) is non-negative and is equal to 0 iff λ+(ε, E) reaches 0, increasing the dimension of H1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Using this functional, we recast the Problem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 as min {ε > 0 : F(ε, E) = 0 for some E ∈ Ωε} (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3) Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' When GK is connected, dim ker L0 = 1 and by Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 dim ker L up 1 = dim ker L1+(n−dim ker L0) = n+β1−1, so the first nonzero eigenvalue of L up 1 is the (n+β1)- th.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While (n + β1) can be a large number in practice, we will discuss in Subsection 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 an efficient method that allows us to compute λ+(ε, E) without computing any of the previous (n + β1 − 1) eigenvalues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 12 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A two-level optimization procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We propose to approach (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3) by means of a two- level iterative method, which is based on the successive minimization of the target functional F(ε, E) and a subsequent tuning of the parameter ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' More precisely, we propose the following two-level scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A similar procedure was used in the context of graph spectral nearness in [2] and in other matrix nearness problems [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Inner level: for fixed ε > 0, solve the minimization problem E(ε) = arg min E∈Ω∩Πε F(ε, E) by a constrained gradient flow which we formulate below, where we denote the computed minimizer by E(ε).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Outer level: given the function ε �→ E(ε), we consider the optimization problem: (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1) F(ε, E(ε)) = 0 and look for the smallest value ε∗ > 0 that solves (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Details on the implementation of both levels are given below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' By construction, the resulting algorithm converges to a minimum of F(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Although global convergence to the global optimum is not guaranteed, in our experiments we always observe the method reaches the expected global solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' However, we point out that there exists one pathological configuration of the simplicial complex K which would prevent global optimality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This happens when there are several holes and simplices from V2(K) being adjacent to each other only through a common edge, which has a small weight.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In that case, the algorithm would always eliminate that common edge, whilst the correctness of this answer would depend on the balance between the number of holes and adjacent simplices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' These configurations are rarely present in real-life systems (we did not encounter them in any of our tests);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' nevertheless, one can tackle such unsuccessful runs through manual weight modification of the undesired edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Inner Level Iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Here we consider the minimization problem with respect to E and fixed scalar parameters: the perturbation norm ε is inherited from the outer level, and the connectedness parameters α and µ are assumed to be given (we will discuss the choice of µ and α later).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We solve the resulting minimization problem min{F(ε, E) : E ∈ Ω ∩ Πε} by solving the associated constrained gradient system E (t) = −PΩ∩ΠεG(ε, E(t)) (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) where G(ε, E) = ∇EFk(ε, E) and PΩ∩Πε is a projector onto the admissible set Ω∩Πε (where ε is fixed).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Since the system integrates the anti-gradient, a minimizer (at least local) of F(ε, E) is obtained at t → ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Equation (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) introduces a dummy time dependence outlining the difference between the discrete gradient descent and the gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In the current work, we use the latter which benefits from a variety of known integrators and simpler implementation of constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We devote the next two Subsection 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 and Subsection 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 to computing the projected gradient in (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The idea is to express the derivative of F in terms of the derivative of the QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 13 perturbation E , and to identify the constrained gradient of the functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To this end, we first compute the free gradient and then we discuss how to deal with the projection onto the admissible set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, in Subsection 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 we will discuss the free gradient transition phase that characterizes the outer iteration level for (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The free gradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We compute here the free gradient of F with respect to E, given a fixed ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to proceed, we need a few preliminary results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The following is a standard perturbation result for eigenvalues;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [23], where we denote by ⟨X, Y ⟩ = � i,j xijyij = Tr(X⊤Y ) the inner product in Rn×n that induces the Frobenius norm ∥X∥ = ⟨X, X⟩1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 (Derivative of simple eigenvalues).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Consider a continuously differentiable path of square symmetric matrices A(t) for t in an open interval I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let λ(t), t ∈ I, be a continuous path of simple eigenvalues of A(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let x(t) be the eigenvector associated to the eigenvalue λ(t) and assume ∥x(t)∥ = 1 for all t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then λ is continuously differentiable on I with the derivative (denoted by a dot) (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3) λ = x⊤A x = ⟨xx⊤, A ⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Moreover, “continuously differentiable” can be replaced with “analytic” in the assumption and the conclusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let us denote the perturbed weight matrix by � W1(t) = W1 +εE(t), and the corresponding � W0(t) = W0(� W1(t)) and � W2(t) = W2(� W1(t)), defined accordingly as discussed in Section 3;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' we omit the time dependence for the perturbed matrices to simplify the notation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Since � W0, � W1 and � W2 are necessarily diagonal, by the chain rule we have � W i(t) = ε diag � Ji 1E 1 � , where 1 is the vector of all ones, diag(v) is the diagonal matrix with diagonal entries the vector v, and Ji 1 is the Jacobian matrix of the i-th weight matrix with respect to � W1, which for any u1 ∈ V1 and u2 ∈ Vi, has entries [Ji 1]u1,u2 = ∂ ∂ �w1(u1) �wi(u2) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Next, in the following two lemmas, we express the time derivative of the Laplacian L0 and L up 1 as functions of E(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The proofs of these results are straightforward and omitted for brevity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In what follows, Sym[A] denotes the symmetric part of the matrix A, namely Sym[A] = (A + A⊤)/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 (Derivative of L0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For the simplicial complex K with the initial edges’ weight matrix W1 and fixed perturbation norm ε, let E(t) be a smooth path and � W0, � W1, � W2 be cor- responding perturbed weight matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, 1 2ε d dtL0(t) = � W −1 0 B1� W1E B⊤ 1 � W −1 0 − Sym � � W −1 0 diag � J0 1E 1 � L0 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4) 14 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 (Derivative of L up 1 ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For the simplicial complex K with the initial edges’ weight matrix W1 and fixed perturbation norm ε, let E(t) be a smooth path and � W0, � W1, � W2 be cor- responding perturbed weight matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, 1 2ε d dtL up 1 (t) = − Sym � � W −1 1 B2� W 2 2 B⊤ 2 � W −1 1 E � W −1 1 � + (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5) + � W −1 1 B2� W2 diag � J0 1E 1 � B⊤ 2 � W −1 1 (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6) Combining Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 with Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 and Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 we obtain the following expres- sion for the gradient of the functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 (The free gradient of F(ε, E)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Assume the initial weight matrices W0, W1 and W2, as well as the parameters ε > 0, α > 0 and µ > 0, are given.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Additionally assume that E(t) is a differentiable matrix-valued function such that the first non-zero eigenvalue λ+(ε, E) of L up 1 (ε, E) and the second smallest eigenvalue µ2(ε, E) of L0(ε, E) are simple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let � W0, � W1, � W2 be corresponding perturbed weight matrices;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' then: 1 ε∇EF(ε,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E)(t) = λ+(ε,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E) � Sym � −� W −1 1 B2� W 2 2 B⊤ 2 � W −1 1 x+x⊤ +� W −1 1 � + diag � J2 1 ⊤ diagvec � B⊤ 2 � W −1 1 x+x⊤ +� W −1 1 B2� W2 �� � − − α µ max � 0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1 − µ2(ε,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E) µ � � B⊤ 1 � W −1 0 y2y⊤ 2 � W −1 0 B1� W1− − diag � J0 1 ⊤ diagvec � Sym[� W −1 0 y2y⊤ 2 L0] �� � where x+ is a unit eigenvector of L up 1 corresponding to λ+,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' y2 is a unit eigenvector of L0 corresponding to µ2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' and the operator diagvec(X) returns the main diagonal of X as a vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To derive the expression for the gradient ∇EF, we exploit the chain rule for the time derivative: λ = ⟨ d dtA(E(t)), xx⊤⟩ = ⟨∇Eλ, E ⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then it is sufficient to apply the cyclic perturbation for the scalar products of Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 and Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 with x+x⊤ + and y2y⊤ 2 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The final transition requires the formula: ⟨A, diag(BE1))⟩ = � diag � B⊤(diagvec A) � , E � Remark 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The derivation above assumes the simplicity of both µ2(ε, E) and λ+(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This assumption is not too restrictive as simplicity for these extremal eigenvalues is a generic property.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We observe simplicity in all our numerical tests.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The constrained gradient system and its stationary points.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In this section we are deriving from the free gradient determined in Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 the constrained gradient of the considered functional, that is the projected gradient (with respect to the Frobenius inner product) onto the manifold Ω∩Πε, composed of perturbations E which preserve the structure of W and the unit norm constraint of E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 15 In order to obtain the constrained gradient system, we need to project the unconstrained gradient given by Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 onto the feasible set and also to normalize E to preserve its unit norm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' On a time interval where the set of 0-weight edges remains unchanged, the norm- unconstrained gradient is given by (a formal derivation can be established through the KKT conditions): P+G(ε, E) (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7) where P+ is the non-negativity projection such that [P+X]ij = � Xij, [W1 + εE]ij > 0 0, otherwise .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Further, in order to comply with the constraint ∥E(t)∥2 = 1, we must have (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='8) 0 = 1 2 d dt∥E(t)∥2 = ⟨E(t), E (t)⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We are thus led to the following constrained optimization problem for the admissible direction of the steepest descent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6 (Direction of steepest admissible descent).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let E, G ∈ Rn,n with G given by (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7) and ∥E∥F = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' On a time interval where the set of 0-weight edges remains unchanged, the gradient system reads E (t) = −P+G(ε, E(t)) + κP+E(t), where κ = ⟨ε, G(E(t)), P+E(t)⟩ ∥P+E(t)∥2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We need to orthogonalize E (t) with respect to E(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As usual, this can be obtained by the introduction of a linear orthogonality correction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The gradient system reads (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10) E = P+(−G − κE), where κ is determined from the constraint ⟨E, E ⟩ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We then have 0 = ⟨E, E ⟩ = ⟨E, P+(−G − κE)⟩ = −⟨P+E, G⟩ − κ⟨P+E, P+E⟩, and the result follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Equation (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9) suggests that the systems goes “primarily” in the direction of the antigra- dient −G(E, ε), thus the functional is expected to decrease along it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 (Monotonicity).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let E(t) of unit Frobenius norm satisfy the differential equa- tion (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10) with G given by (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, the functional Fk(ε, E)(t) decreases monotonically with t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 16 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We consider first the simpler case where the non-negativity projection does nt apply so that G = G(E, ε) (without P+).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then d dtF(ε, E)(t) = � ∇EFk(ε, E), E � = ⟨εG(ε, E(t)), −G(ε, E(t)) + κE(t)⟩ = = −ε∥G(ε, E)∥2 + ε⟨G(ε, E), E⟩ ⟨E, E⟩ ⟨G(ε, E), E⟩ = = ε � −∥G(ε, E)∥2 + |⟨G(ε, E), E⟩|2 ∥E∥2 � ≤ 0 (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11) where the final estimate is given by the Cauchy-Bunyakovsky-Schwarz inequality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The derived inequality holds on the time interval without the change in the support of P+ (so that no new edges are prohibited by the non-negativity projection).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As a direct consequence, we observe that the stationary points of the differential equation (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10) are characterized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='8 (Stationary points).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let E⋆ be an admissible perturbation with ∥E⋆∥F = 1 be such that (i) The eigenvalue λ+(ε, E) is simple at E = E⋆ and depends continuously on E in a neigh- borhood of E⋆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' (ii) Penalization is not active, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' µ2(ε, E) > µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' (iii) The gradient G(ε, E⋆) is nonzero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Let E(t) be the solution of (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10) passing through E⋆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then the following are equivalent: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' d dtF (ε, E(t)) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E⋆ is a real multiple of G(ε, E⋆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' It is immediate to see that 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' implies 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=', which implies 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The proof is concluded noting that (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11) shows that 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' implies 3 by the strict form of Cauchy-Bunyakovsky-Schwarz inequality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Free Gradient Transition in the Outer Level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The optimization problem in the inner level is non-convex due to the non-convexity of the functional F(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hence, for a given ε, the computed minimizer E(ε) may depend on the initial guess E0 = E0(ε).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The effects of the initial choice are particularly important upon the transition �ε → ε = �ε + ∆ε between constrained inner levels: given reasonably small ∆ε, one should expect rela- tively close optimizers E(�ε) and E(ε), and, hence, the initial guess E0(ε) being close to and dependent on E(ε).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This choice, which seems very natural, determines however a discontinuity F(�ε, E(�ε)) ̸= F(ε, E(�ε)), which may prevent the expected monotonicity property with respect to ε in the (likely unusual case) where F(�ε, E(�ε)) < F(ε, E(�ε)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This may happen in particular when ∆ε is not taken small;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' since the choice of ∆ε is driven by a Newton-like iteration we are interested in finding a way to prevent this situation and making the whole iterative method more robust.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The goal QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 17 E0(ε0) E(ε0) constrain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E(t) E0(ε1) E(ε1) constrain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E(t) free flow, �E(t) ε0 → ε0 + ∆ε0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E0(εv) E(εv) constrain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E(t) F(εv, E(εv)) = 0 Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The scheme of alternating constrained (blue) and free gradient (red) flows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Each stage inherits the final iteration of the previous stage as initial E0(εi) or �E0(εi) respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The scheme alternates until the target functional vanishes (F(ε, E) = 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' is that of guaranteeing monotonicity of the functional both with respect to time and with respect to ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' When in the outer iteration we increase ε from a previous value �ε < ε, we have the problem of choosing a suitable initial value for the constrained gradient system (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9), such that at the stationary point E(�ε) we have F(�ε, E(�ε)) < F(ε, E(ε)) (which we assume both positive, that is on the left of the value ε⋆ which identifies the closest zero of the functional).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to maintain monotonicity with respect to time and also with respect to ε, it is convenient to start to look at the optimization problem with value ε, with the initial datum δW1 = �εE(�ε) of norm �ε < ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This means we have essentially to optimize with respect to the inequality constraint ∥δW1∥F ≤ ε, or equivalently solve the problem (now with inequality constrain on ∥E∥F ): E(ε) = arg min E∈Ω,∥E∥F ≤1 F(ε, E) The situation changes only slightly from the one discussed above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' If ∥E∥F < 1, every direction is admissible, and the direction of the steepest descent is given by the negative gradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' So we choose the free gradient flow (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12) E = −P+G(ε, E(t)) as long as ∥E(t)∥F < 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' When ∥E(t)∥F = 1, then there are two possible cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' If ⟨P+G(ε, E), E⟩ ≥ 0, then the solution of (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12) has d dt∥E(t)∥2 F = 2 ⟨E , E⟩ = −2 ⟨P+G(ε, E(t)), E⟩ ≤ 0, and hence the solution of (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12) remains of Frobenius norm at most 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Otherwise, if ⟨P+G(ε, E), E⟩ < 0, the admissible direction of steepest descent is given by the right-hand side of (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9), i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' −P+G(ε, E) + κE, κ = ⟨G(ε,E), P+E⟩ ∥P+E∥2 and so we choose that differential equation to evolve E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The situation can be summarized as taking, if ∥E(t)∥F = 1, (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13) E = −P+G(ε, E) + µE with µ = min � 0, κ � with κ = ⟨G(ε,E), P+E⟩ ∥P+E∥2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Along the solutions of (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13), the functional F decays monotoni- cally, and stationary points of (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13) with P+G(ε, E(t)) ̸= 0 are characterized, by the same 18 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO Algorithm 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 Pseudo-code of the complete constrained- and free-gradient flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Require: initial edge perturbation guess E0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' initial ε0 > 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' ε-stepsize ∆ε > 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' bounds α∗, α∗ for the α-phase;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1: α, E ← AlphaPhase(E0, ε0, α∗, α∗) ▷ for details see Section 7 2: while |F(ε, E)| < 10−6 do 3: ε ← ε + ∆ε 4: E ← ε ε+∆εE ▷ before the free gradient ∥E∥ < 1 5: Ei ← FreeGradientFlow(E, ∆ε, ε) ▷ see Subsection 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 6: E ← ConstrainedGradientFlow(E, ε) ▷ see Section 6 7: end while arguments used before, as (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='14) E is a negative real multiple of P+G(ε, E(t)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' If it can be excluded that the gradient P+G(ε, E(t)) vanishes at an optimizer, it can thus be concluded that the optimizer of the problem with inequality constraints is a stationary point of the gradient flow (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9) for the problem with equality constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Remark 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As a result, F(ε, E(t)) monotonically decreases both with respect to time t and to the value of the norm ε, when ε ≤ ε⋆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Algorithm details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In Algorithm 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 we provide the pseudo-code of the whole bi-level procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The initial “α-phase” is used to choose an appropriate value for the regularization parameter α.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In order to avoid the case in which the penalizing term on the right-hand side of (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2) dominates the loss F(ε, E(t)) in the early stages of the descent flow, we select α by first running such an initial phase, prior to the main alternated constrained/free gradient loop.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In this phase, we fix a small ε = ε0 and run the constrained gradient integration for an initial α = α∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' After the computation of a local optimum E∗, we then increase α and rerun for the same ε0 with E∗ as starting point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We iterate until no change in E∗ is observed or until α reaches an upper bound α∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The resulting value of α and E∗ are then used in the main loop where we first increase ε by the chosen step size, we rescale Ei by 0 < ε/(ε + ∆ε) < 1, and then we perform the free gradient integration described in Subsection 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 until we reach a new point Ei on the unit sphere ∥Ei∥ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Then, we perform the inner constrained gradient step by integrating Equation (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9), iterating the following two-step norm-corrected Euler scheme: � Ei+1/2 = Ei − hi (P+G(Ei, ε) − κiP+Ei) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Ei+1 = PΠεEi+1/2/∥PΠεEi+1/2∥ (6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1) where the second step is necessary to numerically guarantee the Euler integration remains in the set of admissible flows since the discretization does not conserve the norm and larger steps hi may violate the non-negativity of the weights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In both the free and constrained integration phases, since we aim to obtain the solution at t → ∞ instead of the exact trajectory, we favor larger steps hi given that the established QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 19 monotonicity is conserved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Specifically, if F(ε, Ei+1) < F(ε, Ei), then the step is accepted and we set hi+1 = βhi with β > 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' otherwise, the step is rejected and repeated with a smaller step hi ← hi/β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The step acceleration strategy described above, where βhi is immediately increased after one accepted step, may lead to “oscillations” between accepted and rejected steps in the event the method would prefer to maintain the current step size hi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To avoid this potential issue, in our experiments we actually increase the step length after two consecutive accepted steps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Alternative step-length selection strategies are also possible, for example, based on Armijo’s rule or non-monotone stabilization techniques [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' When the weight of an edge e is moved to zero, we are formally reducing the initial complex K to a smaller �K with V1( �K) = � V1 = V1\\{e}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While the Hodge Laplacian of the new �K should have a smaller dimension than the initial one, in our perturbative approach, we want to maintain the dimension of L1 unchanged so as to be able to explore the set of possible perturbations Ω(ε) ∩ Π(W1) in a continuous way.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' While this may create a complication in the general unweighted setting, as it might give rise to an almost-zero row in B1 and thus a degenerate almost-zero entry in σ(Lup 1 ) which does not correspond to a different homology, we emphasize that in our weighted set-up, this sort of faux edges are automatically ignored.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In fact, in our model, the weights of the 2-simplicies W2 evolve as a function of the edge weights W1 as in (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, when an entry of W1 approaches zero, the corresponding entries of W2 approach 0 with the same order, and the weight normalization in the definition of B1 prevents the formation of zero rows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Computational costs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Each step of either the free or the constrained flows requires one step of explicit Euler integration along the anti-gradient −∇EF(ε, E(t)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As discussed in Section 5, the construction of such a gradient requires several sparse and diagonal matrix- vector multiplications as well as the computation of the smallest nonzero eigenvalue of both L up 1 (ε, E) and L0(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The latter two represent the major computational requirements of the numerical procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Fortunately, as both matrices are of the form A⊤A, with A being either of the two boundary or co-boundary operators B2 and B ⊤ 1 , we have that both the two eigenvalue problems boil down to a problem of the form min x⊥ ker A ∥Ax∥ ∥x∥ i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=', the computation of the smallest singular value of the sparse matrix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This problem can be approached by a sparse singular value solver based on a Krylov subspace scheme for the pseudo inverse of A⊤A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In practice, we implement the pseudo inversion by solving the corresponding least squares problems min x ∥L up 1 (ε, E)x − b∥, min x ∥L0(ε, E)x − b∥ , which, in our experiments, we solved using the least square minimal-residual method (LSMR) from [15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' This approach allows us to use a preconditioner for the normal equation correspond- ing to the least square problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For simplicity, in our tests we used a constant preconditioner computed by means of an incomplete Cholesky factorization of the initial unperturbed L up 1 , or 20 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO L0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Possibly, much better performance can be achieved with a tailored preconditioner that is efficiently updated throughout the matrix flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We also note that the eigenvalue problem for the graph Laplacian L0(ε, E) may be alternatively approached by a combinatorial multigrid strategy [30].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' However, designing a suitable preconditioning strategy goes beyond the scope of this work and will be the subject of future investigations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Numerical experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In this section, we provide several synthetic and real-world example applications of the proposed stability algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The code for all the examples is available at https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='com/COMPiLELab/HOLaGraF.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' All experiments are run until the global stopping criterion |F(ε, E(t))| < 10−6 is met and the parameters µ and α are chosen as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Concerning µ, since the effect of weight perturbations on µ2(ε, E) diminishes with the growth of the network, F(ε, E) becomes less sensitive to it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Thus, in the computations we set µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='75µ2, where µ2 is the smallest nonzero eigenvalue of the initial Laplacian L0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As for α, we run the α-phase described in Section 6 with parameters ε0 = 10−3, α∗ = 1 and α∗ = 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Illustrative Example.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We consider here a small example of a simplicial complex K of order 2 consisting of eight 0-simplicies (vertices), twelve 1-simplicies (edges), four 2-simplicies V2 = {[1, 2, 3], [1, 2, 8], [4, 5, 6], [5, 6, 7]} and one corresponding hole [2, 3, 4, 5], hence, β1 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' By design, the dimensionality of the homology group H1 can be increased only by eliminating edges [1, 2] or [5, 6];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' for the chosen weight profile w1([1, 2]) > w1([5, 6]), hence, the method should converge to the minimal perturbation norm ε = w1([5, 6]) by eliminating the edge [5, 6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The exemplary run of the optimization framework in time is shown on Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The top panel of Figure 6 provides the continued flow of the target functional F(ε, E(t)) consisting of the initial α-phase (in green) and alternated constrained (in blue) and free gradient (in orange) stages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As stated above, F(ε, E(t)) is strictly monotonic along the flow since the support of P+ does not change.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Since the initial setup is not pathological with respect to the connectivity, the initial α-phase essentially reduces to a single constrained gradient flow and terminates after one run with α = α∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The constrained gradient stages are characterized by a slow changing E(t), which is essentially due to the flow performing small adjustments to find the correct rotation on the unit sphere, whereas the free gradient stage quickly decreases the target functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The second panel shows the behaviour of first non-zero eigenvalue λ+(ε, E(t)) (solid line) of L up 1 (ε, E(t)) dropping through the ranks of σ(L1(ε, E(t))) (semi-transparent);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' similar to the case of the target functional F(ε, E(t)), λ+(ε, E(t)) monotonically decreases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The rest of the eigenvalues exhibit only minor changes, and the rapidly changing λ+ successfully passes through the connectivity threshold µ (dotted line).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The third and the fourth panels show the evolution of the norm of the perturbation ∥E(t)∥ and the perturbation E(t) itself, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The norm ∥E(t)∥ is conserved during the constrained-gradient and the α- stages;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' these stages correspond to the optimization of the perturbation shape, as shown by the small positive values at the beginning of the bottom panel which eventually vanish.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' During the free gradient integration the norm ∥E(t)∥ increases, but the relative change of the norm declines with the growth of εi to avoid jumping over the smallest possible ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Finally, due to the simplicity of the complex, the edge we want to eliminate, 56, dominates the flow from the very beginning (see bottom panel);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' such a clear QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 21 0 100 200 300 400 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='000 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='002 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='004 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='006 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='008 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='010 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='012 functional F(ε, E(t)) α−stage free gradient constrained gradient 0 100 200 300 400 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 spectrum σ(L1) target λ+ other λi threshold µ 0 100 200 300 400 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='00 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='25 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='50 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='75 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='00 norm ∥E(t)∥ 0 100 200 300 400 [1 2] [1 3] [1 8] [2 3] [2 4] [2 8] [3 5] [4 5] [4 6] [5 6] [5 7] [6 7] perturbation E(t) −1 −0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 1 1 2 3 4 5 6 7 8 ε = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='025 1 2 3 4 5 6 7 8 ε = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='075 1 2 3 4 5 6 7 8 ε = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='125 1 2 3 4 5 6 7 8 ε = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='31 Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Illustrative run of the framework determining the topological stability: the top pane — the flow of the functional F(ε, E(t));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the second pane — the flow of σ(L1), λ+ is highlighted;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' third pane — the change of the perturbation norm ∥E(t)∥;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the bottom pane — the heatmap of the perturbation profile E(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' pattern persists only in small examples, whereas for large networks the perturbation profile is initially spread out among all the edges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Triangulation Benchmark.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' To provide more insight into the computational behavior of the method, we synthesize here an almost planar graph dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Namely, we assume N uniformly sampled vertices on the unit square with a network built by the Delaunay triangula- tion;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' then, edges are randomly added or erased to obtain the sparsity ν (so that the graph has 1 2νN(N − 1) edges overall).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' An order-2 simplicial complex K = (V0, V1, V2) is then formed by letting V0 be the generated vertices, V1 the edges, and V2 every 3-clique of the graph;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' edges’ weights are sampled uniformly between 1/4 and 3/4, namely w1(ei) ∼ U[ 1 4, 3 4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' An example of such triangulation is shown in Figure 7a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' here N = 8 and edges [6, 8] and [2, 7] were eliminated to achieve the desired sparsity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We sample networks with a varying number of vertices N = 10, 16, 22, 28 and vary- ing sparsity pattern ν = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 which determine the number of edges in the output as m = ν N(N−1) 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Due to the highly randomized procedure, topological structures of a sam- pled graph with a fixed pair of parameters may differ substantially, so 10 networks with the same (N, ν) pair are generated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For each network, the working time (without considering the sampling itself) and the resulted perturbation norm ε, and are reported in Figure 7b and Figure 7c, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' As anticipated in Subsection 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1, we show the performance of two implementations of the method, one based on LSMR and one based on LSMR preconditioned 22 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO 1 2 3 4 5 6 7 8 (a) Example of Triangulation and Holes 102 103 10 102 execution time, s ν = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35, LSMR LSMR, ichol 102 103 10 102 103 number of edges, m execution time, s ν = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5, LSMR LSMR, ichol (b) Time (in seconds) 102 103 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='0 pertubation norm ν = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35, LSMR 102 103 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='0 number of edges, m perturbation norm ν = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5, LSMR (c) Perturbation norm, ε Figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benchmarking Results on the Synthetic Triangulation Dataset: varying sparsities ν = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 and N = 16, 22, 28, 34, 40;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' each network is sampled 10 times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Shapes correspond to the number of eliminated edges in the final perturbation: 1 : �, 2 : □, 3 : �, 4 : △.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For each pair (ν, N), the un-preconditioned and Cholesky-preconditioned execution times are shown.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' by using the incomplete Cholesky factorization of the initial matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We observe that, the computational cost of the whole procedure lies between O(m2) and O(m3) denser structures, with a higher number of vertices, result in the higher number of edges being eliminated;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' at the same time, even most dense cases still can exhibit structures requiring the elimination of a single edge, showing that the flow does not necessarily favor multi-edge optima;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the required perturbation norm ε is growing with the size of the graph, Figure 7c, but not too fast: it is expected that denser networks would require larger ε to create a new hole;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' at the same time if the perturbation were to grow drastically with the sparsity ν, it would imply that the method tries to eliminate sufficiently more edges, a behavior that resembles convergence to a sub-optimal perturbation;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' preconditioning with a constant incomplete Cholesky multiplier, computed for the initial Laplacians, provides a visible execution time gain for medium and large networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Since the quality of the preconditioning deteriorates as the flow approaches the minimizer (as a non-zero eigenvalue becomes 0), it is worth investigating the design of a preconditioner for the up-Laplacian that can be efficiently updated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Transportation Networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Finally, we provide an application to real-world examples based on city transportation networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We consider networks for Bologna, Anaheim, Berlin Mitte, and Berlin Tiergarten;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' each network consists of nodes — intersections/public transport stops — connected by edges (roads) and subdivided into zones;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' for each road the free flow time, length, speed limit are known;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' moreover, the travel demand for each pair of nodes is provided through the dataset of recorded trips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' All the datasets used here are publicly available at https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='com/bstabler/TransportationNetworks;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Bologna network is provided by the Physic Department of the University of Bologna (enriched through the Google Maps API https://developers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='google.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='com/maps).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The regularity of city maps naturally lacks 3-cliques,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' hence forming the simplicial complex ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='26 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='27 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='28 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='29 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='31 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='33 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='34 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='36 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='37 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='38 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='39 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='41 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='42 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='43 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='44 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='46 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='47 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='48 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='49 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='51 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='52 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='53 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='54 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='55 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='56 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='57 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='58 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='59 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='Bologna: Regional Network ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='26 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='27 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='28 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='29 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='31 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='33 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='34 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='36 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='37 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='38 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='39 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='41 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='42 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='43 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='44 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='46 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='47 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='48 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='49 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='51 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='52 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='53 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='54 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='55 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='56 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='57 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='58 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='59 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1st original eigenflow ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='26 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='27 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='28 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='29 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='31 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='33 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='34 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='36 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='37 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='38 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='39 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='41 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='42 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='43 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='44 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='46 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='47 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='48 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='49 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='51 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='52 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='53 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='54 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='55 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='56 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='57 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='58 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='59 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2nd original eigenflow ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='25 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='26 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='27 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='28 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='29 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='30 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='31 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='32 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='33 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='34 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='35 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='36 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='37 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='38 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='39 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='41 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='42 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='43 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='44 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='45 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='46 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='47 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='48 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='49 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='50 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='51 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='52 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='53 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='54 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='55 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='56 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='57 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='58 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='59 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='−∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='∥v∥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='created eigenflow ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Example of the Transportation Network for Bologna.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Left pane: original zone graph where the width of edges corresponds to the weight, to-be-eliminated edge is colored in red.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Right pane: eigenflows, original and created;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' color and width correspond to the magnitude of entries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' based on triangulations as done before frequently leads to trivial outcomes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Instead, here we “lift” the network to city zones, thus more effectively grouping the nodes in the graph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Specifically: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' we consider the completely connected graph where the nodes are zones in the city/region;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' the free flow time between two zones is temporarily assigned as a weight of each edge: the time is as the shortest path between the zones (by the classic Dijkstra algorithm) on the initial graph;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' similarly to what is done in the filtration used for persistent homology, we filter out exces- sively distant nodes;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' additionally, we exclude the longest edges in each triangle in case it is equal to the sum of two other edges (so the triangle is degenerate and the trip by the longest edge is always performed through to others);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' finally, we use the travel demand as an actual weight of the edges in the final network;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' travel demands are scaled logarithmically via the transformation wi �→ log10 � wi 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='95 min wi � ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' see the example on the left panel of Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Given the definition of weights in the network, high instability (corresponding to small per- turbation norm ε) implies structural phenomena around the “almost-hole”, where the faster and shorter route is sufficiently less demanded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In the case of Bologna, Figure 8, the algorithm eliminates the edge [11, 47] (Casalecchio 24 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO Table 1 Topological instability of the transportation networks: filtered zone networks with the corresponding pertur- bation norm ε and its percentile among w1(·) profile.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For each simplicial complex the number of nodes, edges and triangles in V2(K) are provided alongside the initial number of holes β1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The results of the algorithm consist of the perturbation norm, ε, computation time, and approximate percentile p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' network β1 logarithmic weights n m ∆ time ε p Bologna 60 175 171 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='43s 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='003 [11, 47] (4th smallest) Anaheim 38 159 221 1 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='39s 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='003 [10, 29] (11th smallest) Berlin-Tiergarten 26 63 55 0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='46s 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='18 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='015 [6, 16] (20th smallest) Berlin-Mitte 98 456 900 1 127s 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='887 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='0016 [57, 87] (6th), [58, 87], (17th) di Reno – Pianoro) creating a new hole 6 − 11 − 57 − 47.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We also provide examples of the eigenflows in the kernel of the Hodge Laplacian (original and additional perturbed): original eigenvectors correspond to the circulations around holes 7−26−12−20 and 8−21−20−16−37 non-locally spread in the neighborhood [29].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' The results for four different networks are summarized in the Table 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' p mimics the percentile, ε/ � e∈V1 wi(e), showing the overall small perturbation norm contextually.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' At the same time, we emphasize that except Bologna (which is influenced by the geographical topology of the land), the algorithm does not choose the smallest weight possible;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' indeed, given our interpretation of the topological instability, the complex for Berlin-Tiergarten is stable and the transportation network is effectively constructed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Discussion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' In the current work, we formulate the notion of k-th order topological stability of a simplicial complex K as the smallest data perturbation required to create one additional k-th order hole in K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' By introducing an appropriate weighting and normaliza- tion, the stability is reduced to a matrix nearness problem solved by a bi-level optimization procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Despite the highly nonconvex landscape, our proposed procedure alternating con- strained and free gradient steps yields a monotonic descending scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Our experiments show that this approach is generally successful in computing the minimal perturbation increasing β1(ε, E), even for potentially difficult cases, as the one proposed in Subsection 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' For simplicity, here we limit our attention to the smallest perturbation that introduces only one new hole.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' However, a simple modification may be employed to address the case of the introduction of m new holes: include the sum of m nonzero eigenvalues of Lup 1 (ε, E) rather than just the first one in the spectral functional (4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' We also remark that, due to the spectral inheritance principle Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='7, the proposed framework for H1 can be in principle extended to a general Hk;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' however, this extension requires nontrivial considerations on the data modification procedure and on the numerical linear algebra tools, as a nontrivial topology of higher-order requires a much denser network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Different improvements are possible in terms of numerical implementation, including in- vestigating the use of more sophisticated (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' implicit) integrators for the gradient flow system (Equation (5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='9)), which would additionally require the use of higher-order derivatives QUANTIFYING THE STRUCTURAL STABILITY OF SIMPLICIAL HOMOLOGY 25 of λ+(ε, E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Moreover, as already mentioned in Subsection 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1, the numerical method for the computation of the small singular values would benefit from the use of an efficient precondi- tioner that can be effectively updated throughout the flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Investigations in this direction are in progress and will be the subject of future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' REFERENCES [1] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Altenburger and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Ugander, Monophily in social networks introduces similarity among friends-of-friends, Nature human behaviour, 2 (2018), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 284–290.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [2] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Andreotti, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Edelmann, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Guglielmi, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lubich, Graph partitioning using matrix differ- ential equations, 2017, https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='48550/ARXIV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1712.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='05993.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [3] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Arrigo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Higham, and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tudisco, A framework for second-order eigenvector centralities and clustering coefficients, Proceedings of the Royal Society A, 476 (2020), p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 20190724.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [4] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Battiston, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Cencetti, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Iacopini, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Latora, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lucas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Patania, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Young, and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Petri, Networks beyond pairwise interactions: Structure and dynamics, Physics Reports, 874 (2020), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1–92, https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='physrep.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='05.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [5] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benson, Three hypergraph eigenvector centralities, SIAM Journal on Mathematics of Data Science, 1 (2019), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 293–312.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [6] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benson, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Abebe, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Schaub, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Jadbabaie, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Kleinberg, Simplicial closure and higher-order link prediction, Proceedings of the National Academy of Sciences, 115 (2018), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E11221–E11230.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [7] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benson, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Gleich, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Leskovec, Higher-order organization of complex networks, Science, 353 (2016), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 163–166, https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='1126/science.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='aad9029.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [8] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benson, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Gleich, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Leskovec, Higher-order organization of complex networks, Science, 353 (2016), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 163–166.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [9] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Bick, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Gross, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Harrington, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Schaub, What are higher-order networks?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=', CoRR, abs/2104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11329 (2021), https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='org/abs/2104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='11329.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [10] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Chen and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Meila, The decomposition of the higher-order homology embedding constructed from the k-Laplacian, Advances in Neural Information Processing Systems, 34 (2021), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 15695–15709.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [11] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Meil˘a, and I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Kevrekidis, Helmholtzian eigenmap: Topological feature discovery and edge flow learning from point cloud data, rXiv:2103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='07626, (2021), https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='48550/ ARXIV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='2103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='07626.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [12] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Ebli and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Spreemann, A notion of harmonic clustering in simplicial complexes, in 2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA), IEEE, 2019, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1083– 1090.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [13] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Estrada and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Higham, Network properties revealed through matrix functions, SIAM review, 52 (2010), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 696–714.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [14] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Fiedler, Laplacian of graphs and algebraic connectivity, Banach Center Publications, 25 (1989), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 57–70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [15] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Fong and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Saunders, Lsmr: An iterative algorithm for sparse least-squares problems, SIAM Journal on Scientific Computing, 33 (2011), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2950–2971.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [16] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Fortunato and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hric, Community detection in networks: A user guide, Physics reports, 659 (2016), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1–44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [17] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Fountoulakis, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Li, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Yang, Local hyper-flow diffusion, Advances in Neural Information Processing Systems, 34 (2021), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 27683–27694.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [18] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Friedkin, Theoretical foundations for centrality measures, American journal of Sociology, 96 (1991), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1478–1504.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [19] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Ghaderi and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Srikant, Opinion dynamics in social networks with stubborn agents: Equilibrium and convergence rate, Automatica, 50 (2014), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 3209–3215.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [20] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Grippo, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lampariello, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lucidi, A class of nonmonotone stabilization methods in uncon- strained optimization, Numerische Mathematik, 59 (1991), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 779–805.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [21] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Guglielmi and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lubich, Matrix nearness problems and eigenvalue optimization, in preparation, 26 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' GUGLIELMI, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' SAVOSTIANOV, AND F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' TUDISCO 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [22] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Horak and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Jost, Spectra of combinatorial Laplace operators on simplicial complexes, Advances in Mathematics, 244 (2013), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 303–336.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [23] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Horn and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Johnson, Matrix Analysis, Cambridge University Press, 1990.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [24] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lim, Hodge Laplacians on graphs, SIAM Review, (2015), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 685–715.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [25] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Muhammad and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Egerstedt, Control using higher order laplacians in network topologies, in Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' of 17th International Symposium on Mathematical Theory of Networks and Systems, Citeseer, 2006, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1024–1038.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [26] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Nettasinghe, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Krishnamurthy, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lerman, Diffusion in social networks: Effects of monophilic contagion, friendship paradox and reactive networks, IEEE Transactions on Network Sci- ence and Engineering, (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [27] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Newman, Finding community structure in networks using the eigenvectors of matrices, Physical review E, 74 (2006), p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 036104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [28] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Otter, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Porter, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tillmann, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Grindrod, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Harrington, A roadmap for the computation of persistent homology, EPJ Data Science, 6 (2017), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1–38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [29] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Schaub, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benson, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Horn, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Lippner, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Jadbabaie, Random walks on simplicial complexes and the normalized Hodge 1-Laplacian, SIAM Review, 62 (2020), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 353–391.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [30] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Spielman and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Teng, Nearly linear time algorithms for preconditioning and solving sym- metric, diagonally dominant linear systems, SIAM Journal on Matrix Analysis and Applications, 35 (2014), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 835–885.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [31] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Strogatz, From kuramoto to crawford: exploring the onset of synchronization in populations of coupled oscillators, Physica D: Nonlinear Phenomena, 143 (2000), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 1–20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [32] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tudisco, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Arrigo, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Gautier, Node and layer eigenvector centralities for multiplex networks, SIAM Journal on Applied Mathematics, 78 (2018), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 853–876.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [33] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tudisco, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Benson, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Prokopchik, Nonlinear higher-order label spreading, in Proceedings of the Web Conference 2021, 2021, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2402–2413.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [34] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tudisco and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hein, A nodal domain theorem and a higher-order cheeger inequality for the graph p-laplacian, (2016), https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='org/abs/1602.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content='05567.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [35] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tudisco and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Higham, Core-periphery detection in hypergraphs, SIAM Journal on Mathematics of Data Science, (to appear).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [36] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Tudisco, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Mercado, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Hein, Community detection in networks via nonlinear modularity eigenvectors, SIAM Journal on Applied Mathematics, 78 (2018), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 2393–2419.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' [37] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' Vigna, Spectral ranking, Network Science, 4 (2016), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
+page_content=' 433–445.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/W9E2T4oBgHgl3EQfDwYa/content/2301.03627v1.pdf'}
diff --git a/WNFOT4oBgHgl3EQf7TSv/content/tmp_files/2301.12962v1.pdf.txt b/WNFOT4oBgHgl3EQf7TSv/content/tmp_files/2301.12962v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7d0686de55588a59054108eb79ae518fdd6662da
--- /dev/null
+++ b/WNFOT4oBgHgl3EQf7TSv/content/tmp_files/2301.12962v1.pdf.txt
@@ -0,0 +1,2735 @@
+Hierarchical Imitation Learning with Vector Quantized Models
+Kalle Kujanp¨a¨a1, Joni Pajarinen2, and Alexander Ilin1
+1Department of Computer Science, Aalto University, Finland
+2Department of Electrical Engineering and Automation, Aalto University, Finland
+Abstract
+The ability to plan actions on multiple levels of abstraction enables intelligent agents to solve complex tasks
+effectively. However, learning the models for both low and high-level planning from demonstrations has proven
+challenging, especially with higher-dimensional inputs. To address this issue, we propose to use reinforcement
+learning to identify subgoals in expert trajectories by associating the magnitude of the rewards with the predictability
+of low-level actions given the state and the chosen subgoal. We build a vector-quantized generative model for the
+identified subgoals to perform subgoal-level planning. In experiments, the algorithm excels at solving complex,
+long-horizon decision-making problems outperforming state-of-the-art. Because of its ability to plan, our algorithm
+can find better trajectories than the ones in the training set.
+1
+Introduction
+Learning from expert demonstrations has proven successful in many sequential decision-making settings that can be
+modeled with Markov decision processes (Abbeel and Ng, 2004). Imitation learning (IL) is a technique for learning to
+imitate the behavior of an expert by discovering the mapping between states and actions without access to information
+such as rewards (Osa et al., 2018). IL has proven useful in aviation (Sammut et al., 1992), autonomous driving (Chen
+and Kr¨ahenb¨uhl, 2022), robotics (Kober and Peters, 2008), video games (Vinyals et al., 2019) and even healthcare
+(Mayer et al., 2008).
+Recent advances in planning with learned dynamics models have improved our ability to solve complex long-horizon
+problems when interacting with the environment is possible (Hafner et al., 2022; Ye et al., 2021; Schrittwieser et al.,
+2020). Learning models for planning in the offline setting is possible as well (Argenson and Dulac-Arnold, 2020), but
+these model-based reinforcement learning (RL) methods assume access to environment rewards and are not directly
+applicable to the IL setting. Hierarchical RL with decision-making at multiple time scales has succeeded in tasks
+where flat RL struggles (Hafner et al., 2022; Nachum et al., 2019). A hierarchy should also be useful in the IL setting
+as many real-world problems have a natural hierarchical structure (Sharma et al., 2019; Jing et al., 2021). Moreover,
+planning with a hierarchy may shorten the effective planning horizon and avoid compounding model errors also in the
+IL setting (Nair and Finn, 2019).
+We propose a method for hierarchical planning in the IL setting that relies on segmenting expert trajectories into
+subtasks without any high-level supervision. Unlike prior work that assumes a fixed number of subgoals (Pertsch et al.,
+2020b), fixed-length subtasks (Czechowski et al., 2021), or trains multiple models to deal with subtasks of different
+lengths (Zawalski et al., 2022), our algorithm segments the trajectories into a variable number of variable-length
+subtasks. We use the segmentation to learn a generative model over the subgoals and a subgoal-conditioned low-level
+policy to execute the subtasks. To perform high-level planning, we use standard search methods such as Policy-Guided
+Heuristic Search (Orseau and Lelis, 2021), Monte Carlo Tree Search (Coulom, 2006), or A* (Hart et al., 1968) in which
+our generative model is used for node expansion. Our method outperforms strong search, hierarchical IL, and offline
+RL algorithms at complex long-horizon decision-making problems. Our experiments also show that our algorithm can
+handle suboptimal expert trajectories and self-improve. In summary, the main contributions of this work are:
+1. A novel yet conceptually simple RL approach for identifying subgoals from trajectories based on the prediction
+performance of a low-level policy.
+2. A VQVAE generative model for proposing subgoals for planning with temporal abstraction.
+3. In experiments, our generative model combined with the learned low-level policy and a suitable high-level
+search algorithm solves complex problems with sparse rewards better and with fewer node expansions than
+state-of-the-art subgoal search and outperforms offline RL algorithms.
+1
+arXiv:2301.12962v1 [cs.AI] 30 Jan 2023
+
+2
+Related Work
+Our method combines hierarchical discrete planning with imitation learning to solve complex planning problems.
+Hierarchical planning allows shortening the effective planning horizon, which is beneficial in long-horizon tasks
+(Pertsch et al., 2020a; Nair and Finn, 2019). We use offline data to learn a VQVAE (Van Den Oord et al., 2017) that
+generates adaptive horizon subgoals and a low-level policy to reach the subgoals for high-level planning. Previous
+work proposes a wide variety of different approaches for learning hierarchical behavior from data, also in combination
+with planning. However, according to our knowledge, our method is the first method that segments a trajectory into
+a varying number of adaptive-length subtrajectories and uses the subgoals to learn a generative model that can solve
+difficult long-term decision-making problems in a discrete setting.
+Hierarchical Continuous Planning. In hierarchical planning with continuous control, the Cross-Entropy
+Method (CEM) is typically used. With continuous controls, search methods with convergence guarantees such as
+MCTS (Coulom, 2006), PHS∗ (Orseau and Lelis, 2021), or A∗ (Hart et al., 1968), utilized with our method in the
+experiments, cannot be directly applied. HiGoC (Li et al., 2022) uses CEM for hierarchical planning in offline RL
+assuming access to rewards. HVF (Nair and Finn, 2019) adds a hierarchical structure with predicted subgoal images
+to Visual Model Predictive Control (Ebert et al., 2018) that plans in image space using CEM. Visual hierarchical
+methods that rely on identifying keyframes from trajectories using variational inference and planning with CEM
+include TAP (Jayaraman et al., 2018) and KeyIn (Pertsch et al., 2020b). However, these methods assume a fixed
+number of keyframes in the trajectories and have not been successfully applied to highly complex reasoning tasks
+with sparse rewards. KeyIn also uses an environment simulator for planning.
+Goal-Conditioned Hierarchical Planning (Pertsch et al., 2020a) produces plans, executed by a learned inverse
+dynamics model, in a high-dimensional state space in an offline setting. SGT-PT does goal-based RL in a low-
+dimensional setting by planning with subgoal trees (Jurgenson et al., 2020). However, these models require an explicit
+goal state and need to generate subgoals between the initial state and the goal state, which can be a difficult learning
+problem in complex long-horizon environments.
+Hierarchical IL without Planning. Unlike our method, many model-free hierarchical imitation learning
+methods assume some degree of high-level supervision (Le et al., 2018; Fox et al., 2019; Zhang et al., 2021). As an
+alternative Daniel et al. (2016) infer compositional structure in data discovering options (Sutton et al., 1999) (a
+set of low-level policies with termination) with expectation-maximization. CompILE (Kipf et al., 2019) uses VAEs
+to segment trajectories into subtasks and the subtask encodings as subpolicies in hierarchical RL. The model-free
+method Option-GAIL (Jing et al., 2021) infers expert options from trajectories with an EM-like approach. Zhang
+and Paschalidis (2021) directly optimize a hierarchical policy with options by maximizing the probability of expert
+trajectories with a hidden Markov model. Directed-Info GAIL (Sharma et al., 2018) is a variant of hierarchical inverse
+RL that learns latent policies by modeling problems as directed graphs. The method segments expert trajectories
+into sub-tasks and learns structural policies to solve different sub-tasks. OptionGAN (Henderson et al., 2018) learns
+to recover reward and policy options simultaneously. Learning from Guided Play (LfGP) (Ablett et al., 2021) uses
+scheduled auxiliary tasks to address lacking exploration in adversarial online IL. Paul et al. (2019) learn a generative
+model over subgoals from demonstrations and use it to augment the reward function for RL fine-tuning. However,
+these methods do not incorporate high-level planning mechanisms, which may make them unsuitable for solving
+complex reasoning problems.
+Offline RL. In offline RL, the agent’s objective is to learn an optimal policy without interacting with the
+environment. Instead, the agent has access to a dataset of transitions that have been collected by a behavior policy
+πβ that can be suboptimal. A significant benefit that offline RL has over imitation learning is the ability to extract
+strong policies even when the expert trajectories are suboptimal (Kumar et al., 2022). Conservative Q-learning
+(Kumar et al., 2020) is a model-free offline RL method that learns a lower bound on the policy value, which helps
+avoid overestimating state values. Decision Transformer (Chen et al., 2021) treats the offline reinforcement learning
+task as a sequence modeling problem, where the goal is to predict the action conditioned on a desired reward. We use
+offline RL methods as baselines in our experiments.
+Hierarchical IL with Discrete Search. The closest works to ours in search and planning are kSubS (Czechowski
+et al., 2021) and AdaSubS (Zawalski et al., 2022). kSubs learns an autoregressive model for generating subgoals given
+a set of trajectories and uses subgoal search for planning. Unlike our method, kSubS relies on the true environment
+dynamics for low-level search to find fixed-length subtrajectories between subgoals in some environments when
+combined with an autoregressive CNN. AdaSubS replaces the low-level search of kSubS with a learned policy in all
+environments and supports subtrajectories of multiple hard-coded lengths by training an autoregressive generative
+model for each length. Our approach relies on a learned low-level policy and dynamics model and supports varying-
+length segments. We show that it is possible to train a single non-autoregressive adaptive generative model, which
+makes generating subgoals more efficient.
+2
+
+Trajectory
+Segmentation
+Low-Level Policy
+πθ(a | s, sg)
+Subgoals
+Generative Model
+for Subgoals
+VQVAE
+VQVAE
+Low-Level Policy
+πθ(a | s, sg)
+Figure 1: A visualization of our Hierarchical Imitation Planning with Search (HIPS) when learning to solve Sokoban.
+The main components of our method are a detector dξ(sgk|si) that segments the trajectory into subgoals, a subgoal-
+conditioned low-level policy πθ(ai|si, sgk), and the VQVAE, a generative model over the subgoals. The low-level
+policy and VQVAE are used during evaluation for planning, whereas the detector is training-only.
+3
+Method
+We consider goal-oriented, complex reasoning problems, in which the agent’s objective is to act in the environment to
+reach a terminal state. This corresponds to a Markov decision process with a reward of one upon solving the task and
+zero otherwise. We consider environments with fully Markovian, discrete-valued states with full observability. We work
+in the imitation learning (offline) setting: the agent needs to learn to solve tasks only from available demonstrations
+without the possibility to interact with the environment before evaluation. We assume that there is a dataset D of
+trajectories τ = {s0, a0, s1, . . . , aT −1, sT } collected by experts who know how to solve tasks, with only a reward of
+one at the terminal state sT . The experts may not reach the terminal states in the fastest possible way.
+We propose to solve the imitation learning task using an agent which has a hierarchical structure with two levels.
+Our agent learns a hierarchical representation of the available trajectories by identifying likely experts’ subgoals in
+the existing trajectories in an unsupervised manner. The identified subgoals are then used to train a discrete-code
+generative model which can generate reasonable subgoals to perform subgoal-level planning with standard search
+algorithms such as PHS, MCTS, or A*. We also learn a low-level policy that executes the plan generated by the
+planner by sequentially reaching the determined subgoals. A graphical representation of our method is shown in Fig. 1.
+We call our method Hierarchical Imitation Planning with Search (HIPS). Below we describe the main components of
+the approach.
+3.1
+Subgoal Identification
+The goal of the subgoal identification phase is to learn a high-level representation τ∗ = {sg1, sg2, . . . , sgM } of each
+trajectory such that the trajectory is represented as a sequence of subgoals sgk. Each subgoal is a state from the
+trajectory, that is ∀ksgk ∈ τ, and in particular, sgM = sT . This is a time series segmentation problem where the
+solution has two desirable qualities: 1) we want the identified subgoals sgk to be easy to reach by a trainable low-level
+policy πθ(ai|si, sgk) which takes the subgoal sgk as input and 2) we want the subgoals to be easy to sample from a
+generative model.
+Successfully segmenting the trajectories into a variable number of variable-length segments turns out to be a highly
+non-trivial task, as most prior work has focused on fixed-length segments or a fixed number of segments. Our solution
+is to formulate this segmentation task as an RL problem in which we treat each trajectory from D as one episode for
+training. In each episode, the segmentation agent starts at the first state s0 of the trajectory and selects the next
+subgoal state sg1 according to the probabilities produced by its policy dξ(sg1|s0), which we call the detector. We limit
+the detector to consider only the following H states as candidate subgoals, that is sg1 is selected from s1, ..., sH. The
+3
+
+区
+区区
+区区区
+区区区
++
+区区区
+区Algorithm 1 Segmenting Trajectories for Hierarchical IL
+Input: A dataset of trajectories D, untrained low-level policy network πθ, detector dξ
+Parameters: The parameters of the low-level policy network, θ, detector network, ξ
+Output: Trained low-level policy π, dataset D′ of subgoal pairs {(sgk+1, sgk)}
+1: while πθ, dξ not converged do
+2:
+Sample a trajectory τ.
+3:
+Segment τ with dξ
+4:
+Predict low-level actions with πθ conditioned on produced subgoals
+5:
+Compute returns (Equation 1), update ξ with REINFORCE
+6:
+Compute the losses for πθ (Equation 2), update θ
+7: end while
+8: Create a dataset D′ of subgoal pairs {(sgk+1, sgk)} by sampling trajectories τ from D and segmenting them with
+dξ.
+9: return πθ, D′
+agent samples the next subgoal according to the computed probabilities and gets the reward
+R1 = r1 − α,
+(1)
+where r1 is the log-probability that a low-level policy πθ(ai|si, sg1) selects the sequence of actions a0, ..., ag1−1 in the
+first segment:
+r1 =
+g1−1
+�
+i=0
+log πθ(ai|si, sg1)
+and α is a penalty to prevent segmentation into too many subtrajectories. After that, the segmentation agent
+changes its state to sg1, selects the next subgoal according to dξ(sg2|sg1) and gets reward R2 computed using action
+log-probabilities from the second segment, similarly to (1). The episode continues like this until the end of the
+trajectory is reached.
+The low-level policy πθ(ai|si, sgk) is considered as part of the environment of the segmentation agent. It is updated
+during training using goal-conditioned behavioral cloning to minimize
+Lθ = −Eτ∼D
+M(τ)
+�
+k=1
+−1+gk
+�
+i=gk−1
+log πθ(ai|si, sgk),
+(2)
+with subgoals sgk produced by the segmentation agent. Note that the number of identified subgoals M(τ) may vary
+across trajectories.
+Thus, the segmentation agent is trained by giving it a higher reward when selected subgoals lead to more accurate
+action predictions by the low-level policy. The low-level policy is trained concurrently with the subgoals (high-level
+commands) produced by the segmentation agent as input. We train the segmentation agent using the policy gradient
+algorithm REINFORCE with a learnable value function as baseline (Williams, 1992). Note that using a trainable
+detector dξ(sgk+1|sgk) encourages subgoals that are easy to recognize among the states in the training trajectories,
+and therefore, such subgoals might be easy to produce by a learned generative model. Our approach for segmenting
+trajectories is summarized in Algorithm 1.
+3.2
+Generative Model for Subgoals
+To plan in terms of the high-level subgoals, the agent needs the ability to generate reasonable subgoals sg for each
+environment state s. We do this by learning a generative model p(sg|s) over the subgoals using the ones identified in
+the trajectory segmentation step as training data. We implement the model as a VQVAE with discrete latent codes,
+which is inspired by the vector quantized models proposed by Ozair et al. (2021). The VQVAE encoder takes a pair
+of states (sg, s) as input and outputs a continuous latent code ze. Then, the code is quantized by finding the nearest
+code ek from the codebook such that k = arg minm ∥ze − em∥2. The decoder uses the code ek to reconstruct the
+subgoal state ˆsg = gψ(ek, s). The loss minimized during training is
+L = Lrec(ˆsg, sg) + ∥[ze] − ek∥2
+2 + β∥ze − [ek]∥2
+2 ,
+(3)
+4
+
+Sokoban
+Sliding Tile Puzzle
+Box-World
+Traveling Salesman Problem
+Figure 2: The environments we consider in the experiments. Sokoban: the task is to push yellow boxes onto the
+target locations (marked with red squares). Sliding Tile Puzzle (STP): The task is to order the tiles from 1 to 24 by
+moving them. Box-World: The agent must collect colored keys and open color-matching locks to recover more keys
+until it finally reaches a goal target (marked with the $ sign). Traveling salesman problem (TSP): The agent (marked
+with the circle) has to visit all cities (marked with squares) before returning to the start (marked with the black
+square). Visited cities are marked with green squares and unvisited ones with red squares.
+where Lrec is the reconstruction loss and [·] denotes the stop gradient operation.
+We train the VQVAE in two stages. In the pre-training stage, we use random pairs of states (sj, si), i < j ≤ i + H
+from the trajectories as inputs (sg, s). We skip the discretization layer and only use the reconstruction loss Lrec(ˆsg, sg)
+for training the encoder and the decoder. After the pre-training has converged, the complete VQVAE is trained by
+using pairs of consecutive subgoals (sgk+1, sgk) as input. We initialize the codebook by running KMeans++ clustering
+(Arthur and Vassilvitskii, 2006) on the first batches of encoder outputs and using the cluster centers as the initial
+codes. This training strategy was inspired by the strategy proposed by �La´ncucki et al. (2020). Finally, we learn a
+subgoal-conditioned prior p(ek+1|sgk) over the latent codes.
+Once the model has been trained, one can generate subgoals conditioned on the current state s by sampling a
+code e from the learned codebook and running it through the decoder gψ(e, s). Note that the number of possible
+codes e is finite, which means that the number of generated subgoals sg is finite as well. Note also that distinct codes
+e may result in the same generated subgoal sg, which is a desired behavior when the size of the codebook is larger
+than the number of reasonable subgoals for the considered state. The pseudo-code for our VQVAE training is given
+in Algorithm 2 in the Appendix F.
+3.3
+High-Level Planning with Search
+We perform planning in the subgoal space, as it can be more efficient and suitable for long-horizon planning than
+planning in the state space (Nair and Finn, 2019). We demonstrate that our method is compatible with many different
+search algorithms. Each search node represents a subgoal. When a search node is expanded, possible next subgoals
+(child nodes) are generated with the VQVAE. Given a codebook of size K, there are K possible child nodes for each
+subgoal. To limit the size of the search tree, we remove duplicates and unreachable subgoals. The reachability of a
+proposed subgoal sg from state si is evaluated by using the low-level policy πθ(ai|si, sg) trained in the segmentation
+phase. We run the policy iteratively from state si and simulate state transitions by using the true dynamics or a
+learned model fdyn(si+1|ai, si). If the subgoal sg is reached within a specific number of steps, the subgoal is considered
+reachable, and a search node is created. In this work, we work with discrete states and require an exact match
+between the reached state and the subgoal. Using a suitable threshold to evaluate the match is an alternative in the
+continuous setting. We also learn a value function V (s) that predicts the number of low-level steps necessary to reach
+the goal (terminal state) and use it as a heuristic in planning.
+The search methods we use are Greedy Best-First Search (GBFS), Policy-Guided Heuristic Search (PHS*, Orseau
+and Lelis, 2021), A* (Hart et al., 1968) and Monte-Carlo Tree Search (MCTS, Coulom, 2006; Kocsis and Szepesv´ari,
+2006). PHS* is dependent on a good policy, but when the dataset contains suboptimal trajectories, learning a good
+VQVAE prior to act as the policy might be impossible. Then, policy-independent algorithms like A* or GBFS can be
+superior. PHS* is also aimed at minimizing the search loss, not finding a particularly high-quality solution. When
+that is important, A* or MCTS can be superior to PHS*.
+5
+
+3
+9
+13
+12
+5
+7
+2
+4
+16
+1
+X
+18
+8
+14
+11
+15
+10
+19
+6
+17
+XX
++
++
+X
+20
+21
+22
+23
+2410
+0?
+180?
+180
+0?
+10?
+1818
+184
+Experiments
+In our experimental phase, we evaluate our method on complex, sparse reward problems that require reasoning. We
+compare our method to existing search, hierarchical imitation learning, and offline RL methods. We also analyze
+whether our RL-based approach for identifying subgoals is superior to subgoals sampled at fixed intervals.
+4.1
+Environments
+We evaluate our method in four environments that are all complex reasoning domains (see Fig. 2).
+The first
+environment is Sokoban, which is a PSPACE-complete puzzle where the agent must push boxes onto goal locations
+(Culberson, 1997). The moves are irreversible and one wrong push can make the puzzle unsolvable. We use a 10 × 10
+problem size with four boxes, the default configuration in the earlier literature (Orseau and Lelis, 2021; Guez et al.,
+2019; Racani`ere et al., 2017). We use a one-hot encoded tensor with shape 10 × 10 × 4 as the observation space
+(Orseau and Lelis, 2021).
+The second environment is the sliding tile puzzle (STP) which is a classic benchmark in the search literature
+(Korf, 1985). We use a puzzle size of 5 × 5, and the objective is to sort the number tiles in a specific descending order.
+The third environment is Box-World (BW), where the agent must collect colored keys and open color-matching
+locks to recover more keys until it finally reaches a goal target (Zambaldi et al., 2018). Keys can only be used once. If
+the agent uses its key to open the wrong box, the game will become unsolvable. Hence, careful planning and reasoning
+about entities and their relations are required.
+The fourth environment is a grid-based Traveling Salesman Problem (TSP). The agent moves through the 2D
+grid to visit all the cities before returning to the starting point (see Fig. 2). The TSP is an NP-hard combinatorial
+optimization problem. However, finding any solution to TSP is relatively easy. Hence, grid-based TSP serves as an
+environment for evaluating the solution quality.
+In Sokoban, we collect a training set of 10340 trajectories using gym-sokoban (Schrader, 2018). 10240 of the
+problem instances have been solved using Curry (Shoham, 2021) and 100 trajectories were collected by performing
+random actions. The training set consists of 5100 trajectories in STP and 22100 in Box-World, of which 5000 and
+22000 were collected by solving the problem instances with a subgoal-based A* algorithm (Hart et al., 1968) and 100
+by performing random actions. In subgoal-based A*, we generated subgoals progressively closer to the terminal state
+procedurally and executed A* to reach these subgoals, as solving complete problem instances with A* would have
+been computationally very expensive due to the complexity of the environments. In TSP, we do not limit the number
+of demonstrations available to the agent. Instead, we generate highly suboptimal trajectories by running an agent
+that visits 25 cities in a 25 × 25 grid in random order.
+4.2
+Agents
+Our HIPS agent consists of the following neural networks: the detector dξ(sgk+1|sgk), the low-level policy πθ(ai|si, sgk),
+the VQVAE encoder fφ(zek+1|sgk+1, sgk), the VQVAE decoder gψ(sgk+1|ek+1, sgk), the VQVAE prior p(ek+1|sgk), the
+low-level dynamics model fdyn(si+1|ai, si), and the distance function V (si). The encoder fφ and detector dξ are not
+used during evaluation. All networks are ResNet-based CNNs (He et al., 2016). The decoder and low-level dynamics
+CNNs also contain FiLM layers (Perez et al., 2018). We only use the 100 random trajectories for training the low-level
+dynamics model. In Box-World, the neural networks additionally use Deep Recurrent Convolutions (Guez et al.,
+2019). All neural networks are implemented with PyTorch (Paszke et al., 2019) and trained with an Adam optimizer
+(Kingma and Ba, 2014). We evaluate Sokoban and Box-World with PHS* as the high-level search algorithm. Because
+of the suboptimality of the TSP trajectories, learning a good VQVAE prior is difficult, and we use A* in TSP. We
+also observed that GBFS is superior to PHS* with very small search budgets in STP (see Appendix D).
+We compare our agents to two main classes of baselines. The first class of baselines is strong IL and offline
+RL algorithms. We compare our agent to standard flat behavioral cloning (BC), a powerful offline RL algorithm,
+Conservative Q-Learning (CQL, Kumar et al., 2020), and a strong IL algorithm, Inverse Soft-Q Learning (IQ-Learn,
+Garg et al., 2021). We ran IQ-Learn in the online mode, where it could collect additional data during training.
+Furthermore, we include the Decision Transformer (DT, Chen et al., 2021), the hierarchical IL algorithm Option-GAIL
+(Jing et al., 2021), and the online goal-conditioned RL algorithm RIS (Chane-Sane et al., 2021) in our experiments. For
+CQL, we use the implementation of d3rlpy (Seno and Imai, 2021), and for other algorithms, we use the open-sourced
+implementations of the authors. For all methods that need rewards, we give the agent a reward of one upon completing
+the task and 0 otherwise. We do not include the 100 random trajectories in the dataset when evaluating imitation
+learning algorithms.
+6
+
+Table 1: The overall success rates (%) of different algorithms. The algorithms in the bottom part have access to the
+true environment dynamics, and those in the upper part do not.
+Method
+Sokoban
+STP
+BW
+TSP
+HIPS (ours)
+97.5
+94.7
+55.7
+99.9
+HIPS-k (ours)
+99.0
+94.7
+20.1
+100.0
+BC
+18.7
+82.5
+41.1
+28.8
+CQL
+3.3
+11.7
+6.0
+33.6
+DT
+36.7
+0.0
+10.0
+0.0
+RIS
+0.0
+0.0
+7.3
+0.0
+Option-GAIL
+0.3
+0.0
+0.0
+0.0
+IQ-Learn
+0.0
+0.0
+0.0
+0.0
+HIPS-env (ours)
+98.1
+94.6
+99.6
+100.0
+AdaSubS
+91.4
+0.0
+22.4
+kSubS
+90.5
+93.3
+87.9
+Table 2: The success rates (%, higher is better) of different search algorithms after performing N node expansions. In
+Sokoban and Box-World, HIPS was evaluated with PHS*, with GBFS in STP, and with A* in TSP.
+Sokoban
+Sliding Tile Puzzle
+Box-World
+Travelling Salesman
+N
+100
+500
+1000
+100
+500
+1000
+100
+500
+1000
+100
+500
+1000
+HIPS (ours)
+88.5
+94.7
+95.9
+89.8
+91.8
+92.2
+55.7
+55.7
+55.7
+73.5
+99.9
+99.9
+HIPS-k (ours)
+77.7
+90.9
+94.3
+68.0
+79.1
+84.4
+20.1
+20.1
+20.1
+77.0
+100.0
+100.0
+HIPS-env (ours)
+88.5
+94.9
+96.4
+90.2
+92.1
+92.6
+99.6
+99.6
+99.6
+73.9
+100.0
+100.0
+HIPS-env-k (ours)
+76.9
+91.3
+94.0
+69.7
+80.9
+87.2
+99.1
+99.1
+99.1
+77.6
+99.9
+99.9
+AdaSubS
+82.3
+88.8
+90.2
+0.0
+0.0
+0.0
+1.2
+12.2
+20.8
+kSubS
+70.9
+79.9
+82.8
+80.0
+91.7
+92.7
+40.4
+80.9
+85.3
+PHS*
+1.8
+76.1
+91.1
+0.0
+0.0
+0.0
+25.8
+93.8
+94.0
+0.0
+0.0
+0.0
+The second class of baselines is search methods that use the true environment dynamics, a state-of-the-art low-level
+search, Policy Guided Heuristic Search (PHS*, Orseau and Lelis, 2021), and two subgoal-level search methods: kSubS
+(Czechowski et al., 2021) and AdaSubS (Zawalski et al., 2022). In PHS*, we observed that using a policy trained with
+behavioral cloning works better than a policy trained using the loss function proposed by Orseau and Lelis (2021)
+and therefore, we use the BC policy in the experiments. We evaluate kSubS and AdaSubS with the autoregressive
+CNNs on all environments except Box-World because it would have required significant changes to the existing
+implementation. Our method, HIPS, relies on a learned dynamics model instead of the true dynamics. Hence, it
+solves a more complex problem. We also train a more comparable variant of our method, HIPS-env, that uses an
+environment simulator for planning.
+4.3
+Results
+We use the overall success rates reported in Table 1 as the main evaluation metric in all environments. We evaluate
+the performance of each seed and take the mean over the random seeds. We use 10 random seeds to evaluate our
+method, HIPS, at least five seeds per ablation, and at least three seeds per baseline method. When evaluating the
+overall success rate, the search algorithms may perform as many expansions as necessary to find a solution to the
+problems. The critical success factor for our model is the capacity of the generative model to cover the complete
+search space with the proposed subgoals.
+Our method, HIPS, outperforms the baseline methods in all four environments (see Table 1). The table also
+contains an ablation of our method, HIPS-k, where we train the VQVAE with subgoals sampled at fixed intervals
+as done in AdaSubS and kSubS instead of using the detector network (see Appendices D and G for more details).
+Eliminating the detector leads to a clear drop in performance in one of the four environments. HIPS-k is superior to
+kSubS in all environments despite solving a more difficult problem than kSubS. HIPS-k must learn the environment
+dynamics and a low-level policy, whereas kSubS uses a low-level search with the environment dynamics. HIPS-k also
+is superior to AdaSubS which also uses a low-level policy instead of search. The sparse reward structure and the
+required reasoning capabilities prove to be very difficult for the model-free baseline IL and offline RL methods that do
+7
+
+Table 3: The success rates of different algorithms (higher is better) and the average number of steps (lower is better)
+needed to solve TSP.
+Method
+Success rate (%)
+Avg. steps
+HIPS-PHS* (ours)
+100.0
+305.9
+HIPS-MCTS (ours)
+100.0
+213.0
+HIPS-A* (ours)
+99.9
+161.3
+kSubS
+87.9
+268.2
+CQL
+33.6
+336.9
+BC
+28.8
+339.3
+AdaSubS
+22.4
+338.9
+Teacher
+100.0
+336.5
+Oracle MCTS
+100.0
+199.9
+Christofides
+100.0
+139.0
+not rely on planning, which is why they struggle with all four tasks.
+HIPS-env performs equally to HIPS, except in Box-World, where the search exploits the inaccuracies of the
+dynamics model. However, HIPS was evaluated using open-loop planning, where one plan was generated, executed,
+and evaluated. If the agent is allowed to re-plan when the dynamics model deviates from the environment and
+fine-tune the model with the new transition, the performance would most likely increase. The issues with the dynamics
+model do not prevent HIPS from outperforming the baselines.
+Letting a search algorithm perform unlimited expansions is unrealistic in most real-world applications. Following
+Czechowski et al. (2021), we evaluate the percentage of test problems solved after N node expansions. The results are
+given in Table 2. The benefits of using RL to detect subframes become clear, as HIPS outperforms the fixed-length
+ablation HIPS-k in Sokoban and Sliding Tile Puzzle. HIPS-env outperforms the baseline methods in Sokoban and
+Travelling Salesman and is superior to kSubS on STP when the search budget is small. PHS* can solve STP and
+TSP, but cannot make enough progress in 1000 node expansions. AdaSubS cannot solve STP because it struggles to
+reliably reach the generated subgoals with its low-level policy. Therefore, kSubS outperforms AdaSubS.
+TSP is a problem where generating a successful trajectory is easy, but finding an efficient solution is much more
+difficult. Hence, we evaluate the solution lengths of the methods. We also compare the search algorithms PHS*,
+MCTS, and A* when used with HIPS. Given the amount of data, we perform VQVAE pre-training using pairs of
+subgoals instead of random pairs. In addition to the baselines capable of solving TSP, the performance of our method
+is compared to a known approximation algorithm Christofides (Christofides, 1976), to the training dataset (Teacher),
+and to an Oracle variant of subgoal-level MCTS where we replace the VQVAE generator with procedurally generated
+subgoals, where the agent is visiting one of the remaining unvisited cities.
+The solution lengths found in TSP are reported in Table 3. HIPS finds better solutions than the baselines in TSP.
+HIPS with PHS* can only slightly improve the training data, whereas HIPS-MCTS with subgoal-level rollouts can
+find significantly better solutions than the ones in the training dataset. It is inferior to the approximation algorithm,
+but the gap to the Oracle MCTS is small (around 6 %), which shows that the subgoals generated by the VQVAE are
+competitive with the procedurally generated subgoals. Finally, HIPS-A* is the best-performing agent, and the gap
+to Christofides is around 15 %. Note that our learned heuristic is non-admissible, and we trade off optimality for
+speed. kSubS can also improve on the training dataset, but it is uncompetitive against HIPS-MCTS and HIPS-A*.
+CQL, BC, and AdaSubS can make some progress on the task and solve some instances, but they cannot improve
+the solution lengths. A problem of model-free baseline methods is the inability to commit to going to a specific city,
+which highlights the benefits of goal-conditioned learning. Complete results including the standard errors can be
+found in Appendix C.
+4.4
+Visualizations
+We visualize the subgoals and plans generated by our agent to gain further understanding into the agent. An example
+of a high-level plan found by HIPS for Sokoban is visualized in Fig. 3. Fig. 4 illustrates the subgoals proposed by
+the HIPS generative model for an intermediate state in TSP. The model suggests visiting one of the unvisited cities
+as the next subgoal, which is a very reasonable planning strategy. More visualizations of the high-level trajectories
+discovered by our agent and the subgoals proposed by the generative model can be found in Appendix G.
+8
+
+Figure 3: An example of a high-level plan (a sequence of subgoals) found by HIPS in Sokoban.
+Figure 4: Examples of subgoals proposed for the current state (marked with blue boundaries) in TSP.
+5
+Conclusion
+We present a novel method for hierarchical IL that can address difficult reasoning problems that require long-term
+decision-making. Our approach relies on identifying subgoals from trajectories and generating new subgoals for
+search-based planning on new problem instances. Our method outperforms powerful search, IL, and offline RL
+baselines on the benchmark tasks. Our experiments demonstrate that our VQVAE is a suitable generative model for
+subgoal-level search and using a detector to discover subgoals has benefits over subtrajectories of fixed length.
+We see many promising directions for addressing the limitations of our method and improving it. Quantifying the
+model uncertainty could help prevent the search from exploiting the learned models. Combining discrete high-level
+planning with continuous low-level execution could make it possible to solve real-world tasks with robotics. Learning
+more abstract goals not formulated in the observation space to improve the efficiency of high-level planning and
+allowing the agent to ignore task-irrelevant sensory inputs are also left for future work. Combining low-level and
+high-level searches would improve the solution rate. Applying the learned high-level models in an RL setting to
+improve exploration or in a curriculum learning to solve progressively harder problem instances are also promising
+directions for future research.
+References
+Abbeel, P. and Ng, A. Y. (2004). Apprenticeship learning via inverse reinforcement learning. In Proceedings of the
+Twenty-First International Conference on Machine Learning, page 1.
+Ablett, T., Chan, B., and Kelly, J. (2021). Learning from guided play: A scheduled hierarchical approach for improving
+exploration in adversarial imitation learning. arXiv preprint arXiv:2112.08932.
+Argenson, A. and Dulac-Arnold, G. (2020). Model-based offline planning. arXiv preprint arXiv:2008.05556.
+Arthur, D. and Vassilvitskii, S. (2006). K-Means++: The advantages of careful seeding. Technical report, Stanford.
+Chane-Sane, E., Schmid, C., and Laptev, I. (2021). Goal-conditioned reinforcement learning with imagined subgoals.
+In International Conference on Machine Learning, pages 1430–1440. PMLR.
+Chen, D. and Kr¨ahenb¨uhl, P. (2022). Learning from all vehicles. In Proceedings of the IEEE/CVF Conference on
+Computer Vision and Pattern Recognition, pages 17222–17231.
+Chen, L., Lu, K., Rajeswaran, A., Lee, K., Grover, A., Laskin, M., Abbeel, P., Srinivas, A., and Mordatch, I. (2021).
+Decision transformer: Reinforcement learning via sequence modeling. Advances in Neural Information Processing
+Systems, 34:15084–15097.
+9
+
+XTXXXChristofides, N. (1976). Worst-case analysis of a new heuristic for the travelling salesman problem. Technical report,
+Carnegie-Mellon Univ Pittsburgh Pa Management Sciences Research Group.
+Coulom, R. (2006). Efficient selectivity and backup operators in monte-carlo tree search. In International Conference
+on Computers and Games, pages 72–83. Springer.
+Culberson, J. (1997). Sokoban is PSPACE-complete. IEICE Technical Report.
+Czechowski, K., Odrzyg´o´zd´z, T., Zbysi´nski, M., Zawalski, M., Olejnik, K., Wu, Y., Kuci´nski, �L., and Mi�lo´s, P. (2021).
+Subgoal search for complex reasoning tasks. Advances in Neural Information Processing Systems, 34:624–638.
+Daniel, C., Van Hoof, H., Peters, J., and Neumann, G. (2016). Probabilistic inference for determining options in
+reinforcement learning. Machine Learning, 104(2):337–357.
+Ebert, F., Finn, C., Dasari, S., Xie, A., Lee, A., and Levine, S. (2018).
+Visual foresight: Model-based deep
+reinforcement learning for vision-based robotic control. arXiv preprint arXiv:1812.00568.
+Fox, R., Berenstein, R., Stoica, I., and Goldberg, K. (2019). Multi-task hierarchical imitation learning for home
+automation. In 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE), pages
+1–8. IEEE.
+Garg, D., Chakraborty, S., Cundy, C., Song, J., and Ermon, S. (2021). IQ-learn: Inverse soft-Q learning for imitation.
+Advances in Neural Information Processing Systems, 34:4028–4039.
+Guez, A., Mirza, M., Gregor, K., Kabra, R., Racani`ere, S., Weber, T., Raposo, D., Santoro, A., Orseau, L., Eccles, T.,
+et al. (2019). An investigation of model-free planning. In International Conference on Machine Learning, pages
+2464–2473. PMLR.
+Hafner, D., Lee, K.-H., Fischer, I., and Abbeel, P. (2022). Deep hierarchical planning from pixels. arXiv preprint
+arXiv:2206.04114.
+Hart, P. E., Nilsson, N. J., and Raphael, B. (1968). A formal basis for the heuristic determination of minimum cost
+paths. IEEE Transactions on Systems Science and Cybernetics, 4(2):100–107.
+He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the
+IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778.
+Henderson, P., Chang, W.-D., Bacon, P.-L., Meger, D., Pineau, J., and Precup, D. (2018). Optiongan: Learning
+joint reward-policy options using generative adversarial inverse reinforcement learning. In Proceedings of the AAAI
+Conference on Artificial Intelligence, volume 32.
+Jayaraman, D., Ebert, F., Efros, A. A., and Levine, S. (2018). Time-agnostic prediction: Predicting predictable video
+frames. arXiv preprint arXiv:1808.07784.
+Jing, M., Huang, W., Sun, F., Ma, X., Kong, T., Gan, C., and Li, L. (2021). Adversarial option-aware hierarchical
+imitation learning. In International Conference on Machine Learning, pages 5097–5106. PMLR.
+Jurgenson, T., Avner, O., Groshev, E., and Tamar, A. (2020). Sub-goal trees a framework for goal-based reinforcement
+learning. In International Conference on Machine Learning, pages 5020–5030. PMLR.
+Kingma, D. P. and Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.
+Kipf, T., Li, Y., Dai, H., Zambaldi, V., Sanchez-Gonzalez, A., Grefenstette, E., Kohli, P., and Battaglia, P. (2019).
+Compile: Compositional imitation learning and execution. In International Conference on Machine Learning, pages
+3418–3428. PMLR.
+Kober, J. and Peters, J. (2008). Policy search for motor primitives in robotics. Advances in Neural Information
+Processing Systems, 21.
+Kocsis, L. and Szepesv´ari, C. (2006). Bandit based monte-carlo planning. In European Conference on Machine
+Learning, pages 282–293. Springer.
+Korf, R. E. (1985). Depth-first iterative-deepening: An optimal admissible tree search. Artificial Intelligence,
+27(1):97–109.
+10
+
+Kumar, A., Hong, J., Singh, A., and Levine, S. (2022). When should we prefer offline reinforcement learning over
+behavioral cloning? arXiv preprint arXiv:2204.05618.
+Kumar, A., Zhou, A., Tucker, G., and Levine, S. (2020). Conservative q-learning for offline reinforcement learning.
+Advances in Neural Information Processing Systems, 33:1179–1191.
+�La´ncucki, A., Chorowski, J., Sanchez, G., Marxer, R., Chen, N., Dolfing, H. J., Khurana, S., Alum¨ae, T., and Laurent,
+A. (2020). Robust training of vector quantized bottleneck models. In 2020 International Joint Conference on
+Neural Networks (IJCNN), pages 1–7. IEEE.
+Le, H., Jiang, N., Agarwal, A., Dudik, M., Yue, Y., and Daum´e III, H. (2018). Hierarchical imitation and reinforcement
+learning. In International Conference on Machine Learning, pages 2917–2926. PMLR.
+Li, J., Tang, C., Tomizuka, M., and Zhan, W. (2022).
+Hierarchical planning through goal-conditioned offline
+reinforcement learning. arXiv preprint arXiv:2205.11790.
+Mayer, H., Gomez, F., Wierstra, D., Nagy, I., Knoll, A., and Schmidhuber, J. (2008). A system for robotic heart
+surgery that learns to tie knots using recurrent neural networks. Advanced Robotics, 22(13-14):1521–1537.
+Nachum, O., Tang, H., Lu, X., Gu, S., Lee, H., and Levine, S. (2019). Why does hierarchy (sometimes) work so well
+in reinforcement learning? arXiv preprint arXiv:1909.10618.
+Nair, S. and Finn, C. (2019). Hierarchical foresight: Self-supervised learning of long-horizon tasks via visual subgoal
+generation. arXiv preprint arXiv:1909.05829.
+Orseau, L. and Lelis, L. H. (2021). Policy-guided heuristic search with guarantees. In Proceedings of the AAAI
+Conference on Artificial Intelligence, volume 35, pages 12382–12390.
+Osa, T., Pajarinen, J., Neumann, G., Bagnell, J. A., Abbeel, P., and Peters, J. (2018). An algorithmic perspective on
+imitation learning. Foundations and Trends® in Robotics, 7(1-2):1–179.
+Ozair, S., Li, Y., Razavi, A., Antonoglou, I., Van Den Oord, A., and Vinyals, O. (2021). Vector quantized models for
+planning. In International Conference on Machine Learning, pages 8302–8313. PMLR.
+Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L.,
+et al. (2019). Pytorch: An imperative style, high-performance deep learning library. Advances in Neural Information
+Processing Systems, 32.
+Paul, S., Vanbaar, J., and Roy-Chowdhury, A. (2019). Learning from trajectories via subgoal discovery. Advances in
+Neural Information Processing Systems, 32.
+Perez, E., Strub, F., De Vries, H., Dumoulin, V., and Courville, A. (2018). Film: Visual reasoning with a general
+conditioning layer. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32.
+Pertsch, K., Rybkin, O., Ebert, F., Zhou, S., Jayaraman, D., Finn, C., and Levine, S. (2020a). Long-horizon
+visual planning with goal-conditioned hierarchical predictors. Advances in Neural Information Processing Systems,
+33:17321–17333.
+Pertsch, K., Rybkin, O., Yang, J., Zhou, S., Derpanis, K., Lim, J., Daniilidis, K., and Jaegle, A. (2020b). Keyin:
+Keyframing for visual planning. Conference on Learning for Dynamics and Control.
+Racani`ere, S., Weber, T., Reichert, D., Buesing, L., Guez, A., Jimenez Rezende, D., Puigdom`enech Badia, A., Vinyals,
+O., Heess, N., Li, Y., et al. (2017). Imagination-augmented agents for deep reinforcement learning. Advances in
+Neural Information Processing Systems, 30.
+Sammut, C., Hurst, S., Kedzier, D., and Michie, D. (1992). Learning to fly. In Machine Learning Proceedings 1992,
+pages 385–393. Elsevier.
+Schrader, M.-P. B. (2018). gym-sokoban. https://github.com/mpSchrader/gym-sokoban.
+Schrittwieser, J., Antonoglou, I., Hubert, T., Simonyan, K., Sifre, L., Schmitt, S., Guez, A., Lockhart, E., Hassabis,
+D., Graepel, T., et al. (2020). Mastering atari, go, chess and shogi by planning with a learned model. Nature,
+588(7839):604–609.
+11
+
+Seno, T. and Imai, M. (2021). d3rlpy: An offline deep reinforcement learning library. arXiv preprint arXiv:2111.03788.
+Sharma, A., Sharma, M., Rhinehart, N., and Kitani, K. M. (2018). Directed-info gail: Learning hierarchical policies
+from unsegmented demonstrations using directed information. arXiv preprint arXiv:1810.01266.
+Sharma, P., Pathak, D., and Gupta, A. (2019). Third-person visual imitation learning via decoupled hierarchical
+controller. Advances in Neural Information Processing Systems, 32.
+Shoham, Y. (2021). Curry. https://festival-solver.site/curry/.
+Sutton, R. S., Precup, D., and Singh, S. (1999). Between MDPs and semi-MDPs: A framework for temporal
+abstraction in reinforcement learning. Artificial Intelligence, 112(1-2):181–211.
+Van Den Oord, A., Vinyals, O., and Kavukcuoglu, K. (2017). Neural discrete representation learning. Advances in
+Neural Information Processing Systems, 30.
+Vinyals, O., Babuschkin, I., Czarnecki, W. M., Mathieu, M., Dudzik, A., Chung, J., et al. (2019). Alphastar: Mastering
+the real-time strategy game Starcraft II. DeepMind Blog, 24.
+Williams, R. J. (1992). Simple statistical gradient-following algorithms for connectionist reinforcement learning.
+Machine Learning, 8(3):229–256.
+Ye, W., Liu, S., Kurutach, T., Abbeel, P., and Gao, Y. (2021). Mastering atari games with limited data. Advances in
+Neural Information Processing Systems, 34:25476–25488.
+Zambaldi, V., Raposo, D., Santoro, A., Bapst, V., Li, Y., Babuschkin, I., Tuyls, K., Reichert, D., Lillicrap, T.,
+Lockhart, E., et al. (2018). Relational deep reinforcement learning. arXiv preprint arXiv:1806.01830.
+Zawalski, M., Tyrolski, M., Czechowski, K., Stachura, D., Piekos, P., Odrzyg´o´zd´z, T., Wu, Y., Kuci´nski, �L., and
+Mi�lo´s, P. (2022). Fast and precise: Adjusting planning horizon with adaptive subgoal search. arXiv preprint
+arXiv:2206.00702.
+Zhang, D., Li, Q., Zheng, Y., Wei, L., Zhang, D., and Zhang, Z. (2021). Explainable hierarchical imitation learning
+for robotic drink pouring. IEEE Transactions on Automation Science and Engineering.
+Zhang, Z. and Paschalidis, I. (2021). Provable hierarchical imitation learning via EM. In International Conference on
+Artificial Intelligence and Statistics, pages 883–891. PMLR.
+12
+
+A
+Ethical Considerations
+We do not see any immediate negative societal impacts associated with our work. We do not train our models with
+any sensitive or private data, and our model is not directly applicable to, for instance, real-world decision-making
+concerning humans. However, we cannot exclude the method being applied to something harmful that is difficult to
+foresee, for instance, military purposes.
+B
+Infrastructure
+We trained our models on an HPC cluster using one NVIDIA GPU and multiple CPU workers per run. Most runs
+were performed on V100 GPUs with 32 GB of GDDR SDRAM. For some of the runs, the GPU was an A100, a
+K100, or a P80. We used 6 CPU workers per GPU with 10 GB of RAM per worker and each worker running on one
+core. By reducing the number of workers, it is possible to train and evaluate the agent on a workstation with Intel
+i7-8086K, 16 GB of RAM, and an NVIDIA GeForce GTX 1080 Ti GPU with 10 GB of video memory.
+C
+Complete Results
+Table 4: The overall success rates (%) of different algorithms including the standard errors of the means of the runs.
+The algorithms in the bottom part have access to the true environment dynamics, and those in the upper part do not.
+Method
+Sokoban
+STP
+BW
+TSP
+HIPS (ours)
+97.5 ± 0.6
+94.7 ± 1.0
+55.7 ± 2.4
+99.9 ± 0.1
+HIPS-k (ours)
+99.0 ± 0.3
+94.7 ± 1.5
+20.1 ± 1.4
+100.0 ± 0.0
+BC
+18.7 ± 0.7
+82.5 ± 2.2
+41.1 ± 1.6
+28.8 ± 8.5
+CQL
+3.3 ± 0.4
+11.7 ± 3.3
+6.0 ± 1.0
+33.6 ± 2.6
+DT
+36.7 ± 1.2
+0.0 ± 0.0
+10.0 ± 2.3
+0.0 ± 0.0
+RIS
+0.0 ± 0.0
+0.0 ± 0.0
+7.3 ± 2.2
+0.0 ± 0.0
+Option-GAIL
+0.3 ± 0.3
+0.0 ± 0.0
+0.0 ± 0.0
+0.0 ± 0.0
+IQ-Learn
+0.0 ± 0.0
+0.0 ± 0.0
+0.0 ± 0.0
+0.0 ± 0.0
+HIPS-env (ours)
+98.1 ± 0.4
+94.6 ± 1.0
+99.6 ± 0.2
+100.0 ± 0.0
+AdaSubS
+91.4 ± 0.5
+0.0 ± 0.0
+22.4 ± 2.3
+kSubS
+90.5 ± 1.0
+93.3 ± 0.9
+87.9 ± 3.8
+Table 5: The success rates (%, higher is better) of different search algorithms after performing N node expansions in
+Sokoban and STP, including the standard errors of the means of the runs.
+Sokoban
+Sliding Tile Puzzle
+N
+100
+500
+1000
+100
+500
+1000
+HIPS (ours)
+88.5 ± 0.8
+94.7 ± 0.7
+95.9 ± 0.7
+89.8 ± 1.9
+91.8 ± 1.7
+92.2 ± 1.7
+HIPS-k (ours)
+77.7 ± 1.0
+90.9 ± 0.6
+94.3 ± 0.0
+68.0 ± 3.9
+79.1 ± 3.5
+84.4 ± 1.6
+HIPS-env (ours)
+88.5 ± 0.4
+94.9 ± 0.3
+96.4 ± 0.3
+90.2 ± 1.8
+92.1 ± 1.7
+92.6 ± 1.6
+HIPS-env-k (ours)
+76.9 ± 0.8
+91.3 ± 1.5
+94.0 ± 1.6
+69.7 ± 2.2
+80.9 ± 2.9
+87.2 ± 2.3
+AdaSubS
+82.3 ± 0.5
+88.8 ± 0.4
+90.2 ± 0.5
+0.0 ± 0.0
+0.0 ± 0.0
+0.0 ± 0.0
+kSubS
+70.9 ± 2.3
+79.9 ± 2.2
+82.8 ± 1.3
+80.0 ± 3.8
+91.7 ± 1.3
+92.7 ± 1.1
+PHS*
+1.8 ± 0.1
+76.1 ± 0.5
+91.1 ± 0.3
+0.0 ± 0.0
+0.0 ± 0.0
+0.0 ± 0.0
+13
+
+Table 6: The success rates (%, higher is better) of different search algorithms after performing N node expansions in
+Box-World and TSP, including the standard errors of the means of the runs.
+Box-World
+Travelling Salesman
+N
+100
+500
+1000
+100
+500
+1000
+HIPS (ours)
+55.7 ± 2.4
+55.7 ± 2.4
+55.7 ± 2.4
+73.5 ± 13.4
+99.9 ± 0.1
+99.9 ± 0.1
+HIPS-k (ours)
+20.1 ± 1.4
+20.1 ± 1.4
+20.1 ± 1.4
+77.0 ± 2.4
+100.0 ± 0.0
+100.0 ± 0.0
+HIPS-env (ours)
+99.6 ± 0.2
+99.6 ± 0.2
+99.6 ± 0.2
+73.9 ± 13.3
+100.0 ± 0.0
+100.0 ± 0.0
+HIPS-env-k (ours)
+99.1 ± 0.2
+99.1 ± 0.2
+99.1 ± 0.2
+77.6 ± 2.2
+99.9 ± 0.1
+99.9 ± 0.1
+AdaSubS
+1.2 ± 0.6
+12.2 ± 1.6
+20.8 ± 1.2
+kSubS
+40.4 ± 11.1
+80.9 ± 6.8
+85.3 ± 5.3
+PHS*
+25.8 ± 1.0
+93.8 ± 0.7
+94.0 ± 0.7
+0.0 ± 0.0
+0.0 ± 0.0
+0.0 ± 0.0
+Table 7: The success rates of different algorithms (higher is better) and the average number of steps (lower is better)
+needed to solve TSP. The table also includes the standard errors of the means of the runs.
+Method
+Success rate (%)
+Avg. steps
+HIPS-PHS* (ours)
+100.0 ± 0.0
+305.9 ± 5.4
+HIPS-MCTS (ours)
+100.0 ± 0.0
+213.0 ± 4.4
+HIPS-A* (ours)
+99.9 ± 0.1
+161.3 ± 2.7
+kSubS
+87.9 ± 3.8
+268.2 ± 12.0
+CQL
+33.6 ± 2.6
+336.9 ± 4.1
+BC
+28.8 ± 8.5
+339.3 ± 3.9
+AdaSubS
+22.4 ± 2.3
+338.9 ± 18.6
+Teacher
+100.0 ± 0.0
+336.5 ± 0.4
+Oracle MCTS
+100.0 ± 0.0
+199.9 ± 0.7
+Christofides
+100.0 ± 0.0
+139.0 ± 0.1
+14
+
+D
+Ablation: HIPS-k
+Table 8: The success rates (%, higher is better) of different variants of HIPS in STP after performing N node
+expansions
+N
+100
+500
+1000
+∞
+HIPS-PHS*
+4.3 ± 2.9
+94.4 ± 0.8
+95.0 ± 0.9
+95.0 ± 0.9
+HIPS-GBFS
+89.8 ± 1.9
+91.8 ± 1.7
+92.2 ± 1.7
+94.7 ± 1.0
+HIPS-GBFS-3
+N/A
+N/A
+N/A
+N/A
+HIPS-GBFS-5
+68.0 ± 3.9
+79.2 ± 3.5
+84.4 ± 1.6
+94.7 ± 1.5
+HIPS-GBFS-7
+84.3 ± 0.5
+85.6 ± 0.6
+86.1 ± 0.6
+86.4 ± 0.5
+HIPS-GBFS-9
+76.5 ± 0.9
+76.5 ± 0.9
+76.5 ± 0.9
+76.5 ± 0.9
+When we replace the detector dξ with subgoals sampled at fixed intervals, we re-train the low-level policy πθ
+to achieve these subgoals. The discrete VQVAE, including the prior, are also re-trained using the new pairs of
+consecutive subgoals. The distance between the subgoals can, in this case, be controlled by a hyperparameter k. In
+Sokoban, STP, and Box-World, we used ten as the subgoal horizon and five as k (see Table 10). In TSP, selecting a
+segment length half of the subgoal horizon proved to be too much, so we let k be equal to four, the default segment
+length used in kSubS (Czechowski et al., 2021).
+We performed a small ablation study in STP to analyze the impact of the value of k. The results are shown in
+Table 8. We see that given a large research budget, PHS* is slightly superior to GBFS, but GBFS outperforms PHS*
+given a small search budget. HIPS-GBFS-3 doesn’t converge because the value function is noisy. Using a larger k
+allows the value function to ”leap over” the noise, as observed by Czechowski et al. (2021). We see that using a
+larger k improves the percentage of puzzles solved after a smaller number of expansions, but hurts the overall solution
+rate, as the search space is explored less systematically. However, using a k too large is also harmful as training
+the generative model becomes difficult. Furthermore, no value of k was able to outperform HIPS-GBFS, which
+highlights the benefits of our method that can propose subgoals at different distances adaptively in all environments
+(see Figure 5).
+Sokoban
+STP
+Box-World
+TSP
+Figure 5: Lengths of the subtrajectories to reach subgoals proposed by VQVAE when it has been trained using the
+detector dξ.
+15
+
+0.05
+0.04
+0.03
+0.02
+0.01
+0.00
+10
+20
+30
+40
+500.12
+0.10
+0.08
+0.06
+0.04-
+0.02
+0.00
+0
+5
+10
+15
+20
+250.175
+0.150
+0.125
+0.100
+0.075
+0.050
+0.025
+0.000
+0.0
+2.5
+5.0
+7.5
+10.0
+12.5
+15.0
+17.5
+20.00.10
+0.08
+0.06
+0.04 -
+0.02
+0.00
+0
+5
+10
+15
+20
+25
+30E
+Search Methods
+Greedy Best-First Search (GBFS)
+is a priority queue -based search algorithm, where the evaluation function
+has been defined as
+ϕ(n) = h(n),
+where h(n) is a heuristic that predicts the distance to the goal. The node that is predicted to be the closest to the
+goal is expanded.
+Policy-Guided Heuristic Search (PHS)
+is a policy-guided search algorithm (Orseau and Lelis, 2021) which
+uses a priority queue with the evaluation function
+ϕ(n) = η(n)g(n)/π(n),
+where g(n) is the path cost from the root to node n, π(n) is the node policy (the probability of selecting node n) and
+η(n) is a heuristic factor whose purpose is to estimate the cost to the nearest descendant solution node. We use a
+variant of PHS, PHS* where the heuristic factor has been defined as
+ηh(n) = 1 + h(n)/g(n)
+π(n)h(n)/g(n) .
+(4)
+A*
+is a heuristic-based search algorithm that tries to find the shortest path to the goal (Hart et al., 1968). It is
+based on a priority queue with the evaluation function
+ϕ(n) = g(n) + h(n),
+where g(n) is the distance from the root to node n and h(n) is a heuristic that predicts the distance from the node
+n to the goal. If the heuristic h(n) never overestimates the true distance to the goal, the heuristic is said to be
+admissible, and A* is guaranteed to find the shortest path.
+Monte Carlo Tree Search (MCTS)
+is a tree-based search method based on expanding a search tree by performing
+Monte Carlo evaluations. The selection of nodes for expansion is biased towards promising nodes to enable MCTS to
+focus on the relevant parts of the search tree. The most commonly used method is UCT, where nodes with higher
+rewards and lower visitation frequency get the highest priority (Ozair et al., 2021; Kocsis and Szepesv´ari, 2006).
+Listing 1 contains the pseudo-code describing our high-level search with a priority queue. Subroutine phs cost
+calculates the value of the heuristic factor η(n) for PHS* as in (4). Subroutine extract plan collects the subgoals on
+the path to the leaf node (terminal state) from the root (initial state). Subroutine get distances takes as input the
+current state and the proposed children and tries to reach them using the subgoal-conditioned low-level policy πθ.
+The distances between the state and the children are recovered simultaneously.
+16
+
+Listing 1 PyTorch pseudocode for the high-level search
+1
+def
+get_priority (node , alg):
+2
+if alg == ’phs_star ’:
+3
+return
+phs_cost(node.log_p , node.value , node.cum_dist)
+4
+elif
+alg == ’gbfs ’:
+5
+return
+node.value
+6
+elif
+alg == ’a_star ’:
+7
+return
+node.value + node.cum_dist
+8
+9
+10
+def
+init_node(node , alg , vqvae , policy , value_func , dynamics):
+11
+node.value = value_func(node.state)
+12
+node. child_states = vqvae.generate(node.state)
+13
+node. distances_to_children = get_distances (
+14
+node.state ,
+15
+node.child_states ,
+16
+policy ,
+17
+dynamics)
+18
+node. filter_unreachable_children ()
+# Uses
+the
+distances
+computed
+19
+node. children_log_probs = vqvae.prior(node.state)
+20
+node.priority = get_priority (node , alg)
+21
+22
+23
+def
+search(state , alg , vqvae , policy , value_func , dynamics):
+24
+n_nodes = 0
+25
+queue = PriorityQueue ()
+# Create
+empty
+priority
+queue
+26
+expanded = Set ()
+# Create
+empty
+set
+27
+node = Node(
+28
+state ,
+29
+parent=None ,
+30
+cum_dist =0,
+31
+log_p =0,
+32
+)
+33
+init_node(node , alg , vqvae , policy , value_func , dynamics)
+34
+queue.insert(node)
+35
+36
+while
+len(queue) > 0:
+37
+node = queue.pop ()
+38
+expanded.add(node.state)
+39
+n_nodes
++= 1
+40
+for
+c_state , c_dist , c_log_p
+in zip(node.child_states ,
+41
+node. distances_to_children ,
+42
+node. children_log_probs ):
+43
+if
+c_state
+in
+expanded:
+44
+continue
+45
+new_node = Node(
+46
+c_state ,
+47
+parent=node ,
+48
+cum_dist=node.cum_dist + c_dist ,
+49
+log_p=node.log_p + c_log_p ,
+50
+)
+51
+if
+is_terminal (c_state):
+52
+return
+extract_plan (new_node), n_nodes
+# Success
+53
+init_node(new_node , alg , vqvae , policy , value_func , dynamics)
+54
+queue.insert(new_node)
+55
+return
+None , n_nodes
+# Search
+failed , queue
+empty
+17
+
+F
+VQVAE Training
+Algorithm 2 Training VQVAE for Subgoal Generation
+Input: A dataset of trajectories D, a dataset of subgoal pairs D′, untrained encoder fφ, decoder gψ, codebook {ek}
+Parameters: The codebook {ek} and the parameters of the encoder, φ, and the decoder, ψ
+Output: Trained encoder fφ, decoder gψ, codebook {ek}.
+1: while fφ and gψ not converged do
+2:
+Sample a trajectory τ from D
+3:
+For each state si ∈ τ, uniformly sample a pair sj from (si+1, . . . , sH)
+4:
+Reconstruct sj without discretization: ˆsj = gψ(fφ(sj, si), si)
+5:
+Compute reconstruction loss Lrec(ˆsj, sj)
+6:
+Update φ, ψ to minimize reconstruction loss
+7: end while
+8: Sample subsequent subgoal pairs sgj, sgj−1 from D′ and encode them with the encoder: zj = fφ(sgj, sgj−1)
+9: Initialize {ek} as the clusters centers obtained by running KMeans++ on the encodings {zj}
+10: while {ek}, fφ and gψ not converged do
+11:
+Sample a batch of subsequent subgoal pairs sgj, sgj−1 from D′
+12:
+Reconstruct subgoals with VQVAE
+zj = fφ(sgj, sgj−1)
+kj = arg min
+m
+∥zj − em∥2
+ˆsgj = gψ(ekj, sgj−1)
+13:
+Compute the loss in (3)
+14:
+Update φ, ψ, {ek} to minimize the loss computed in Step 13.
+15: end while
+16: return fφ, gψ, {ek}
+G
+Hyperparameters and Visualizations
+Table 9: General hyperparameters of our method.
+Parameter
+Value
+Learning rate for dynamics
+2 · 10−4
+Learning rate for π, d, V
+1 · 10−3
+Learning rate for VQVAE
+2 · 10−4
+Discount rate for REINFORCE
+0.99
+Table 10: Environment-specific hyperparameters of our method.
+Parameter
+Explanation
+Sokoban
+STP
+Box-World
+TSP
+α
+Subgoal penalty
+0.1
+0.1
+0.1
+0.05
+β
+Beta for VQVAE
+0.1
+0.1
+0.1
+0
+c
+Exploration constant for MCTS
+–
+–
+–
+0.1
+D
+Codebook dimensionality
+128
+128
+128
+64
+H
+Subgoal horizon
+10
+10
+10
+50
+K
+VQVAE codebook size
+64
+64
+64
+32
+k
+Segment length w/o REINFORCE
+5
+5
+5
+4
+(N, D)
+DRC size
+–
+–
+(3, 3)
+–
+18
+
+(a) A subgoal-level plan.
+(b) Examples of generated subgoals.
+Figure 6: Visualization of the solution found by HIPS for a Sokoban problem. (a): A subgoal-level plan found by
+HIPS. (b): Subgoals proposed for an intermediate state (marked with blue boundaries). The subgoals have been
+sorted according to the prior probabilities. The subgoal selected for the final plan is marked with red boundaries.
+19
+
+XTX区
+XXX
+XX
+区
+XXXXIXXIXXXXTXX
+XXXXXIXXIXXTXXIX(a) A subgoal-level plan.
+(b) Examples of generated subgoals.
+Figure 7: Visualization of the solution found by HIPS for an STP problem. (a): A subgoal-level plan found by
+HIPS. (b): Subgoals proposed for intermediate states (marked with blue boundaries). The subgoals have been sorted
+according to the prior probabilities. The subgoal selected for the final plan is marked with red boundaries. Red color
+is used to highlight the tiles which are different from the reference state (the previous state in (a) and the current
+state in (b)).
+20
+
+20
+23
+3
+1
+2
+13
+19
+5
+15
+21
+9
+11
+4
+16
+7
+18
+14
+12
+8
+10
+22
+24
+6
+1720
+3
+13
+12
+5
+2
+9
+4
+16
+1
+21
+19
+8
+11
+15
+7
+14
+6
+17
+10
+18
+22
+23
+243
+13
+12
+5
+20
+9
+4
+16
+1
+2
+7
+8
+11
+15
+19
+18
+14
+6
+17
+21
+10
+22
+23
+243
+9
+13
+12
+5
+20
+7
+4
+16
+1
+2
+8
+11
+15
+19
+18
+14
+6
+17
+21
+10
+22
+23
+243
+9
+13
+12
+5
+7
+2
+4
+16
+1
+18
+19
+8
+11
+15
+20
+14
+6
+17
+21
+10
+22
+23
+243
+9
+13
+12
+5
+7
+2
+4
+16
+1
+18
+8
+14
+11
+15
+10
+19
+6
+17
+20
+21
+22
+23
+243
+9
+13
+12
+5
+7
+2
+4
+16
+1
+18
+14
+6
+15
+17
+10
+8
+11
+19
+20
+21
+22
+23
+243
+9
+13
+12
+5
+7
+2
+4
+16
+1
+18
+8
+14
+15
+17
+10
+6
+11
+19
+20
+21
+22
+23
+243
+9
+13
+12
+5
+7
+2
+16
+1
+18
+8
+4
+14
+17
+10
+6
+11
+15
+19
+20
+21
+22
+23
+243
+2
+9
+12
+5
+7
+13
+16
+17
+18
+8
+4
+1
+14
+10
+6
+11
+15
+19
+20
+21
+22
+23
+243
+2
+16
+17
+9
+7
+13
+12
+5
+18
+8
+4
+1
+14
+10
+6
+11
+15
+19
+20
+21
+22
+23
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+6
+8
+10
+22
+14
+24
+173
+2
+17
+4
+9
+7
+13
+16
+5
+18
+8
+1
+12
+14
+10
+6
+11
+15
+19
+20
+21
+22
+23
+243
+2
+5
+16
+4
+7
+13
+17
+12
+9
+18
+8
+1
+14
+10
+6
+11
+15
+19
+20
+21
+22
+23
+243
+2
+5
+16
+4
+7
+8
+13
+12
+9
+18
+17
+1
+14
+10
+6
+11
+15
+19
+20
+21
+22
+23
+243
+2
+5
+16
+4
+7
+8
+13
+12
+9
+10
+11
+1
+15
+14
+6
+17
+18
+19
+20
+21
+22
+23
+243
+2
+5
+16
+4
+7
+1
+8
+13
+9
+10
+15
+12
+14
+6
+11
+17
+18
+19
+20
+21
+22
+23
+243
+2
+8
+5
+4
+7
+1
+16
+13
+9
+10
+15
+12
+14
+6
+11
+17
+18
+19
+20
+21
+22
+23
+243
+2
+8
+5
+4
+15
+10
+7
+13
+9
+16
+1
+12
+14
+6
+11
+17
+18
+19
+20
+21
+22
+23
+243
+2
+8
+5
+4
+10
+11
+7
+13
+9
+6
+1
+12
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+243
+2
+8
+4
+10
+11
+7
+5
+9
+6
+1
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+2410
+3
+8
+4
+11
+7
+2
+5
+9
+6
+1
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2410
+7
+3
+4
+11
+2
+5
+8
+9
+6
+1
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+242
+7
+3
+4
+10
+1
+5
+8
+9
+11
+6
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+242
+1
+7
+3
+4
+10
+6
+5
+8
+9
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+245
+1
+3
+4
+2
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+245
+7
+3
+4
+2
+1
+6
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+241
+2
+3
+4
+5
+6
+7
+8
+9
+10
+11
+12
+13
+14
+15
+16
+17
+18
+19
+20
+21
+22
+23
+2420
+23
+3
+5
+1
+2
+13
+4
+15
+21
+9
+19
+12
+16
+7
+18
+11
+8
+17
+10
+22
+6
+14
+2420
+3
+13
+5
+1
+2
+9
+4
+15
+21
+19
+23
+12
+16
+7
+18
+11
+8
+17
+10
+22
+6
+14
+2420
+3
+13
+5
+1
+2
+9
+4
+12
+15
+21
+19
+11
+23
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2420
+3
+13
+5
+1
+2
+9
+4
+12
+21
+19
+8
+16
+15
+7
+18
+6
+11
+17
+10
+22
+14
+23
+2420
+3
+13
+12
+5
+2
+9
+4
+1
+21
+19
+8
+16
+15
+7
+18
+6
+11
+17
+10
+22
+14
+23
+2420
+3
+13
+12
+5
+2
+9
+4
+16
+1
+21
+19
+8
+11
+15
+7
+18
+6
+17
+10
+22
+14
+23
+2420
+23
+3
+1
+2
+13
+19
+5
+15
+21
+9
+11
+4
+16
+7
+18
+14
+12
+8
+10
+22
+24
+6
+1720
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+6
+8
+10
+22
+14
+24
+1720
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+14
+8
+10
+22
+24
+6
+1720
+23
+19
+3
+1
+2
+13
+11
+5
+15
+21
+9
+4
+12
+16
+7
+18
+6
+8
+10
+22
+14
+24
+1720
+23
+3
+1
+2
+13
+19
+5
+15
+21
+9
+11
+4
+16
+7
+18
+14
+12
+8
+10
+22
+24
+6
+1720
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+6
+8
+10
+22
+14
+24
+1720
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+6
+8
+10
+22
+14
+24 1720
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+6
+8
+17
+10
+18
+22
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+6
+17
+10
+22
+14
+8
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+6
+8
+17
+10
+18
+22
+14
+2420
+3
+19
+5
+1
+2
+23
+4
+15
+21
+13
+9
+12
+16
+7
+18
+11
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+4
+15
+21
+9
+19
+12
+16
+7
+18
+11
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2420
+3
+5
+1
+2
+23
+19
+4
+15
+21
+13
+9
+12
+16
+7
+18
+11
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+9
+11
+12
+16
+7
+18
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+21
+6
+12
+16
+7
+18
+11
+8
+17
+10
+22
+6
+14
+2420
+23
+3
+5
+1
+2
+13
+19
+4
+15
+6
+11
+12
+16
+21
+7
+18
+8
+17
+10
+22
+6
+14
+24(a) A subgoal-level plan.
+(b) Examples of generated subgoals.
+Figure 8: Visualization of the solution found by HIPS for a BW problem. (a): A subgoal-level plan found by HIPS.
+(b): Subgoals proposed for an intermediate state (marked with blue boundaries). The subgoals have been sorted
+according to the prior probabilities. The subgoal selected for the final plan is marked with red boundaries.
+21
+
+10
+0?
+180?
+180
+0?
+10?
+1818
+180
+0?
+100?
+1800
+0?
+10
+100
+0??
+180?
+180?
+瘤
+10
+180?
+180
+0?
+1010
+0?
+180?
+18(a) A subgoal-level plan.
+(b) Examples of generated subgoals.
+Figure 9: Visualization of the solution found by HIPS for a TSP instance. (a): A subgoal-level plan found by HIPS.
+(b): Subgoals proposed for an intermediate state (marked with blue boundaries).
+22
+
diff --git a/WNFOT4oBgHgl3EQf7TSv/content/tmp_files/load_file.txt b/WNFOT4oBgHgl3EQf7TSv/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..c380e7f2736514044dc586011209dd2cad1411f0
--- /dev/null
+++ b/WNFOT4oBgHgl3EQf7TSv/content/tmp_files/load_file.txt
@@ -0,0 +1,2780 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf,len=2779
+page_content='Hierarchical Imitation Learning with Vector Quantized Models Kalle Kujanp¨a¨a1, Joni Pajarinen2, and Alexander Ilin1 1Department of Computer Science, Aalto University, Finland 2Department of Electrical Engineering and Automation, Aalto University, Finland Abstract The ability to plan actions on multiple levels of abstraction enables intelligent agents to solve complex tasks effectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, learning the models for both low and high-level planning from demonstrations has proven challenging, especially with higher-dimensional inputs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' To address this issue, we propose to use reinforcement learning to identify subgoals in expert trajectories by associating the magnitude of the rewards with the predictability of low-level actions given the state and the chosen subgoal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We build a vector-quantized generative model for the identified subgoals to perform subgoal-level planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In experiments, the algorithm excels at solving complex, long-horizon decision-making problems outperforming state-of-the-art.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Because of its ability to plan, our algorithm can find better trajectories than the ones in the training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1 Introduction Learning from expert demonstrations has proven successful in many sequential decision-making settings that can be modeled with Markov decision processes (Abbeel and Ng, 2004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Imitation learning (IL) is a technique for learning to imitate the behavior of an expert by discovering the mapping between states and actions without access to information such as rewards (Osa et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IL has proven useful in aviation (Sammut et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1992), autonomous driving (Chen and Kr¨ahenb¨uhl, 2022), robotics (Kober and Peters, 2008), video games (Vinyals et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019) and even healthcare (Mayer et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Recent advances in planning with learned dynamics models have improved our ability to solve complex long-horizon problems when interacting with the environment is possible (Hafner et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Ye et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Schrittwieser et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning models for planning in the offline setting is possible as well (Argenson and Dulac-Arnold, 2020), but these model-based reinforcement learning (RL) methods assume access to environment rewards and are not directly applicable to the IL setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical RL with decision-making at multiple time scales has succeeded in tasks where flat RL struggles (Hafner et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Nachum et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A hierarchy should also be useful in the IL setting as many real-world problems have a natural hierarchical structure (Sharma et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Jing et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Moreover, planning with a hierarchy may shorten the effective planning horizon and avoid compounding model errors also in the IL setting (Nair and Finn, 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We propose a method for hierarchical planning in the IL setting that relies on segmenting expert trajectories into subtasks without any high-level supervision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Unlike prior work that assumes a fixed number of subgoals (Pertsch et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020b), fixed-length subtasks (Czechowski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021), or trains multiple models to deal with subtasks of different lengths (Zawalski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022), our algorithm segments the trajectories into a variable number of variable-length subtasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use the segmentation to learn a generative model over the subgoals and a subgoal-conditioned low-level policy to execute the subtasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' To perform high-level planning, we use standard search methods such as Policy-Guided Heuristic Search (Orseau and Lelis, 2021), Monte Carlo Tree Search (Coulom, 2006), or A* (Hart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1968) in which our generative model is used for node expansion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our method outperforms strong search, hierarchical IL, and offline RL algorithms at complex long-horizon decision-making problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our experiments also show that our algorithm can handle suboptimal expert trajectories and self-improve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In summary, the main contributions of this work are: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A novel yet conceptually simple RL approach for identifying subgoals from trajectories based on the prediction performance of a low-level policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A VQVAE generative model for proposing subgoals for planning with temporal abstraction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In experiments, our generative model combined with the learned low-level policy and a suitable high-level search algorithm solves complex problems with sparse rewards better and with fewer node expansions than state-of-the-art subgoal search and outperforms offline RL algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12962v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='AI] 30 Jan 2023 2 Related Work Our method combines hierarchical discrete planning with imitation learning to solve complex planning problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical planning allows shortening the effective planning horizon, which is beneficial in long-horizon tasks (Pertsch et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Nair and Finn, 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use offline data to learn a VQVAE (Van Den Oord et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2017) that generates adaptive horizon subgoals and a low-level policy to reach the subgoals for high-level planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Previous work proposes a wide variety of different approaches for learning hierarchical behavior from data, also in combination with planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, according to our knowledge, our method is the first method that segments a trajectory into a varying number of adaptive-length subtrajectories and uses the subgoals to learn a generative model that can solve difficult long-term decision-making problems in a discrete setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical Continuous Planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In hierarchical planning with continuous control, the Cross-Entropy Method (CEM) is typically used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' With continuous controls, search methods with convergence guarantees such as MCTS (Coulom, 2006), PHS∗ (Orseau and Lelis, 2021), or A∗ (Hart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1968), utilized with our method in the experiments, cannot be directly applied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HiGoC (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022) uses CEM for hierarchical planning in offline RL assuming access to rewards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HVF (Nair and Finn, 2019) adds a hierarchical structure with predicted subgoal images to Visual Model Predictive Control (Ebert et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018) that plans in image space using CEM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Visual hierarchical methods that rely on identifying keyframes from trajectories using variational inference and planning with CEM include TAP (Jayaraman et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018) and KeyIn (Pertsch et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, these methods assume a fixed number of keyframes in the trajectories and have not been successfully applied to highly complex reasoning tasks with sparse rewards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' KeyIn also uses an environment simulator for planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Goal-Conditioned Hierarchical Planning (Pertsch et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020a) produces plans, executed by a learned inverse dynamics model, in a high-dimensional state space in an offline setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' SGT-PT does goal-based RL in a low- dimensional setting by planning with subgoal trees (Jurgenson et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, these models require an explicit goal state and need to generate subgoals between the initial state and the goal state, which can be a difficult learning problem in complex long-horizon environments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical IL without Planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Unlike our method, many model-free hierarchical imitation learning methods assume some degree of high-level supervision (Le et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Fox et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' As an alternative Daniel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2016) infer compositional structure in data discovering options (Sutton et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1999) (a set of low-level policies with termination) with expectation-maximization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' CompILE (Kipf et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019) uses VAEs to segment trajectories into subtasks and the subtask encodings as subpolicies in hierarchical RL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The model-free method Option-GAIL (Jing et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021) infers expert options from trajectories with an EM-like approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Zhang and Paschalidis (2021) directly optimize a hierarchical policy with options by maximizing the probability of expert trajectories with a hidden Markov model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Directed-Info GAIL (Sharma et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018) is a variant of hierarchical inverse RL that learns latent policies by modeling problems as directed graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The method segments expert trajectories into sub-tasks and learns structural policies to solve different sub-tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' OptionGAN (Henderson et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018) learns to recover reward and policy options simultaneously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning from Guided Play (LfGP) (Ablett et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021) uses scheduled auxiliary tasks to address lacking exploration in adversarial online IL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Paul et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019) learn a generative model over subgoals from demonstrations and use it to augment the reward function for RL fine-tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, these methods do not incorporate high-level planning mechanisms, which may make them unsuitable for solving complex reasoning problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Offline RL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In offline RL, the agent’s objective is to learn an optimal policy without interacting with the environment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Instead, the agent has access to a dataset of transitions that have been collected by a behavior policy πβ that can be suboptimal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A significant benefit that offline RL has over imitation learning is the ability to extract strong policies even when the expert trajectories are suboptimal (Kumar et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Conservative Q-learning (Kumar et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020) is a model-free offline RL method that learns a lower bound on the policy value, which helps avoid overestimating state values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Decision Transformer (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021) treats the offline reinforcement learning task as a sequence modeling problem, where the goal is to predict the action conditioned on a desired reward.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use offline RL methods as baselines in our experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical IL with Discrete Search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The closest works to ours in search and planning are kSubS (Czechowski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021) and AdaSubS (Zawalski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' kSubs learns an autoregressive model for generating subgoals given a set of trajectories and uses subgoal search for planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Unlike our method, kSubS relies on the true environment dynamics for low-level search to find fixed-length subtrajectories between subgoals in some environments when combined with an autoregressive CNN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' AdaSubS replaces the low-level search of kSubS with a learned policy in all environments and supports subtrajectories of multiple hard-coded lengths by training an autoregressive generative model for each length.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our approach relies on a learned low-level policy and dynamics model and supports varying- length segments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We show that it is possible to train a single non-autoregressive adaptive generative model, which makes generating subgoals more efficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 2 Trajectory Segmentation Low-Level Policy πθ(a | s, sg) Subgoals Generative Model for Subgoals VQVAE VQVAE Low-Level Policy πθ(a | s, sg) Figure 1: A visualization of our Hierarchical Imitation Planning with Search (HIPS) when learning to solve Sokoban.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The main components of our method are a detector dξ(sgk|si) that segments the trajectory into subgoals, a subgoal- conditioned low-level policy πθ(ai|si, sgk), and the VQVAE, a generative model over the subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The low-level policy and VQVAE are used during evaluation for planning, whereas the detector is training-only.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3 Method We consider goal-oriented, complex reasoning problems, in which the agent’s objective is to act in the environment to reach a terminal state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' This corresponds to a Markov decision process with a reward of one upon solving the task and zero otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We consider environments with fully Markovian, discrete-valued states with full observability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We work in the imitation learning (offline) setting: the agent needs to learn to solve tasks only from available demonstrations without the possibility to interact with the environment before evaluation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We assume that there is a dataset D of trajectories τ = {s0, a0, s1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' , aT −1, sT } collected by experts who know how to solve tasks, with only a reward of one at the terminal state sT .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The experts may not reach the terminal states in the fastest possible way.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We propose to solve the imitation learning task using an agent which has a hierarchical structure with two levels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our agent learns a hierarchical representation of the available trajectories by identifying likely experts’ subgoals in the existing trajectories in an unsupervised manner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The identified subgoals are then used to train a discrete-code generative model which can generate reasonable subgoals to perform subgoal-level planning with standard search algorithms such as PHS, MCTS, or A*.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We also learn a low-level policy that executes the plan generated by the planner by sequentially reaching the determined subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A graphical representation of our method is shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We call our method Hierarchical Imitation Planning with Search (HIPS).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Below we describe the main components of the approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 Subgoal Identification The goal of the subgoal identification phase is to learn a high-level representation τ∗ = {sg1, sg2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' , sgM } of each trajectory such that the trajectory is represented as a sequence of subgoals sgk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Each subgoal is a state from the trajectory, that is ∀ksgk ∈ τ, and in particular, sgM = sT .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' This is a time series segmentation problem where the solution has two desirable qualities: 1) we want the identified subgoals sgk to be easy to reach by a trainable low-level policy πθ(ai|si, sgk) which takes the subgoal sgk as input and 2) we want the subgoals to be easy to sample from a generative model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Successfully segmenting the trajectories into a variable number of variable-length segments turns out to be a highly non-trivial task, as most prior work has focused on fixed-length segments or a fixed number of segments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our solution is to formulate this segmentation task as an RL problem in which we treat each trajectory from D as one episode for training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In each episode, the segmentation agent starts at the first state s0 of the trajectory and selects the next subgoal state sg1 according to the probabilities produced by its policy dξ(sg1|s0), which we call the detector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We limit the detector to consider only the following H states as candidate subgoals, that is sg1 is selected from s1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', sH.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The 3 区 区区 区区区 区区区 + 区区区 区Algorithm 1 Segmenting Trajectories for Hierarchical IL Input: A dataset of trajectories D, untrained low-level policy network πθ, detector dξ Parameters: The parameters of the low-level policy network, θ, detector network, ξ Output: Trained low-level policy π, dataset D′ of subgoal pairs {(sgk+1, sgk)} 1: while πθ, dξ not converged do 2: Sample a trajectory τ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3: Segment τ with dξ 4: Predict low-level actions with πθ conditioned on produced subgoals 5: Compute returns (Equation 1), update ξ with REINFORCE 6: Compute the losses for πθ (Equation 2), update θ 7: end while 8: Create a dataset D′ of subgoal pairs {(sgk+1, sgk)} by sampling trajectories τ from D and segmenting them with dξ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 9: return πθ, D′ agent samples the next subgoal according to the computed probabilities and gets the reward R1 = r1 − α, (1) where r1 is the log-probability that a low-level policy πθ(ai|si, sg1) selects the sequence of actions a0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', ag1−1 in the first segment: r1 = g1−1 � i=0 log πθ(ai|si, sg1) and α is a penalty to prevent segmentation into too many subtrajectories.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' After that, the segmentation agent changes its state to sg1, selects the next subgoal according to dξ(sg2|sg1) and gets reward R2 computed using action log-probabilities from the second segment, similarly to (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The episode continues like this until the end of the trajectory is reached.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The low-level policy πθ(ai|si, sgk) is considered as part of the environment of the segmentation agent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' It is updated during training using goal-conditioned behavioral cloning to minimize Lθ = −Eτ∼D M(τ) � k=1 −1+gk � i=gk−1 log πθ(ai|si, sgk), (2) with subgoals sgk produced by the segmentation agent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Note that the number of identified subgoals M(τ) may vary across trajectories.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Thus, the segmentation agent is trained by giving it a higher reward when selected subgoals lead to more accurate action predictions by the low-level policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The low-level policy is trained concurrently with the subgoals (high-level commands) produced by the segmentation agent as input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We train the segmentation agent using the policy gradient algorithm REINFORCE with a learnable value function as baseline (Williams, 1992).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Note that using a trainable detector dξ(sgk+1|sgk) encourages subgoals that are easy to recognize among the states in the training trajectories, and therefore, such subgoals might be easy to produce by a learned generative model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our approach for segmenting trajectories is summarized in Algorithm 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 Generative Model for Subgoals To plan in terms of the high-level subgoals, the agent needs the ability to generate reasonable subgoals sg for each environment state s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We do this by learning a generative model p(sg|s) over the subgoals using the ones identified in the trajectory segmentation step as training data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We implement the model as a VQVAE with discrete latent codes, which is inspired by the vector quantized models proposed by Ozair et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The VQVAE encoder takes a pair of states (sg, s) as input and outputs a continuous latent code ze.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Then, the code is quantized by finding the nearest code ek from the codebook such that k = arg minm ∥ze − em∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The decoder uses the code ek to reconstruct the subgoal state ˆsg = gψ(ek, s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The loss minimized during training is L = Lrec(ˆsg, sg) + ∥[ze] − ek∥2 2 + β∥ze − [ek]∥2 2 , (3) 4 Sokoban Sliding Tile Puzzle Box-World Traveling Salesman Problem Figure 2: The environments we consider in the experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sokoban: the task is to push yellow boxes onto the target locations (marked with red squares).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sliding Tile Puzzle (STP): The task is to order the tiles from 1 to 24 by moving them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Box-World: The agent must collect colored keys and open color-matching locks to recover more keys until it finally reaches a goal target (marked with the $ sign).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Traveling salesman problem (TSP): The agent (marked with the circle) has to visit all cities (marked with squares) before returning to the start (marked with the black square).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Visited cities are marked with green squares and unvisited ones with red squares.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' where Lrec is the reconstruction loss and [·] denotes the stop gradient operation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We train the VQVAE in two stages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In the pre-training stage, we use random pairs of states (sj, si), i < j ≤ i + H from the trajectories as inputs (sg, s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We skip the discretization layer and only use the reconstruction loss Lrec(ˆsg, sg) for training the encoder and the decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' After the pre-training has converged, the complete VQVAE is trained by using pairs of consecutive subgoals (sgk+1, sgk) as input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We initialize the codebook by running KMeans++ clustering (Arthur and Vassilvitskii, 2006) on the first batches of encoder outputs and using the cluster centers as the initial codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' This training strategy was inspired by the strategy proposed by �La´ncucki et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Finally, we learn a subgoal-conditioned prior p(ek+1|sgk) over the latent codes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Once the model has been trained, one can generate subgoals conditioned on the current state s by sampling a code e from the learned codebook and running it through the decoder gψ(e, s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Note that the number of possible codes e is finite, which means that the number of generated subgoals sg is finite as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Note also that distinct codes e may result in the same generated subgoal sg, which is a desired behavior when the size of the codebook is larger than the number of reasonable subgoals for the considered state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The pseudo-code for our VQVAE training is given in Algorithm 2 in the Appendix F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 High-Level Planning with Search We perform planning in the subgoal space, as it can be more efficient and suitable for long-horizon planning than planning in the state space (Nair and Finn, 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We demonstrate that our method is compatible with many different search algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Each search node represents a subgoal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' When a search node is expanded, possible next subgoals (child nodes) are generated with the VQVAE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Given a codebook of size K, there are K possible child nodes for each subgoal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' To limit the size of the search tree, we remove duplicates and unreachable subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The reachability of a proposed subgoal sg from state si is evaluated by using the low-level policy πθ(ai|si, sg) trained in the segmentation phase.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We run the policy iteratively from state si and simulate state transitions by using the true dynamics or a learned model fdyn(si+1|ai, si).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' If the subgoal sg is reached within a specific number of steps, the subgoal is considered reachable, and a search node is created.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In this work, we work with discrete states and require an exact match between the reached state and the subgoal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Using a suitable threshold to evaluate the match is an alternative in the continuous setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We also learn a value function V (s) that predicts the number of low-level steps necessary to reach the goal (terminal state) and use it as a heuristic in planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The search methods we use are Greedy Best-First Search (GBFS), Policy-Guided Heuristic Search (PHS*, Orseau and Lelis, 2021), A* (Hart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1968) and Monte-Carlo Tree Search (MCTS, Coulom, 2006;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kocsis and Szepesv´ari, 2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PHS* is dependent on a good policy, but when the dataset contains suboptimal trajectories, learning a good VQVAE prior to act as the policy might be impossible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Then, policy-independent algorithms like A* or GBFS can be superior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PHS* is also aimed at minimizing the search loss, not finding a particularly high-quality solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' When that is important, A* or MCTS can be superior to PHS*.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 5 3 9 13 12 5 7 2 4 16 1 X 18 8 14 11 15 10 19 6 17 XX + + X 20 21 22 23 2410 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 10?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1818 184 Experiments In our experimental phase, we evaluate our method on complex, sparse reward problems that require reasoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We compare our method to existing search, hierarchical imitation learning, and offline RL methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We also analyze whether our RL-based approach for identifying subgoals is superior to subgoals sampled at fixed intervals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 Environments We evaluate our method in four environments that are all complex reasoning domains (see Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The first environment is Sokoban, which is a PSPACE-complete puzzle where the agent must push boxes onto goal locations (Culberson, 1997).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The moves are irreversible and one wrong push can make the puzzle unsolvable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use a 10 × 10 problem size with four boxes, the default configuration in the earlier literature (Orseau and Lelis, 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Guez et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Racani`ere et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use a one-hot encoded tensor with shape 10 × 10 × 4 as the observation space (Orseau and Lelis, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The second environment is the sliding tile puzzle (STP) which is a classic benchmark in the search literature (Korf, 1985).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use a puzzle size of 5 × 5, and the objective is to sort the number tiles in a specific descending order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The third environment is Box-World (BW), where the agent must collect colored keys and open color-matching locks to recover more keys until it finally reaches a goal target (Zambaldi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Keys can only be used once.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' If the agent uses its key to open the wrong box, the game will become unsolvable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hence, careful planning and reasoning about entities and their relations are required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The fourth environment is a grid-based Traveling Salesman Problem (TSP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The agent moves through the 2D grid to visit all the cities before returning to the starting point (see Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The TSP is an NP-hard combinatorial optimization problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, finding any solution to TSP is relatively easy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hence, grid-based TSP serves as an environment for evaluating the solution quality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Sokoban, we collect a training set of 10340 trajectories using gym-sokoban (Schrader, 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 10240 of the problem instances have been solved using Curry (Shoham, 2021) and 100 trajectories were collected by performing random actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The training set consists of 5100 trajectories in STP and 22100 in Box-World, of which 5000 and 22000 were collected by solving the problem instances with a subgoal-based A* algorithm (Hart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1968) and 100 by performing random actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In subgoal-based A*, we generated subgoals progressively closer to the terminal state procedurally and executed A* to reach these subgoals, as solving complete problem instances with A* would have been computationally very expensive due to the complexity of the environments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In TSP, we do not limit the number of demonstrations available to the agent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Instead, we generate highly suboptimal trajectories by running an agent that visits 25 cities in a 25 × 25 grid in random order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 Agents Our HIPS agent consists of the following neural networks: the detector dξ(sgk+1|sgk), the low-level policy πθ(ai|si, sgk), the VQVAE encoder fφ(zek+1|sgk+1, sgk), the VQVAE decoder gψ(sgk+1|ek+1, sgk), the VQVAE prior p(ek+1|sgk), the low-level dynamics model fdyn(si+1|ai, si), and the distance function V (si).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The encoder fφ and detector dξ are not used during evaluation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' All networks are ResNet-based CNNs (He et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The decoder and low-level dynamics CNNs also contain FiLM layers (Perez et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We only use the 100 random trajectories for training the low-level dynamics model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Box-World, the neural networks additionally use Deep Recurrent Convolutions (Guez et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' All neural networks are implemented with PyTorch (Paszke et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2019) and trained with an Adam optimizer (Kingma and Ba, 2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We evaluate Sokoban and Box-World with PHS* as the high-level search algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Because of the suboptimality of the TSP trajectories, learning a good VQVAE prior is difficult, and we use A* in TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We also observed that GBFS is superior to PHS* with very small search budgets in STP (see Appendix D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We compare our agents to two main classes of baselines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The first class of baselines is strong IL and offline RL algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We compare our agent to standard flat behavioral cloning (BC), a powerful offline RL algorithm, Conservative Q-Learning (CQL, Kumar et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2020), and a strong IL algorithm, Inverse Soft-Q Learning (IQ-Learn, Garg et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We ran IQ-Learn in the online mode, where it could collect additional data during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Furthermore, we include the Decision Transformer (DT, Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021), the hierarchical IL algorithm Option-GAIL (Jing et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021), and the online goal-conditioned RL algorithm RIS (Chane-Sane et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021) in our experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' For CQL, we use the implementation of d3rlpy (Seno and Imai, 2021), and for other algorithms, we use the open-sourced implementations of the authors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' For all methods that need rewards, we give the agent a reward of one upon completing the task and 0 otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We do not include the 100 random trajectories in the dataset when evaluating imitation learning algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 6 Table 1: The overall success rates (%) of different algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The algorithms in the bottom part have access to the true environment dynamics, and those in the upper part do not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Method Sokoban STP BW TSP HIPS (ours) 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 HIPS-k (ours) 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 BC 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 CQL 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 DT 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 RIS 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 Option-GAIL 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 IQ-Learn 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-env (ours) 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 AdaSubS 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 kSubS 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 Table 2: The success rates (%, higher is better) of different search algorithms after performing N node expansions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Sokoban and Box-World, HIPS was evaluated with PHS*, with GBFS in STP, and with A* in TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sokoban Sliding Tile Puzzle Box-World Travelling Salesman N 100 500 1000 100 500 1000 100 500 1000 100 500 1000 HIPS (ours) 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 HIPS-k (ours) 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-env (ours) 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-env-k (ours) 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 AdaSubS 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 kSubS 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 PHS* 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 The second class of baselines is search methods that use the true environment dynamics, a state-of-the-art low-level search, Policy Guided Heuristic Search (PHS*, Orseau and Lelis, 2021), and two subgoal-level search methods: kSubS (Czechowski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021) and AdaSubS (Zawalski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In PHS*, we observed that using a policy trained with behavioral cloning works better than a policy trained using the loss function proposed by Orseau and Lelis (2021) and therefore, we use the BC policy in the experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We evaluate kSubS and AdaSubS with the autoregressive CNNs on all environments except Box-World because it would have required significant changes to the existing implementation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our method, HIPS, relies on a learned dynamics model instead of the true dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hence, it solves a more complex problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We also train a more comparable variant of our method, HIPS-env, that uses an environment simulator for planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 Results We use the overall success rates reported in Table 1 as the main evaluation metric in all environments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We evaluate the performance of each seed and take the mean over the random seeds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use 10 random seeds to evaluate our method, HIPS, at least five seeds per ablation, and at least three seeds per baseline method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' When evaluating the overall success rate, the search algorithms may perform as many expansions as necessary to find a solution to the problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The critical success factor for our model is the capacity of the generative model to cover the complete search space with the proposed subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our method, HIPS, outperforms the baseline methods in all four environments (see Table 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The table also contains an ablation of our method, HIPS-k, where we train the VQVAE with subgoals sampled at fixed intervals as done in AdaSubS and kSubS instead of using the detector network (see Appendices D and G for more details).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Eliminating the detector leads to a clear drop in performance in one of the four environments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS-k is superior to kSubS in all environments despite solving a more difficult problem than kSubS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS-k must learn the environment dynamics and a low-level policy, whereas kSubS uses a low-level search with the environment dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS-k also is superior to AdaSubS which also uses a low-level policy instead of search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The sparse reward structure and the required reasoning capabilities prove to be very difficult for the model-free baseline IL and offline RL methods that do 7 Table 3: The success rates of different algorithms (higher is better) and the average number of steps (lower is better) needed to solve TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Method Success rate (%) Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' steps HIPS-PHS* (ours) 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 305.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 HIPS-MCTS (ours) 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 213.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-A* (ours) 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 161.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 kSubS 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 268.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 CQL 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 336.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 BC 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 339.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 AdaSubS 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 338.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 Teacher 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 336.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 Oracle MCTS 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 199.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 Christofides 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 139.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 not rely on planning, which is why they struggle with all four tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS-env performs equally to HIPS, except in Box-World, where the search exploits the inaccuracies of the dynamics model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, HIPS was evaluated using open-loop planning, where one plan was generated, executed, and evaluated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' If the agent is allowed to re-plan when the dynamics model deviates from the environment and fine-tune the model with the new transition, the performance would most likely increase.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The issues with the dynamics model do not prevent HIPS from outperforming the baselines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Letting a search algorithm perform unlimited expansions is unrealistic in most real-world applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Following Czechowski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021), we evaluate the percentage of test problems solved after N node expansions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The results are given in Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The benefits of using RL to detect subframes become clear, as HIPS outperforms the fixed-length ablation HIPS-k in Sokoban and Sliding Tile Puzzle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS-env outperforms the baseline methods in Sokoban and Travelling Salesman and is superior to kSubS on STP when the search budget is small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PHS* can solve STP and TSP, but cannot make enough progress in 1000 node expansions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' AdaSubS cannot solve STP because it struggles to reliably reach the generated subgoals with its low-level policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Therefore, kSubS outperforms AdaSubS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' TSP is a problem where generating a successful trajectory is easy, but finding an efficient solution is much more difficult.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hence, we evaluate the solution lengths of the methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We also compare the search algorithms PHS*, MCTS, and A* when used with HIPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Given the amount of data, we perform VQVAE pre-training using pairs of subgoals instead of random pairs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In addition to the baselines capable of solving TSP, the performance of our method is compared to a known approximation algorithm Christofides (Christofides, 1976), to the training dataset (Teacher), and to an Oracle variant of subgoal-level MCTS where we replace the VQVAE generator with procedurally generated subgoals, where the agent is visiting one of the remaining unvisited cities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The solution lengths found in TSP are reported in Table 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS finds better solutions than the baselines in TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS with PHS* can only slightly improve the training data, whereas HIPS-MCTS with subgoal-level rollouts can find significantly better solutions than the ones in the training dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' It is inferior to the approximation algorithm, but the gap to the Oracle MCTS is small (around 6 %), which shows that the subgoals generated by the VQVAE are competitive with the procedurally generated subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Finally, HIPS-A* is the best-performing agent, and the gap to Christofides is around 15 %.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Note that our learned heuristic is non-admissible, and we trade off optimality for speed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' kSubS can also improve on the training dataset, but it is uncompetitive against HIPS-MCTS and HIPS-A*.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' CQL, BC, and AdaSubS can make some progress on the task and solve some instances, but they cannot improve the solution lengths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A problem of model-free baseline methods is the inability to commit to going to a specific city, which highlights the benefits of goal-conditioned learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Complete results including the standard errors can be found in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 Visualizations We visualize the subgoals and plans generated by our agent to gain further understanding into the agent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' An example of a high-level plan found by HIPS for Sokoban is visualized in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 4 illustrates the subgoals proposed by the HIPS generative model for an intermediate state in TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The model suggests visiting one of the unvisited cities as the next subgoal, which is a very reasonable planning strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' More visualizations of the high-level trajectories discovered by our agent and the subgoals proposed by the generative model can be found in Appendix G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 8 Figure 3: An example of a high-level plan (a sequence of subgoals) found by HIPS in Sokoban.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Figure 4: Examples of subgoals proposed for the current state (marked with blue boundaries) in TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 5 Conclusion We present a novel method for hierarchical IL that can address difficult reasoning problems that require long-term decision-making.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our approach relies on identifying subgoals from trajectories and generating new subgoals for search-based planning on new problem instances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our method outperforms powerful search, IL, and offline RL baselines on the benchmark tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Our experiments demonstrate that our VQVAE is a suitable generative model for subgoal-level search and using a detector to discover subgoals has benefits over subtrajectories of fixed length.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We see many promising directions for addressing the limitations of our method and improving it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Quantifying the model uncertainty could help prevent the search from exploiting the learned models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Combining discrete high-level planning with continuous low-level execution could make it possible to solve real-world tasks with robotics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning more abstract goals not formulated in the observation space to improve the efficiency of high-level planning and allowing the agent to ignore task-irrelevant sensory inputs are also left for future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Combining low-level and high-level searches would improve the solution rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Applying the learned high-level models in an RL setting to improve exploration or in a curriculum learning to solve progressively harder problem instances are also promising directions for future research.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' References Abbeel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Ng, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Apprenticeship learning via inverse reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Proceedings of the Twenty-First International Conference on Machine Learning, page 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Ablett, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Kelly, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning from guided play: A scheduled hierarchical approach for improving exploration in adversarial imitation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='08932.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Argenson, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Dulac-Arnold, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Model-based offline planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='05556.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Arthur, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Vassilvitskii, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' K-Means++: The advantages of careful seeding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Technical report, Stanford.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Chane-Sane, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Schmid, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Laptev, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Goal-conditioned reinforcement learning with imagined subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 1430–1440.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Kr¨ahenb¨uhl, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning from all vehicles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17222–17231.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Chen, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Rajeswaran, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Grover, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Laskin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Abbeel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Srinivas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Mordatch, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Decision transformer: Reinforcement learning via sequence modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 34:15084–15097.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 9 XTXXXChristofides, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1976).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Worst-case analysis of a new heuristic for the travelling salesman problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Technical report, Carnegie-Mellon Univ Pittsburgh Pa Management Sciences Research Group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Coulom, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Efficient selectivity and backup operators in monte-carlo tree search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Computers and Games, pages 72–83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Culberson, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1997).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sokoban is PSPACE-complete.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IEICE Technical Report.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Czechowski, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Odrzyg´o´zd´z, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zbysi´nski, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zawalski, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Olejnik, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Wu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kuci´nski, �L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Mi�lo´s, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Subgoal search for complex reasoning tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 34:624–638.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Daniel, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Van Hoof, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Peters, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Neumann, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Probabilistic inference for determining options in reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Machine Learning, 104(2):337–357.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Ebert, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Finn, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Dasari, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Xie, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lee, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Levine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Visual foresight: Model-based deep reinforcement learning for vision-based robotic control.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1812.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='00568.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Fox, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Berenstein, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Stoica, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Goldberg, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Multi-task hierarchical imitation learning for home automation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE), pages 1–8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IEEE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Garg, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chakraborty, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Cundy, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Song, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Ermon, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IQ-learn: Inverse soft-Q learning for imitation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 34:4028–4039.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Guez, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Mirza, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Gregor, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kabra, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Racani`ere, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Weber, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Raposo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Santoro, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Orseau, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Eccles, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' An investigation of model-free planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 2464–2473.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hafner, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Fischer, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Abbeel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Deep hierarchical planning from pixels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='04114.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hart, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Nilsson, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Raphael, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1968).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A formal basis for the heuristic determination of minimum cost paths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IEEE Transactions on Systems Science and Cybernetics, 4(2):100–107.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' He, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Ren, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Sun, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Deep residual learning for image recognition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Henderson, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='-D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Bacon, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Meger, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Pineau, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Precup, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Optiongan: Learning joint reward-policy options using generative adversarial inverse reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Jayaraman, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Ebert, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Efros, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Levine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Time-agnostic prediction: Predicting predictable video frames.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1808.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='07784.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Jing, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Huang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Sun, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Ma, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kong, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Gan, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Li, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Adversarial option-aware hierarchical imitation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 5097–5106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Jurgenson, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Avner, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Groshev, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Tamar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sub-goal trees a framework for goal-based reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 5020–5030.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kingma, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Ba, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Adam: A method for stochastic optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1412.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6980.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kipf, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Dai, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zambaldi, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Sanchez-Gonzalez, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Grefenstette, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kohli, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Battaglia, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Compile: Compositional imitation learning and execution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 3418–3428.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kober, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Peters, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Policy search for motor primitives in robotics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kocsis, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Szepesv´ari, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Bandit based monte-carlo planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In European Conference on Machine Learning, pages 282–293.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Korf, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1985).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Depth-first iterative-deepening: An optimal admissible tree search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Artificial Intelligence, 27(1):97–109.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 10 Kumar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Hong, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Levine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' When should we prefer offline reinforcement learning over behavioral cloning?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='05618.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kumar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zhou, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Tucker, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Levine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Conservative q-learning for offline reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 33:1179–1191.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' �La´ncucki, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chorowski, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Sanchez, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Marxer, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chen, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Dolfing, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Khurana, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Alum¨ae, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Laurent, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Robust training of vector quantized bottleneck models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In 2020 International Joint Conference on Neural Networks (IJCNN), pages 1–7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IEEE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Le, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Jiang, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Agarwal, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Dudik, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Yue, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Daum´e III, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical imitation and reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 2917–2926.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Li, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Tang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Tomizuka, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Zhan, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical planning through goal-conditioned offline reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2205.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11790.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Mayer, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Gomez, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Wierstra, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Nagy, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Knoll, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Schmidhuber, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A system for robotic heart surgery that learns to tie knots using recurrent neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advanced Robotics, 22(13-14):1521–1537.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Nachum, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Tang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Gu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lee, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Levine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Why does hierarchy (sometimes) work so well in reinforcement learning?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1909.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10618.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Nair, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Finn, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Hierarchical foresight: Self-supervised learning of long-horizon tasks via visual subgoal generation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1909.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='05829.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Orseau, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Lelis, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Policy-guided heuristic search with guarantees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 12382–12390.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Osa, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Pajarinen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Neumann, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Bagnell, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Abbeel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Peters, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' An algorithmic perspective on imitation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Foundations and Trends® in Robotics, 7(1-2):1–179.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Ozair, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Razavi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Antonoglou, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Van Den Oord, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Vinyals, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Vector quantized models for planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 8302–8313.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Paszke, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Gross, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Massa, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lerer, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Bradbury, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chanan, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Killeen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lin, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Gimelshein, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Antiga, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Pytorch: An imperative style, high-performance deep learning library.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Paul, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Vanbaar, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Roy-Chowdhury, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning from trajectories via subgoal discovery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Perez, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Strub, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', De Vries, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Dumoulin, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Courville, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Film: Visual reasoning with a general conditioning layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Pertsch, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Rybkin, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Ebert, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zhou, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Jayaraman, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Finn, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Levine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Long-horizon visual planning with goal-conditioned hierarchical predictors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 33:17321–17333.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Pertsch, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Rybkin, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Yang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zhou, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Derpanis, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lim, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Daniilidis, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Jaegle, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Keyin: Keyframing for visual planning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Conference on Learning for Dynamics and Control.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Racani`ere, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Weber, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Reichert, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Buesing, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Guez, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Jimenez Rezende, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Puigdom`enech Badia, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Vinyals, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Heess, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Imagination-augmented agents for deep reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sammut, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Hurst, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kedzier, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Michie, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1992).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Learning to fly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Machine Learning Proceedings 1992, pages 385–393.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Elsevier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Schrader, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='-P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' gym-sokoban.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='com/mpSchrader/gym-sokoban.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Schrittwieser, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Antonoglou, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Hubert, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Simonyan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Sifre, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Schmitt, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Guez, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lockhart, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Hassabis, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Graepel, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Mastering atari, go, chess and shogi by planning with a learned model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Nature, 588(7839):604–609.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 11 Seno, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Imai, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' d3rlpy: An offline deep reinforcement learning library.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='03788.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sharma, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Sharma, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Rhinehart, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Kitani, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Directed-info gail: Learning hierarchical policies from unsegmented demonstrations using directed information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1810.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='01266.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sharma, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Pathak, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Gupta, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Third-person visual imitation learning via decoupled hierarchical controller.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Shoham, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Curry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' https://festival-solver.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='site/curry/.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sutton, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Precup, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Singh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Artificial Intelligence, 112(1-2):181–211.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Van Den Oord, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Vinyals, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Kavukcuoglu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Neural discrete representation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Vinyals, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Babuschkin, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Czarnecki, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Mathieu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Dudzik, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Chung, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Alphastar: Mastering the real-time strategy game Starcraft II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' DeepMind Blog, 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Williams, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (1992).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Simple statistical gradient-following algorithms for connectionist reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Machine Learning, 8(3):229–256.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Ye, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kurutach, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Abbeel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Gao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Mastering atari games with limited data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 34:25476–25488.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Zambaldi, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Raposo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Santoro, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Bapst, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Babuschkin, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Tuyls, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Reichert, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lillicrap, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Lockhart, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Relational deep reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:1806.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='01830.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Zawalski, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Tyrolski, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Czechowski, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Stachura, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Piekos, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Odrzyg´o´zd´z, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Wu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Kuci´nski, �L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Mi�lo´s, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Fast and precise: Adjusting planning horizon with adaptive subgoal search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' arXiv preprint arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='00702.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Zhang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Li, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zheng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Wei, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', Zhang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', and Zhang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Explainable hierarchical imitation learning for robotic drink pouring.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' IEEE Transactions on Automation Science and Engineering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Zhang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' and Paschalidis, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Provable hierarchical imitation learning via EM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In International Conference on Artificial Intelligence and Statistics, pages 883–891.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 12 A Ethical Considerations We do not see any immediate negative societal impacts associated with our work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We do not train our models with any sensitive or private data, and our model is not directly applicable to, for instance, real-world decision-making concerning humans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, we cannot exclude the method being applied to something harmful that is difficult to foresee, for instance, military purposes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' B Infrastructure We trained our models on an HPC cluster using one NVIDIA GPU and multiple CPU workers per run.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Most runs were performed on V100 GPUs with 32 GB of GDDR SDRAM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' For some of the runs, the GPU was an A100, a K100, or a P80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We used 6 CPU workers per GPU with 10 GB of RAM per worker and each worker running on one core.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' By reducing the number of workers, it is possible to train and evaluate the agent on a workstation with Intel i7-8086K, 16 GB of RAM, and an NVIDIA GeForce GTX 1080 Ti GPU with 10 GB of video memory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' C Complete Results Table 4: The overall success rates (%) of different algorithms including the standard errors of the means of the runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The algorithms in the bottom part have access to the true environment dynamics, and those in the upper part do not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Method Sokoban STP BW TSP HIPS (ours) 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 HIPS-k (ours) 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 BC 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 CQL 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 DT 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 RIS 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 Option-GAIL 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 IQ-Learn 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-env (ours) 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 AdaSubS 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 kSubS 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 Table 5: The success rates (%, higher is better) of different search algorithms after performing N node expansions in Sokoban and STP, including the standard errors of the means of the runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sokoban Sliding Tile Puzzle N 100 500 1000 100 500 1000 HIPS (ours) 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 HIPS-k (ours) 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 HIPS-env (ours) 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 HIPS-env-k (ours) 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 AdaSubS 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 kSubS 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 PHS* 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 13 Table 6: The success rates (%, higher is better) of different search algorithms after performing N node expansions in Box-World and TSP, including the standard errors of the means of the runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Box-World Travelling Salesman N 100 500 1000 100 500 1000 HIPS (ours) 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 HIPS-k (ours) 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-env (ours) 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-env-k (ours) 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 AdaSubS 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 kSubS 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 PHS* 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 Table 7: The success rates of different algorithms (higher is better) and the average number of steps (lower is better) needed to solve TSP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The table also includes the standard errors of the means of the runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Method Success rate (%) Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' steps HIPS-PHS* (ours) 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 305.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 HIPS-MCTS (ours) 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 213.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 HIPS-A* (ours) 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 161.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 kSubS 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 268.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 CQL 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 336.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 BC 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 339.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 AdaSubS 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 338.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 Teacher 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 336.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 Oracle MCTS 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 199.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 Christofides 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 139.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 14 D Ablation: HIPS-k Table 8: The success rates (%, higher is better) of different variants of HIPS in STP after performing N node expansions N 100 500 1000 ∞ HIPS-PHS* 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 HIPS-GBFS 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 HIPS-GBFS-3 N/A N/A N/A N/A HIPS-GBFS-5 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ± 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ± 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 HIPS-GBFS-7 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 HIPS-GBFS-9 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 When we replace the detector dξ with subgoals sampled at fixed intervals, we re-train the low-level policy πθ to achieve these subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The discrete VQVAE, including the prior, are also re-trained using the new pairs of consecutive subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The distance between the subgoals can, in this case, be controlled by a hyperparameter k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In Sokoban, STP, and Box-World, we used ten as the subgoal horizon and five as k (see Table 10).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' In TSP, selecting a segment length half of the subgoal horizon proved to be too much, so we let k be equal to four, the default segment length used in kSubS (Czechowski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We performed a small ablation study in STP to analyze the impact of the value of k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The results are shown in Table 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We see that given a large research budget, PHS* is slightly superior to GBFS, but GBFS outperforms PHS* given a small search budget.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' HIPS-GBFS-3 doesn’t converge because the value function is noisy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Using a larger k allows the value function to ”leap over” the noise, as observed by Czechowski et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We see that using a larger k improves the percentage of puzzles solved after a smaller number of expansions, but hurts the overall solution rate, as the search space is explored less systematically.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' However, using a k too large is also harmful as training the generative model becomes difficult.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Furthermore, no value of k was able to outperform HIPS-GBFS, which highlights the benefits of our method that can propose subgoals at different distances adaptively in all environments (see Figure 5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Sokoban STP Box-World TSP Figure 5: Lengths of the subtrajectories to reach subgoals proposed by VQVAE when it has been trained using the detector dξ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='00 10 20 30 40 500.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='04- 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='00 0 5 10 15 20 250.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='175 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='150 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='125 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='100 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='075 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='050 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='025 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='000 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='0 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='00.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='04 - 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='00 0 5 10 15 20 25 30E Search Methods Greedy Best-First Search (GBFS) is a priority queue -based search algorithm, where the evaluation function has been defined as ϕ(n) = h(n), where h(n) is a heuristic that predicts the distance to the goal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The node that is predicted to be the closest to the goal is expanded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Policy-Guided Heuristic Search (PHS) is a policy-guided search algorithm (Orseau and Lelis, 2021) which uses a priority queue with the evaluation function ϕ(n) = η(n)g(n)/π(n), where g(n) is the path cost from the root to node n, π(n) is the node policy (the probability of selecting node n) and η(n) is a heuristic factor whose purpose is to estimate the cost to the nearest descendant solution node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' We use a variant of PHS, PHS* where the heuristic factor has been defined as ηh(n) = 1 + h(n)/g(n) π(n)h(n)/g(n) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (4) A* is a heuristic-based search algorithm that tries to find the shortest path to the goal (Hart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 1968).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' It is based on a priority queue with the evaluation function ϕ(n) = g(n) + h(n), where g(n) is the distance from the root to node n and h(n) is a heuristic that predicts the distance from the node n to the goal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' If the heuristic h(n) never overestimates the true distance to the goal, the heuristic is said to be admissible, and A* is guaranteed to find the shortest path.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Monte Carlo Tree Search (MCTS) is a tree-based search method based on expanding a search tree by performing Monte Carlo evaluations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The selection of nodes for expansion is biased towards promising nodes to enable MCTS to focus on the relevant parts of the search tree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The most commonly used method is UCT, where nodes with higher rewards and lower visitation frequency get the highest priority (Ozair et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Kocsis and Szepesv´ari, 2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Listing 1 contains the pseudo-code describing our high-level search with a priority queue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Subroutine phs cost calculates the value of the heuristic factor η(n) for PHS* as in (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Subroutine extract plan collects the subgoals on the path to the leaf node (terminal state) from the root (initial state).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Subroutine get distances takes as input the current state and the proposed children and tries to reach them using the subgoal-conditioned low-level policy πθ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The distances between the state and the children are recovered simultaneously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 16 Listing 1 PyTorch pseudocode for the high-level search 1 def get_priority (node , alg): 2 if alg == ’phs_star ’: 3 return phs_cost(node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='log_p , node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='value , node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='cum_dist) 4 elif alg == ’gbfs ’: 5 return node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='value 6 elif alg == ’a_star ’: 7 return node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='value + node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='cum_dist 8 9 10 def init_node(node , alg , vqvae , policy , value_func , dynamics): 11 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='value = value_func(node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='state) 12 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' child_states = vqvae.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='generate(node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='state) 13 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' distances_to_children = get_distances ( 14 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='state , 15 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='child_states , 16 policy , 17 dynamics) 18 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' filter_unreachable_children () # Uses the distances computed 19 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' children_log_probs = vqvae.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='prior(node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='state) 20 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='priority = get_priority (node , alg) 21 22 23 def search(state , alg , vqvae , policy , value_func , dynamics): 24 n_nodes = 0 25 queue = PriorityQueue () # Create empty priority queue 26 expanded = Set () # Create empty set 27 node = Node( 28 state , 29 parent=None , 30 cum_dist =0, 31 log_p =0, 32 ) 33 init_node(node , alg , vqvae , policy , value_func , dynamics) 34 queue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='insert(node) 35 36 while len(queue) > 0: 37 node = queue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='pop () 38 expanded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='add(node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='state) 39 n_nodes += 1 40 for c_state , c_dist , c_log_p in zip(node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='child_states , 41 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' distances_to_children , 42 node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' children_log_probs ): 43 if c_state in expanded: 44 continue 45 new_node = Node( 46 c_state , 47 parent=node , 48 cum_dist=node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='cum_dist + c_dist , 49 log_p=node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='log_p + c_log_p , 50 ) 51 if is_terminal (c_state): 52 return extract_plan (new_node), n_nodes # Success 53 init_node(new_node , alg , vqvae , policy , value_func , dynamics) 54 queue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='insert(new_node) 55 return None , n_nodes # Search failed , queue empty 17 F VQVAE Training Algorithm 2 Training VQVAE for Subgoal Generation Input: A dataset of trajectories D, a dataset of subgoal pairs D′, untrained encoder fφ, decoder gψ, codebook {ek} Parameters: The codebook {ek} and the parameters of the encoder, φ, and the decoder, ψ Output: Trained encoder fφ, decoder gψ, codebook {ek}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1: while fφ and gψ not converged do 2: Sample a trajectory τ from D 3: For each state si ∈ τ, uniformly sample a pair sj from (si+1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sH) 4: Reconstruct sj without discretization: ˆsj = gψ(fφ(sj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' si),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' si) 5: Compute reconstruction loss Lrec(ˆsj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sj) 6: Update φ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' ψ to minimize reconstruction loss 7: end while 8: Sample subsequent subgoal pairs sgj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sgj−1 from D′ and encode them with the encoder: zj = fφ(sgj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sgj−1) 9: Initialize {ek} as the clusters centers obtained by running KMeans++ on the encodings {zj} 10: while {ek},' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' fφ and gψ not converged do 11: Sample a batch of subsequent subgoal pairs sgj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sgj−1 from D′ 12: Reconstruct subgoals with VQVAE zj = fφ(sgj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sgj−1) kj = arg min m ∥zj − em∥2 ˆsgj = gψ(ekj,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' sgj−1) 13: Compute the loss in (3) 14: Update φ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' {ek} to minimize the loss computed in Step 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 15: end while 16: return fφ, gψ, {ek} G Hyperparameters and Visualizations Table 9: General hyperparameters of our method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Parameter Value Learning rate for dynamics 2 · 10−4 Learning rate for π, d, V 1 · 10−3 Learning rate for VQVAE 2 · 10−4 Discount rate for REINFORCE 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='99 Table 10: Environment-specific hyperparameters of our method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Parameter Explanation Sokoban STP Box-World TSP α Subgoal penalty 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='05 β Beta for VQVAE 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 0 c Exploration constant for MCTS – – – 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 D Codebook dimensionality 128 128 128 64 H Subgoal horizon 10 10 10 50 K VQVAE codebook size 64 64 64 32 k Segment length w/o REINFORCE 5 5 5 4 (N, D) DRC size – – (3, 3) – 18 (a) A subgoal-level plan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b) Examples of generated subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Figure 6: Visualization of the solution found by HIPS for a Sokoban problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (a): A subgoal-level plan found by HIPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b): Subgoals proposed for an intermediate state (marked with blue boundaries).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The subgoals have been sorted according to the prior probabilities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The subgoal selected for the final plan is marked with red boundaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 19 XTX区 XXX XX 区 XXXXIXXIXXXXTXX XXXXXIXXIXXTXXIX(a) A subgoal-level plan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b) Examples of generated subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Figure 7: Visualization of the solution found by HIPS for an STP problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (a): A subgoal-level plan found by HIPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b): Subgoals proposed for intermediate states (marked with blue boundaries).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The subgoals have been sorted according to the prior probabilities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The subgoal selected for the final plan is marked with red boundaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Red color is used to highlight the tiles which are different from the reference state (the previous state in (a) and the current state in (b)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='173 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='243 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2410 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2410 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='242 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='242 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='245 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='245 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='241 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24 1720 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='9 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2420 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='23 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='13 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='19 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='11 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='21 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='17 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='22 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content='24(a) A subgoal-level plan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b) Examples of generated subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Figure 8: Visualization of the solution found by HIPS for a BW problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (a): A subgoal-level plan found by HIPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b): Subgoals proposed for an intermediate state (marked with blue boundaries).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The subgoals have been sorted according to the prior probabilities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' The subgoal selected for the final plan is marked with red boundaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 21 10 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 10?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1818 180 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 100?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1800 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 10 100 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' ?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 瘤 10 180?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 1010 0?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 180?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 18(a) A subgoal-level plan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b) Examples of generated subgoals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' Figure 9: Visualization of the solution found by HIPS for a TSP instance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (a): A subgoal-level plan found by HIPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' (b): Subgoals proposed for an intermediate state (marked with blue boundaries).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
+page_content=' 22' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WNFOT4oBgHgl3EQf7TSv/content/2301.12962v1.pdf'}
diff --git a/WdFQT4oBgHgl3EQfbzbk/content/tmp_files/2301.13325v1.pdf.txt b/WdFQT4oBgHgl3EQfbzbk/content/tmp_files/2301.13325v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4d4d45fb609d5d45a360edd936ce6f20f3e32622
--- /dev/null
+++ b/WdFQT4oBgHgl3EQfbzbk/content/tmp_files/2301.13325v1.pdf.txt
@@ -0,0 +1,1858 @@
+manuscript submitted to Earth and Planetary Physics
+Hybrid-Vlasov simulation of soft X-ray emissions at the
+Earth’s dayside magnetospheric boundaries
+M. Grandin1, H. K. Connor2, S. Hoilijoki1, M. Battarbee1, Y. Pfau-Kempf1,
+U. Ganse1, K. Papadakis1, and M. Palmroth1,3
+1Department of Physics, University of Helsinki, Helsinki, Finland
+2NASA Goddard Space Flight Center, Greenbelt, MD, 20771, USA
+3Space and Earth Observation Centre, Finnish Meteorological Institute, Helsinki, Finland
+Key Points:
+• We produce soft X-ray images of near-Earth space with a global hybrid-Vlasov
+simulation with southward interplanetary magnetic field
+• Flux transfer events can produce X-ray signatures despite being transient phenom-
+ena if they cumulatively increase the proton density locally
+• Mirror-mode structures in the magnetosheath can also produce soft X-ray signa-
+tures in time-integrated images
+Corresponding author: Maxime Grandin, maxime.grandin@helsinki.fi
+–1–
+arXiv:2301.13325v1 [physics.space-ph] 30 Jan 2023
+
+manuscript submitted to Earth and Planetary Physics
+Abstract
+Solar wind charge exchange produces emissions in the soft X-ray energy range which can
+enable the study of near-Earth space regions such as the magnetopause, the magnetosheath
+and the polar cusps by remote sensing techniques. The Solar wind–Magnetosphere–Ionosphere
+Link Explorer (SMILE) and Lunar Environment heliospheric X-ray Imager (LEXI) mis-
+sions aim to obtain soft X-ray images of near-Earth space thanks to their Soft X-ray Im-
+ager (SXI) instruments. While earlier modeling works have already simulated soft X-ray
+images as might be obtained by SMILE SXI during its mission, the numerical models
+used so far are all based on the magnetohydrodynamics description of the space plasma.
+To investigate the possible signatures of ion-kinetic-scale processes in soft X-ray images,
+we use for the first time a global hybrid-Vlasov simulation of the geospace from the Vlasi-
+ator model. The simulation is driven by fast and tenuous solar wind conditions and purely
+southward interplanetary magnetic field. We first produce global X-ray images of the day-
+side near-Earth space by placing a virtual imaging satellite at two different locations,
+providing meridional and equatorial views. We then analyze regional features present
+in the images and show that they correspond to signatures in soft X-ray emissions of mirror-
+mode wave structures in the magnetosheath and flux transfer events (FTEs) at the mag-
+netopause. Our results suggest that, although the time scales associated with the mo-
+tion of those transient phenomena will likely be significantly smaller than the integra-
+tion time of of the SMILE and LEXI imagers, mirror-mode structures and FTEs can cu-
+mulatively produce detectable signatures in the soft X-ray images. For instance, a lo-
+cal increase by 30% in the proton density at the dayside magnetopause resulting from
+the transit of multiple FTEs leads to a 12% enhancement in the line-of-sight- and time-
+integrated soft X-ray emissivity originating from this region. Likewise, a proton density
+increase by 14% in the magnetosheath associated with mirror-mode structures can re-
+sult in an enhancement in the soft X-ray signal by 4%. These are likely conservative es-
+timates, given that the solar wind conditions used in the Vlasiator run can be expected
+to generate weaker soft X-ray emissions than the more common denser solar wind. These
+results will contribute to the preparatory work for the SMILE and LEXI missions by pro-
+viding the community with quantitative estimates of the effects of small-scale, transient
+phenomena occurring on the dayside.
+1 Introduction
+Over the past two decades, interest has grown in studying near-Earth space by ob-
+serving the soft X-ray emissions created by charge-exchange interactions between heavy,
+multiply charged ions and neutral species. For example, charge-exchange interactions
+between O7+ or O8+ ions present in the solar wind and neutral hydrogen atoms from
+the Earth’s exosphere are known to lead to photon emissions through the de-excitation
+of the product ion species, which include photon energies in the soft X-ray range (0.5–
+0.7 keV). This process is known as solar wind charge exchange (SWCX). A review by
+Sibeck et al. (2018) provides in-depth details on the processes at play as well as pioneer-
+ing works on the study of terrestrial and planetary space through the observation of soft
+X-ray emissions.
+Currently, several satellite missions aim at imaging the geocorona in soft X-rays.
+Two major upcoming missions making use of SWCX soft X-ray imaging to reseach near-
+Earth space are the Lunar Environment heliospheric X-ray Imager (LEXI), led by the
+Boston University in collaboration with NASA GSFC and several universities (http://
+sites.bu.edu/lexi), and the Solar wind–Magnetosphere–Ionosphere Link Explorer (SMILE)
+mission, designed jointly by the European Space Agency and the Chinese Academy of
+Sciences (Branduardi-Raymont et al., 2018). Both of these missions will provide soft X-
+ray images of the polar cusps and the magnetosheath, which are known to be the most
+prominent sources of SWCX soft X-ray emissions in near-Earth space, with a goal of un-
+derstanding global interaction between the solar wind and the Earth’s magnetosphere.
+–2–
+
+manuscript submitted to Earth and Planetary Physics
+After its expected launch in 2024, LEXI will observe the dayside magnetosheath from
+the lunar surface for up to two weeks of a short mission period due to the harsh lunar
+environment. SMILE will be launched in 2025 into a highly elliptical polar orbit with
+an apogee of ∼20 Earth radii (RE), observing the dayside magnetosheath and cusps for
+up to 40 continuous hours per orbit during three years of its mission period.
+While there have been no wide-field-of-view soft X-ray observations of the geospace,
+the existence of near-Earth SWCX soft X-ray emissions has been studied in various works
+using data from the XMM and ROSAT astrophysics missions (e.g. Carter et al., 2010,
+2011; Cravens et al., 2001; Snowden et al., 2004; Connor & Carter, 2019; Jung et al., 2022;
+Zhang et al., 2022). Subsequently, modeling efforts were undertaken to understand the
+soft X-ray emissions detected by the ROSAT mission (Snowden et al., 1995) and to quan-
+tify the contributions of various source mechanisms (e.g., Cox, 1998; Cravens et al., 2001).
+Robertson and Cravens (2003b) were the first to produce images of the dayside magne-
+topause and magnetosheath in soft X-ray emissions through the SWCX mechanism us-
+ing a numerical model, and based on their results they suggested that it might be pos-
+sible to make use of those emissions to monitor the solar wind–magnetosphere interac-
+tions through remote sensing observations. The same authors also produced simulated
+images in soft X-rays for a virtual instrument placed on the Earth’s surface and observ-
+ing various directions in the sky (Robertson & Cravens, 2003a), and investigated the po-
+lar cusps’ signatures in soft X-ray images in further simulations (Robertson et al., 2006).
+More recent studies have taken up the modeling efforts as part of the preparatory
+phases of the LEXI and SMILE missions. Collier and Connor (2018) determined that
+the surface of the magnetopause could in principle be determined by soft X-ray imag-
+ing of near-Earth space, as the line-of-sight-integrated soft X-ray emissions maximize for
+observations tangential to the magnetopause surface. Based on these results, Sun et al.
+(2020) developed a tangent-fitting method to derive the magnetopause position in an-
+ticipated images from the Soft X-ray Imager (SXI) instrument onboard SMILE. Alter-
+native methods to determine the magnetopause shape and position from SXI observa-
+tions have been devised by C. Wang and Sun (2022). Among the processes taking place
+in near-Earth space and which could be studied using soft X-ray imaging, Sun et al. (2019)
+showed that SXI will be able to monitor the motions of the dayside magnetopause and
+of the polar cusps in response to changes in the solar wind driving conditions. Connor
+et al. (2021) also addressed similar research questions by simulating soft X-ray images
+which could be obtained during the SMILE and LEXI missions in an event when the in-
+terplanetary magnetic field (IMF) turns southward and an event when the solar wind
+density abruptly increases. They demonstrated that it might be possible to get infor-
+mation on the time variations in the magnetic reconnection rate at the dayside magne-
+topause using such remote-sensing techniques, and they argued that these missions could
+enable the tracing of the energy flow from the solar wind all the way to the cusps with
+the same instrument.
+The above-mentioned modeling studies all used numerical models of space plasma
+based on the magnetohydrodynamics (MHD) paradigm, which treats the plasma as a
+fluid. To complement the results obtained with this approach, it would be interesting
+to carry out modeling studies of SWCX soft X-ray imaging based on a global model of
+near-Earth space relying on a kinetic description of the plasma. Indeed, a certain num-
+ber of space plasma processes are intrinsically of kinetic origin, and therefore cannot emerge
+in MHD simulations. Examples of such processes include the growth of instabilities, wave–
+particle interactions as well as ion-scale physics associated with magnetic reconnection.
+In this paper, we will for the first time carry out a study of soft X-ray emissions
+relying on a 3D global hybrid-Vlasov simulation. We will make use of the kinetic nature
+of the model to investigate to what extent near-Earth space processes associated with
+ion-scale physics might produce signatures in soft X-ray images as will be obtained by
+the SMILE SXI instrument. To that aim, we will make use of Vlasiator (Palmroth et al.,
+–3–
+
+manuscript submitted to Earth and Planetary Physics
+2018), which is a global hybrid-Vlasov model of near-Earth space. In particular, we will
+quantify the signatures in terms of line-of-sight- and time-integrated soft X-ray emissions
+of two transient phenomena occurring on the dayside: mirror-mode waves in the mag-
+netosheath and flux transfer events (FTEs) produced by magnetic reconnection at the
+dayside magnetopause. Mirror-mode waves are generated by temperature anisotropy in
+the magnetosheath and appear as nonpropagating structures (in the plasma frame) with
+anticorrelated variations in plasma density and magnetic field magnitude (Hasegawa, 1969).
+FTEs are flux-rope-like structures forming along the dayside magnetopause in presence
+of multiple or bursty reconnection lines associated with southward IMF conditions, which
+often form near the subsolar point and propagate toward the polar cusps (Russell & El-
+phic, 1978; Rijnbeek et al., 1984; Berchem & Russell, 1984).
+The paper is organized as follows. The Vlasiator model and the methodology to
+construct soft X-ray images in the utilized simulation run are introduced in Sect. 2. Then,
+in Sect. 3, we present examples of global images of the dayside near-Earth space obtained
+with virtual imaging spacecraft placed at two locations: on the dawnside (meridional view)
+and above the north pole (equatorial view). Section 4 describes the analysis and results
+on the detectability of transient processes (FTEs and mirror-mode waves) in the sim-
+ulated soft X-ray images. A discussion of the results is provided in Sect. 5, and Sect. 6
+gives a summary of the main conclusions of the paper.
+2 Models and Methods
+2.1 Vlasiator
+2.1.1
+Model
+Vlasiator (von Alfthan et al., 2014; Palmroth et al., 2018) is a global hybrid-Vlasov
+code simulating near-Earth plasma at ion-kinetic scales. In the hybrid-Vlasov approach,
+ions are described through their velocity distribution function (VDF) which is discretized
+in a 3-dimensional (3D) velocity grid, resulting in ion populations evolving in a 6D phase
+space (3D in ordinary space + 3D in velocity space). In practice, Vlasiator solves the
+Vlasov equation for ions and treats electrons as a massless, charge-neutralizing fluid. While
+most existing Vlasiator simulation runs consider protons as the sole ion species, heav-
+ier ions such as He2+ have also been included in a few runs (e.g., Battarbee, Blanco-Cano,
+et al., 2020).
+Electromagnetic fields are propagated in time by solving Maxwell’s equations un-
+der the Darwin approximation, i.e., neglecting the displacement current term in Amp`ere’s
+law. The equation system is closed through a generalized Ohm’s law including the Hall
+term and a polytropic (adiabatic) description of the electron pressure gradient term. The
+model considers an ideal geomagnetic dipole field using a nonscaled strength 8·1015 Tm3
+and with a zero tilt between its axis and the Z direction in the Geocentric Solar Eclip-
+tic (GSE) frame of reference — in the GSE frame, the Earth’s center is at the origin, the
+X axis points toward the Sun, the Z axis toward the north, and the Y axis completes
+the orthonormal frame and points toward dusk. The geomagnetic dipole is described through
+a vector potential, which is scaled to zero at the inflow boundary in order to prevent mag-
+netic divergence entering the simulation domain. Magnetic and electric fields are prop-
+agated using a finite difference upwind field solver (Londrillo & del Zanna, 2004).
+The simulation domain extent varies from one run to another, but it generally in-
+cludes the dayside magnetosphere, the magnetosheath, the bow shock, the ion foreshock
+(if driving conditions allow for its existence), and part of the magnetotail. This has al-
+lowed for a great variety of studies focusing on a wide range of processes, such as fore-
+shock waves (Turc et al., 2018), foreshock cavitons (Tarvus et al., 2021), bow shock non-
+locality (Battarbee, Ganse, et al., 2020), magnetosheath waves (Hoilijoki et al., 2016; Dubart
+et al., 2020), energy transfer across the magnetopause (Ala-Lahti et al., 2022), magne-
+–4–
+
+manuscript submitted to Earth and Planetary Physics
+totail current sheet flapping (Juusola et al., 2018), or auroral proton precipitation (Grandin
+et al., 2019, 2020). While those past studies were based on 2D–3V runs (2D in ordinary
+space, 3D in velocity space), recent code developments have enabled the production of
+the first Vlasiator 3D–3V runs. This study makes use of one such 3D–3V run, described
+below.
+2.1.2
+Simulation Run
+The Vlasiator run used in this study has its simulation domain extending from −110 RE
+to 50 RE in the X direction (with RE the Earth’s radius; RE = 6371 km) and confined
+within |Y | < 58 RE and |Z| < 58 RE. The inner boundary lies at 4.7 RE from the ori-
+gin; it is implemented as a perfectly conducting sphere on which VDFs are fixed Maxwellian
+distributions. External boundaries have Neumann boundary conditions, except for the
++X wall from which the solar wind and IMF enter the simulation domain.
+Driving conditions in this 3D–3V run are as follows: purely southward IMF with
+Bx = By = 0 and Bz = −5 nT, solar wind with proton number density of 1 cm−3,
+speed along the −X direction at 750 km s−1, temperature of 500 kK. Solar wind con-
+ditions are homogeneous and constant throughout the simulation. Protons are the sole
+ion species in this run. Initial conditions are such that the whole simulation domain is
+filled with solar-wind-like Maxwellian VDFs and the superposition of the dipole geomag-
+netic field with the IMF, and the near-Earth space regions thus form self-consistently
+during the first few hundred seconds of the simulation. In this study, we will focus on
+the time interval starting at t = 800 s, when the dayside magnetosphere is well formed
+and lasting until the end of the simulation at t = 1506 s. In this run, an output file was
+written at a cadence of 1 s, which contains the plasma bulk parameters as well as elec-
+tromagnetic field components in every simulation cell.
+One development which made 3D–3V runs possible was the implementation of adap-
+tive mesh refinement (AMR) for the ordinary-space mesh. In this run, least-refined re-
+gions have a base grid with 8000 km resolution, and there are three refinement levels at
+4000, 2000 and 1000 km resolution to improve the description of regions of interest where
+ion-scale kinetic processes are important (bow shock, magnetosheath, magnetopause, mag-
+netotail current sheet). The velocity space is a uniform 3D Cartesian grid with a reso-
+lution of 40 km s−1. The electric and magnetic fields are solved on a uniform Cartesian
+grid at constant resolution of 1000 km in a process described by Papadakis et al. (2022).
+Figure 1 gives an overview of the simulation domain at t = 1100 s. It consists of
+three slices in the X = 0, Y = 0 and Z = 0 planes showing the proton number den-
+sity. While the chosen viewing angle and slice extents enable seeing dayside structures
+and processes relevant to this study, the actual simulation domain extends beyond this
+figure (see above). In the figure, one can identify the bow shock, the magnetosheath ex-
+hibiting wave-like structures, the magnetopause as well as the northern polar cusp. The
+magnetotail, partly hidden behind the X = 0 slice, will not be the focus of this study
+but exhibits complex dynamics.
+2.2 Soft X-ray Image Generation
+2.2.1
+Analytical Expression of Soft X-ray Emissivity
+In this study, the method to derive the local soft X-ray emissivity within the sim-
+ulation will follow the same approach as past studies (e.g. Connor et al., 2021; Sun et
+al., 2020). It is calculated at position r in the simulation domain with the following ex-
+pression
+Qloc(r) = αX
+4π np(r)nH(r)Veff(r),
+(1)
+–5–
+
+manuscript submitted to Earth and Planetary Physics
+Figure 1.
+Proton number density in the X = 0, Y
+= 0 and Z = 0 planes of the simulation
+domain at t = 1100 s in the Vlasiator simulation, wherein the bow shock, the magnetopause and
+the northern polar cusp are prominent. One can also identify wave-like structures in the dayside
+magnetosheath.
+–6–
+
+t=1100.0 s - origin at (0,0,0)[RE)
+Tick every 10 Re
+nproton
+[cm-3i
+5
+Z[RE]
+4
+-3
+2
+0.8
+0.6
+0.5
+0.4
+0.3
+X [RE]
+0.2
+Y [RE]
+0.1manuscript submitted to Earth and Planetary Physics
+where αX is the interaction efficiency factor, np is the proton number density, nH is the
+neutral hydrogen atom density, and Veff is the so-called “effective velocity”, defined as
+Veff(r) =
+�
+Vp(r)2 + 5
+3
+kBT(r)
+mp
+.
+(2)
+In Eq. (2), Vp is the proton bulk velocity, kB is Boltzmann’s constant, T is the plasma
+temperature, and mp is the proton mass. This equation expresses Veff in m s−1. Note
+that protons do not emit soft X-rays. The highly charged, heavy solar wind ions like C6+,
+N6+, N7+, Ne9+, S10+, O7+, and O8+ emit soft X-rays through SWCX (Sibeck et al.,
+2018). By multiplying the interaction efficiency factor αX, the proton-based quantity in
+Eq. (1) is transformed into the soft X-ray emissivity caused by the source ions.
+Since Vlasiator does not simulate the exosphere, we use the analytical model from
+Cravens et al. (2001) for the neutral density, given as
+nH(r) = 25
+�10 RE
+r
+�3
+,
+(3)
+with r the distance of the considered location to the Earth’s center. The above expres-
+sion gives nH in cm−3, and in the continuation of the study, we will express soft X-ray
+emissivity quantities using centimeter as the unit of length, following the common us-
+age for this specific application. For instance, Qloc will be expressed in keV cm−3 s−1 sr−1.
+In this study, we will use an interaction efficiency factor value of αX = 1×10−15 eV cm2,
+following Sun et al. (2019) and references therein.
+2.2.2
+Line-of-Sight Integration from a Virtual Spacecraft
+In order to simulate soft X-ray images close to as they would be obtained by a space-
+craft such as LEXI and SMILE, we place a virtual spacecraft in the Vlasiator simula-
+tion domain and calculate the line-of-sight-integrated value of the local soft X-ray emis-
+sivity along multiple viewing directions within the instrument field-of-view, which cor-
+responds to many pixels in the images. We define this quantity Qint(ϕ, λ) as a function
+of the azimuth ϕ and elevation λ of a given line of sight.
+Qint(ϕ, λ) =
+�
+Qloc(lϕ,λ) dlϕ,λ,
+(4)
+with lϕ,λ the distance from the spacecraft along the line of sight associated with the (ϕ, λ)
+azimuth–elevation pair. By convention, we will have the (0, 0) pair corresponding to the
+direction toward the Earth’s center. The obtained instantaneous line-of-sight emissiv-
+ity Qint will be given in keV cm−2 s−1 sr−1 in this study.
+Soft X-ray images simulated using this Vlasiator run will have an angular resolu-
+tion of 0.33◦ in both azimuth and elevation, and the line-of-sight integration will take
+place between the virtual satellite location and the outer boundary of the simulation do-
+main. If a given line of sight intersects the inner boundary, Qint will not be calculated.
+The virtual spacecraft will be placed either at (0, −30 RE, 0) in GSE coordinates (i.e.,
+observing from the dawnside) or at (0, 0, 30 RE) (i.e., observing from above the north
+pole). The two virtual satellites are selected to provide the side and polar views of the
+Earth’s magnetosphere similar to LEXI and SMILE, respectively, while keeping the ra-
+dial distances fixed at 30 RE between the apogees of LEXI and SMILE for easy compar-
+ison of soft X-ray signatures in polar and side views. Both instantaneous values of Qint
+and time-integrated ones over 300 s, Qint 300s, will be shown, as the latter correspond
+to the integration time that SMILE and LEXI soft X-ray imagers require for good signal-
+to-noise ratios (Branduardi-Raymont et al., 2018; Connor et al., 2021).
+–7–
+
+manuscript submitted to Earth and Planetary Physics
+3 Simulated Soft X-ray Images
+In this section, we present the simulated soft X-ray images obtained from the Vlasi-
+ator run when the virtual imaging spacecraft is viewing the dayside magnetosphere ei-
+ther from the dawnside or from above the north pole. We first show instantaneous im-
+ages alongside relevant plasma parameters in the noon–midnight meridional plane or equa-
+torial plane, respectively. Animations showing such instantanous images every second
+from t = 800 s to t = 1506 s are provided in supplementary material. We then time-
+integrate these instantaneous images during three 300 s time intervals and discuss the
+main features that can be observed.
+3.1 Instantaneous Simulated Soft X-ray Images
+3.1.1
+View from Dawn
+Figure 2 shows plasma parameters in the noon–midnight meridional plane (Y = 0)
+at t = 1100 s in the simulation. Figure 2a displays the proton number density, in which
+the magnetosheath and the polar cusps stand out. One can identify the bow shock, whose
+subsolar point lies approximately at 15 RE. Given the purely southward IMF orienta-
+tion, the bow shock is essentially quasi-perpendicular, which is why no ion foreshock is
+present. Within the magnetosheath, density irregularities are particularly prominent. We
+will show in Sect. 4.2 that these correspond to mirror-mode waves. The magnetopause
+subsolar point lies at about 10 RE.
+Figure 2b gives the proton bulk velocity. When crossing the bow shock, the solar
+wind slows down from its inflow speed of 750 km s−1 to velocities on the order of 200–
+300 km s−1. We can see that plasma convection at the dayside magnetopause gets faster
+at higher latitudes, before slowing down again when the plasma reaches the polar cusps.
+Figure 2c presents the local soft X-ray emissivity in the plane, Qloc (calculated with
+Eq. 1), which depends not only on the previous two parameters (np and Vp), but also
+on the plasma temperature and the neutral density, not shown in the figure. The most
+prominent features in Qloc are the polar cusps and the magnetosheath, with values reach-
+ing 4 × 10−10 keV cm−3 s−1 sr−1. In the magnetosheath, the wave field is particularly
+visible. Brighter areas along the dayside magnetopause can also be identified; we will show
+in Sect. 4.1 that these are flux transfer events (FTEs).
+Figure 2d shows the instantaneous image of line-of-sight-integrated soft X-ray emis-
+sions, Qint, viewed from a virtual imaging satellite placed at (0, −30 RE, 0), at this same
+time. The field of view associated with this image intersects the noon–midnight merid-
+ional plane within the red trapezoid in Fig. 2c. Elevation is the angle along the Z di-
+rection, whereas azimuth is along X. In this instantaneous image, the brightest areas
+correspond to the cusps and the magnetopause, with values of Qint on the order of 3 keV cm−2 s−1 sr−1.
+It was shown in previous studies that lines-of-sight with maximum brightness correspond
+to directions tangential to the magnetopause (Collier & Connor, 2018). The very bright
+area close to the inner boundary is caused by boundary effects, as proton density is el-
+evated at low latitudes near the inner boundary (Fig. 2a), due to leakage of cold plasma
+resulting from the boundary condition. However, the bright SWCX soft X-ray emission
+near the inner boundary does not exist in reality and is an artificial effect by calculat-
+ing soft X-ray emissivity based on proton parameters (Eq. 1).
+In the instantaneous image of Qint, one can notice signatures likely associated with
+the wave field in the magnetosheath, as well as brighter spots along the dayside mag-
+netopause. This means that, despite the line-of-sight integration and parallax effects, some
+of the structures identified in Qloc (Fig. 2c) also show in Qint. The possible relation be-
+tween such structures and transient processes will be investigated in Sect. 4.
+–8–
+
+manuscript submitted to Earth and Planetary Physics
+Figure 2.
+(a) Proton number density, (b) proton bulk velocity, and (c) local soft X-ray emis-
+sivity in the Y = 0 plane at t = 1100 s. (d) Instantaneous soft X-ray image with 1 s integration
+time from a virtual spacecraft placed at (0, −30 RE, 0) at t = 1100 s. The red trapezoid in panel
+(c) indicates the intersection of the instrument’s field of view shown in panel (d) with the Y = 0
+plane.
+–9–
+
+np [cm-3]
+Vp [kms-1]
+10
+900
+(a)
+(b)
+20
+20
+750
+10
+5
+10
+600
+[Re]
+[Re]
+0
+D
+0
+450
+Z
+Z
+300
+10
+2
+-10
+150
+-20
+-20
+0
+0
+10
+20
+0
+10
+20
+X[Re]
+X[Re]
+t = 1100 s
+4.0
+3.0
+(C)
+(d)
+30
+20
+3.3
+2.5
+20
+10
+2.7
+2.0
+10
+Elevation
+[Re]
+0
+D
+2.0
+0
+D
+1.5
+Z
+1.3
+-10
+1.0
+-10
+20
+0.7
+0.5
+-20
+-30
+0.0
+0.0
+0
+10
+20
+0
+10
+20
+30
+X[RE]
+Azimuth [omanuscript submitted to Earth and Planetary Physics
+Figure 3.
+(a) Proton number density, (b) proton bulk velocity, and (c) local soft X-ray emis-
+sivity in the Z = 0 plane at t = 1100 s. (d) Instantaneous soft X-ray image with 1 s integration
+time from a virtual spacecraft placed at (0, 0, 30 RE) at t = 1100 s. The red trapezoid in panel
+(c) indicates the intersection of the instrument’s field of view shown in panel (d) with the Z = 0
+plane.
+An animated version of Fig. 2 is provided in the supplementary material as Movie S1.
+In this animation, we can especially visualize how the magnetosheath wave signatures
+and dayside magnetopause bright spots move as a function of time. The fact that the
+latter are colocated with confined regions of increased proton density forming near the
+subsolar point and transiting along the magnetopause toward the polar cusps strongly
+suggests that these are signatures of FTEs (see Sect. 4.1 and discussion in Sect. 5).
+3.1.2
+View from North
+Figure 3 is analogous to Fig. 2, but this time the virtual imaging spacecraft has
+been placed above the north pole, at (0, 0, 30 RE) in GSE coordinates. The chosen time
+step for this instantaneous snapshot of plasma parameters and soft X-ray emissions is
+–10–
+
+np [cm-3]
+Vp [kms-1]
+10
+900
+(a)
+(b)
+20
+20
+750
+10
+5
+10
+600
+[Re]
+[Re]
+0
+450
+Y
+Y
+300
+10
+2
+-10
+150
+-20
+-20
+0
+0
+10
+20
+0
+10
+20
+X[Re]
+X [Re]
+t = 1100 s
+4.0
+3.0
+(C)
+(d)
+30
+20
+3.3
+2.5
+20
+10
+2.7
+2.0
+10
+Elevation
+[Re]
+D
+2.0
+1.5
+Y
+1.3
+-10
+1.0
+-10
+20
+0.7
+0.5
+-20
+-30
+0.0
+0.0
+0
+10
+20
+0
+10
+20
+30
+X[RE]
+Azimuth [omanuscript submitted to Earth and Planetary Physics
+Figure 4.
+Soft X-ray image obtained from a virtual spacecraft located at (0, −30 RE, 0) with
+300 s integration time starting at (a) t = 800 s, (b) t = 950 s, and (c) t = 1100 s in the simula-
+tion.
+t = 1100 s, like previously. Figures 3a–c show parameters in the equatorial plane (Z=0),
+and Fig. 3d shows the line-of-sight integrated soft X-ray emissions as a function of az-
+imuth (along the X direction) and elevation (this time along Y ).
+The bow shock and dayside magnetopause are again prominent in the proton den-
+sity panel (Fig. 3a), as are the magnetosheath waves. The proton bulk velocity panel in-
+dicates the abrupt decrease in velocity across the bow shock, while the magnetosheath
+plasma accelerates once it reaches the flanks (Fig. 3b). In local soft X-ray emissivity (Fig. 3c),
+it is again the magnetosheath that shows up the most (ignoring the artifacts near the
+inner boundary).
+In the instantaneous soft X-ray emission image (Fig. 3d), a bright arc is visible across
+elevations, corresponding to the dayside magnetopause observed tangentially. Beyond
+the magnetopause signature, one can also see elongated structures originating from the
+magnetosheath. Due to the viewing angle, polar cusps are not visible in this image. An
+animated version of Fig. 3 is provided in the supplementary material as Movie S2. In
+this animation, we can see how the elongated magnetosheath structures drift Earthward
+until they merge with the magnetopause signature, which occasionally brightens and ex-
+hibits undulations (e.g., around t = 1000 s and t = 1440 s).
+3.2 Time-Integrated Soft X-ray Images
+3.2.1
+View from Dawn
+While the instantaneous soft X-ray images presented above show many interest-
+ing features, they do not correspond to images which could be obtained by SMILE and
+LEXI, as the instruments will need an integration time on the order of 300 s to obtain
+a sufficient number of counts. We therefore produce time-integrated images over three
+300 s time intervals during the studied part of the simulation. We first consider the view
+from a virtual imaging spacecraft in the dawn sector; Figure 4 shows time-integrated soft
+X-ray images obtained during t = 800–1100 s (Fig. 4a), t = 950–1250 s (Fig. 4b), and
+–11–
+
+t = 800-1100 s
+t = 950-1250 s
+t = 1100-1400 s
+820
+(a)
+(b)
+(c)
+30
+30
+30
+700
+20
+20
+20
+sr-11
+580
+10
+10
+10
+N
+(along
+460
+0
+0
+0
+-10
+-10
+-10
+340
+-20
+-20
+-20
+220
+-30
+-30
+-30
+100
+0
+20
+0
+20
+0
+20
+Azimuth[°]
+Azimuth[°]
+Azimuth[°]
+(along X)
+(along X)
+(along X)manuscript submitted to Earth and Planetary Physics
+Figure 5.
+(a) Time-integrated soft X-ray image from t = 950 to t = 1250 s obtained from a
+virtual spacecraft located at (0, −30 RE, 0) (same as Fig. 4b), with several cuts along and across
+boundary region signatures. (b) Soft X-ray emission along the black cut through the northern-
+hemisphere magnetopause and magnetosheath signatures. (c) Same along the grey cut through
+the southern-hemisphere magnetopause and magnetosheath signatures. (d) Same along the red
+cut across the northern cusp signature. (e) Same along the orange cut across the southern cusp
+signature. (f) Same along the magenta cut following the dayside magnetopause signature. In pan-
+els b–f, the three lines correspond to the three studied 300 s time intervals for image acquisition.
+t = 1100–1400 s (Fig. 4c). In these images, the brightest structures correspond to the
+cusps as well as the dayside magnetopause.
+It is clear from the figures that the time integration blurs out most small-scale struc-
+tures that are associated with transient phenomena at the magnetopause and in the mag-
+netosheath. However, one can see some irregularity in the shape and brightness of the
+magnetopause signature, such as a dimmer and thinner region at elevations near −3◦ in
+Fig. 4a, and also at elevations near −7◦ in Fig. 4c, for instance (indicated with black ar-
+rows). Conversely, parts of the dayside magnetopause signature appear brighter, like re-
+gions near +3◦ and −9◦ elevations in Fig. 4a (indicated with white arrows). We will in-
+vestigate the possible connection between these brighter Qint 300s values at the dayside
+magnetopause and FTEs in Sect. 4.1.
+Besides, Figs 4b–c exhibit brighter stripes in their magnetosheath signatures (in-
+dicated with magenta arrows). Such stripes have oblique orientations with respect to the
+Earth–Sun line and can be seen in both the northern and the southern part of the do-
+main. In Sect. 4.2, we will show that these signatures are likely the result of mirror-mode
+waves in the magnetosheath.
+In order to better visualize differences between the three images, we will look at
+Qint 300s values along a few selected cuts. Fig. 5a reproduces the time-integrated image
+within t = 950–1250 s (i.e., Fig. 4b) and indicates where five cuts have been performed,
+along which the Qint 300s values are extracted and displayed in Figs. 5b–f for the three
+time intervals.
+Figs. 5b and 5c show time-integrated soft X-ray emission values along the black and
+grey straight lines, respectively, which both cut across the magnetopause and through
+the magnetosheath. The horizontal axis is the angular distance, noted d, of a point on
+–12–
+
+t = 950-1250 s
+820
+800
+(a)
+800
+800-1100 s
+C
+30
+600
+950-1250 s
+600
+[keV cm-}
+1100-1400
+400
+400
+700
+200
+200
+20
+18 21 24 27 30 33 36
+18 21 24 27 30 33 36
+d[°]
+[。]p
+b
+580
+10
+levation[
+N
+Qint_300s
+(d)
+(e)
+700
+700
+(along )
+-2
+460
+[keV cm'
+0
+600
+600
+500
+500
+E
+-10
+50
+55
+60
+65
+-65
+-60
+-55
+-50
+340
+ESL[°]
+ESL[°]
+-20
+Qint_300s
+800
+220
+(f)
+700
+-30
+600
+100
+500
+0
+10
+20
+30
+-70
+-50
+-30
+-10
+10
+30
+50
+70
+EsL[°]
+Azimuth[°]
+(along X)manuscript submitted to Earth and Planetary Physics
+Figure 6.
+Same as Fig. 4 but with the virtual spacecraft placed at (0, 0, 30 RE).
+the line from the (0, 0) viewing direction (calculated as
+�
+ϕ2 + λ2). In both panels, Qint 300s
+peaks at d ≈ 21◦, which corresponds to the tangent direction to the magnetopause. While
+the curves corresponding to the three integrating time intervals are almost perfectly su-
+perimposed on top of each other in Fig. 5b, one can see that the peak was slightly re-
+duced and drifted to larger d values with time in Fig. 5c. Around d ≈ 24◦, there are
+small-amplitude fluctuations visible in the violet and orange curves, in both panels, cor-
+responding to the stripes identified in Figs. 4b–c. A last observation that can be made
+from these two panels is a change in the slope of Qint 300s occurring at d ≈ 36◦, indi-
+cated with black arrows. This change of slope is likely associated with the line of sight
+tangent to the bow shock, as shown in Fig. 4 of Connor et al. (2021). The correspond-
+ing locations in Fig. 5a are indicated with white arrows.
+Values of Qint 300s along the red and orange curved lines, cutting through the north-
+ern and southern cusp, respectively, are shown in Figs. 5d and 5e. The horizontal axis
+is the angle between the Earth–point direction and the Earth–Sun line, noted θESL. In
+both hemispheres, the cusp signature drifted slightly poleward with time, although the
+differences are relatively small.
+Figure 5f give the Qint 300s values along the magenta curve, which essentially fol-
+lows the magnetopause signature in the images, as a function of θESL. It shows quite clearly
+the variations in Qint 300s previously identified in Fig. 4 and enables estimating the am-
+plitude of those variations, which can reach up to 90 keV cm−2 sr−1 (orange curve peak-
+to-trough amplitude at θESL within [−20◦, −10◦]), i.e. representing about 10% of the
+average value. Other peaks can be identified at θESL ≈ ±55◦, corresponding to loca-
+tions in the high-altitude polar cusps.
+3.2.2
+View from North
+By applying the same methodology as above, we can also produce time-integrated
+soft X-ray images as would be obtained from the virtual imaging spacecraft placed above
+the north pole. Figure 6 presents the three images corresponding to the same integra-
+tion intervals (i.e, t = 800–1100 s, t = 950–1250 s, and t = 1100–1400 s).
+In the three panels, the brightest signature comes from the dayside magnetopause,
+with brightness maximizing at locations corresponding to low nose angles and signatures
+–13–
+
+t = 800-1100 s
+t = 950-1250 s
+t = 1100-1400 s
+820
+(a)
+(b)
+(c)
+30
+30
+30
+700
+20
+20
+20
+sr-11
+580
+10
+10
+10
+(along )
+0
+0
+0
+460
+-10
+-10
+-10
+340
+-20
+-20
+-20
+220
+-30
+-30
+-30
+100
+0
+20
+0
+20
+0
+20
+Azimuth[°]
+Azimuth[°]
+Azimuth[°]
+(along X)
+(along X)
+(along X)manuscript submitted to Earth and Planetary Physics
+Figure 7.
+(a) Time-integrated soft X-ray image from t
+=
+950 to t
+=
+1250 s obtained from
+a virtual spacecraft located at (0, 0, 30 RE) (same as Fig. 6b), with several cuts along and across
+boundary region signatures. (b) Soft X-ray emission along the black cut through the post-noon
+magnetopause and magnetosheath signatures. (c) Same along the grey cut through the pre-noon
+magnetopause and magnetosheath signatures. (d) Same along the magenta cut following the day-
+side magnetopause signature. In panels b–d, the three lines correspond to the three studied 300 s
+time intervals for image acquisition.
+becoming fainter toward the flanks of the magnetopause. Magnetosheath stripes also show
+from this perspective, and are visible in all three panels (indicated with magenta arrows),
+with mostly oblique orientations. Beyond the magnetopause signature near the flanks,
+a second, fainter line parallel to it can be seen, especially in Figs. 6b–c (indicated with
+black arrows). This parallel line might be associated with the same processes as the stripes
+closer to the subsolar point, as they appear similar in terms of brightness and orienta-
+tion.
+Figure 7 shows the values of Qint 300s along a few selected cuts, in a same way as
+in Fig. 5. Figure 7a reproduces Fig. 5b and indicates where three cuts are considered to
+study the variations of Qint 300s along the magnetopause X-ray signature as well as across
+the magnetosheath in the pre-noon and post-noon sectors.
+We can see from Figs. 7b–c that, along the cuts crossing the magnetosheath, the
+integrated soft X-ray emissivity peaks at d ≈ 21◦ in both the post-noon and pre-noon
+sectors, respectively, with a slight reduction in the peak value and very slight outward
+shift with time. This is consistent with the trend observed in the virtual images obtained
+with the view from dawn, and can be interpreted as a slight sunward motion of the day-
+side magnetopause. Sunward from the peak (i.e. at larger d values), fluctuations in Qint 300s
+can be seen, almost reaching a secondary peak at d ≈ 23◦ (more visible in the dawnside
+cut, Fig. 7c). This corresponds to the oblique stripes indicated with magenta arrows in
+Fig. 6. Like in Figs. 5b–c, we can notice a change in the slope of the curves in Figs. 7b–
+c, around d ≈ 33◦ (indicated with black arrows). The corresponding locations in the soft
+X-ray image (Fig. 7a) are indicated with white arrows. As mentioned previously, this
+slope change is likely associated with lines of sight tangent to the bow shock (Connor
+et al., 2021).
+–14–
+
+t = 950-1250 s
+820
+(a)
+800
+(b)
+800-1100 s
+800
+(c)
+30
+ sr-1]
+950-1250 s
+1100-1400 s
+Qint_300s
+600
+600
+[keV cm-2 s
+700
+20
+400
+400
+580
+200
+10
+200
+。.
+levation[
+18 21 24 27 30 33 36
+18 21 24 27 30 33 36
+460
+[。]p
+[。]p
+0
+E
+(d)
+-10
+340
+800
+Qint_300s
+700
+-20
+220
+600
+-30
+100
+500
+-50
+0
+10
+20
+30
+-70
+-30
+-10
+10
+30
+50
+70
+EsL[°]
+Azimuth[°]
+(along X)manuscript submitted to Earth and Planetary Physics
+Finally, looking at Qint 300s values along the magenta cut following the magnetopause
+signature (Fig. 7d), we note that the signal roughly follows linear slopes in the flanks (|θESL| > 30◦),
+with little to no evolution with time. On the other hand, the section corresponding to
+the nose of the magnetopause exhibits fluctuations and variability amounting to about
+5% of the Qint 300s values (up to 40 keV cm−2 sr−1 between the violet and the orange lines
+at θESL ≈ −20◦). In a given image (i.e., focusing on a single line in Fig. 7d), fluctua-
+tions can lead to the appearance of several local maxima in Qint 300s, denoting the ex-
+istence of structures near the magnetopause nose. These structures might be associated
+with the nature of the reconnection taking place at the dayside magnetopause, which can
+either be over an extended X-line or on the contrary more patchy (Atz et al., 2022; Walsh
+et al., 2017).
+4 Signatures of Transient Processes in Soft X-ray Images
+In this section, we will investigate to what extent transient processes taking place
+in the magnetosheath and at the dayside magnetopause can induce signatures in soft X-
+ray images as will be measured by SMILE and LEXI SXI. We will focus on two processes:
+flux transfer events forming by magnetic reconnection at the dayside magnetopause and
+mirror-mode waves developing in the magnetosheath.
+4.1 Flux Transfer Events at the Magnetopause
+In Vlasiator 2D–3V runs driven by IMF with a southward Bz component, flux trans-
+fer events regularly form at the magnetopause near the subsolar point (Hoilijoki et al.,
+2017, 2019), and they propagate poleward toward the cusps, with consequences in terms
+of energy transfer into the magnetosphere (Ala-Lahti et al., 2022) and proton precipi-
+tation into the cusps (Grandin et al., 2020).
+In this 3D–3V run, FTEs also form near the subsolar point and gradually follow
+the magnetopause toward the cusps, in both hemispheres. This can be seen in Supple-
+mentary Movie S3, where FTEs appear as regions of increased proton density enclosed
+within a magnetic island. To investigate whether these FTEs can have a signature on
+soft X-ray images, we first find a simple way to detect and track them along the day-
+side magnetopause, and we then look at their time-integrated parameters over 300 s win-
+dows.
+Figure 8a shows (background color), as a function of time in the simulation, the
+magnetic field component along the radial direction (away from the Earth), Br, along
+the magenta cut following the magnetopause signature shown in Fig. 5a. More precisely,
+the values of Br are taken along the intersection of the lines of sight forming this cut with
+the Y = 0 plane (noon–midnight meridional plane), and they approximately correspond
+to the magnetic field component normal to the magnetopause surface. One can see that,
+at a given value of θESL along the intersected cut, Br exhibits pseudo-oscillations with
+amplitudes up to ±10 nT. These correspond to bipolar signatures as FTEs transit across
+the location. In the northern hemisphere, when the leading edge of a FTE approaches
+a given location at the dayside magnetopause, Br is increased (red values in the plot),
+and as the trailing edge of the FTE passes the location, Br shows a negative deflection
+(blue values). Due to the geometry, the situation is opposite in the southern hemisphere.
+Hence, we can see how FTEs tend to form within a few degrees from the subsolar point
+at the dayside magnetopause, and how they propagate toward one of the polar cusps in
+∼150 s. It is noteworthy that the propagation time of the identified FTEs is shorter than
+the image acquisition time which is retained in this study. However, one may want to
+determine whether FTEs can still create a signature in soft X-ray images integrated over
+300 s.
+–15–
+
+manuscript submitted to Earth and Planetary Physics
+Figure 8.
+(a) Background color: Time evolution of the radial (i.e., nearly normal to the mag-
+netopause) component of the magnetic field along the intersection of the magenta cut following
+the magnetopause signature in Fig. 5a with the Y
+= 0 plane. FTEs generated near the subsolar
+point create a bipolar signature moving toward higher magnetopause angles as a function of time,
+with opposite polarity in opposite hemispheres. Isocontours: Time evolution of the line-of-sight-
+integrated proton number density along the magenta cut in Fig. 5a. (b) Proton number density
+along the intersection of the cut with the Y
+= 0 plane, averaged over 300 s for the three studied
+time intervals (starting at t = 800 s, t = 950 s and t = 1100 s). (c) Value of the 300 s-integrated
+soft X-ray emission along the cut in the corresponding three images shown in Fig. 4 (same data
+as Fig. 5f).
+–16–
+
+800-1100 s
+950-1250 s
+Br [nT]
+(a)
+1100-1400 s
+10.0
+8.3
+(b)
+(c)
+40
+40
+40
+7.5
+8.0
+5.0
+20
+20
+20
+2.5
+7.7
+0
+0.0
+0
+0
+7.4
+-2.5
+-20
+-20
+-20
+-5.0
+0
+7.1
+-7.5
+-40 :
+-40
+-40
+-10.0
+6.8
+800
+900 1000 1100 1200 1300 1400 1500
+2.0 2.5 3.0
+700 750 800
+[np[1010cm-2]
+Time [s]
+(np)300s [cm-3]
+Qint_300s [keV cm-2 sr-1]manuscript submitted to Earth and Planetary Physics
+In Fig. 8b, we show the time-averaged value of the proton number density, ⟨np⟩300s,
+along the intersection of the cut with the Y = 0 plane (i.e., at the locations where Br
+is monitored). The time averaging is done over 300 s intervals corresponding to the in-
+tegration time of the three images shown in Fig. 4. It can be seen that (i) the time-averaged
+density along the magnetopause exhibits some variations as a function of θESL, and (ii) these
+variations change from one 300 s interval to another. The largest differences occur at θESL
+values comprised between 0◦ and −20◦ during the third time interval (1100–1400 s). Dur-
+ing this time interval, three bipolar signatures can be identified in Fig. 8a at those θESL
+values. We can see in Movie S3 that these FTEs are indeed associated with higher-than-
+background proton number densities.
+Figure 8c reproduces the corresponding values of Qint 300s during the three time
+intervals along the cut in the soft X-ray images (same data as in Fig. 5f). While the match
+between the spatial and temporal variations in ⟨np⟩300s and Qint 300s is not perfect, the
+corresponding curves do nonetheless exhibit striking resemblance. This suggests that the
+presence of FTEs at the dayside magnetopause during the image acquisition time can
+affect the average proton density along corresponding lines of sight and induce a signa-
+ture in soft X-ray emission values. The imperfect match can be expected given that Qint 300s
+values result from a line-of-sight integration, whereas the ⟨np⟩300s values are taken lo-
+cally in the Y = 0 plane.
+To investigate this further, overlaid with the Br data in Fig. 8a are isocontours of
+the proton number density integrated along the lines of sight forming the magenta cut
+in Fig. 5a. As for the Qint 300s shown in Fig. 8c, this line-of-sight-integrated density likely
+contains effects of variations occurring outside of the Y = 0 plane; however, one can
+see a tendency for density enhancements to follow the pattern formed by the bipolar sig-
+natures associated with FTEs in the Y = 0 plane. This suggests that the proton den-
+sity enhancements occurring in the core of the simulated FTEs can produce soft X-ray
+signatures even in line-of-sight and time-integrated images. It is worth mentioning that,
+in observations, FTEs often exhibit a density decrease in their core (Akhavan-Tafti et
+al., 2018). Yet, some observed FTEs do show a proton density increase and might there-
+fore produce signatures in soft X-ray images similar to those described here (see Sect. 5)
+4.2 Mirror-Mode Waves in the Magnetosheath
+In Movie S3, one can identify wave-like structures in the magnetosheath consist-
+ing of patches of increased proton number density appearing in the central part of the
+magnetosheath (Earthward from the bow shock) and slowly drifting toward the mag-
+netopause. In 2D–3V Vlasiator simulations, mirror-mode waves have been identified in
+the magnetosheath (Hoilijoki et al., 2016; Dubart et al., 2020). In a nearly 3D–3V setup
+wherein a noon–midnight meridional-plane slice with a thickness of 7RE along the dawn–
+dusk dimension was simulated with Vlasiator, Pfau-Kempf et al. (2020) identified an-
+ticorrelated magnetic field and proton density fluctuations in the magnetosheath, attributed
+to mirror-mode waves. We will first check whether in this 3D–3V run those waves ex-
+hibit similar properties as mirror-mode waves, after which we will investigate to what
+extent they can create signatures in soft X-ray images.
+Figure 9 presents the magnetic field magnitude (Fig. 9a–b), proton number den-
+sity (Fig. 9c–d) and local soft X-ray emissivity (Fig. 9e–f) along the intersections of the
+black and grey cuts in Fig. 5a with the Y = 0 plane. These data are shown at four time
+steps in the simulation: t = 1000, 1010, 1020, and 1030 s. Here, the horizontal axis cor-
+responds to the X coordinate of the cut intersection points in the Y = 0 plane to give
+a better intuition of spatial scales.
+At X < 9 RE, the magnetic field magnitude decreases steadily and the proton den-
+sity is on the order of 0.5 cm−3, which corresponds to magnetosphere field and plasma.
+As X increases, the magnetic field magnitude drops more drastically, whereas the pro-
+–17–
+
+manuscript submitted to Earth and Planetary Physics
+Figure 9.
+Investigation of the magnetosheath waves signatures along the black and grey cuts
+in Fig. 5a. The shown variables are taken along the intersections of the cuts with the Y
+=
+0
+plane at four time steps in the simulation (i.e., these are instantaneous values). (a–b) Magnetic
+field magnitude. (c–d) Proton number density. (e–f) Local soft X-ray emissivity.
+–18–
+
+Northern hemisphere
+Southern hemisphere
+60
+t = 1000 s
+60
+[BI [nT]
+t = 1010 s
+t = 1020 s
+40
+40
+t = 1030 s
+20
+20
+(a)
+(b)
+4
+4
+(d)
+C
+3
+3
+2
+2
+du
+1
+1
+0
+0
+1e-10
+1e-10
+4
+4
+(e)
+(f)
+3
+2
+2
+[keV cm-
+1
+1
+0
+0
+8
+9
+10
+11
+12
+8
+9
+10
+11
+12
+X[Re]
+X[Re]manuscript submitted to Earth and Planetary Physics
+Figure 10.
+Time-averaged values of the magnetosheath parameters shown in Fig. 9 over 300 s
+intervals starting at t = 800 s, t = 950 s, and t = 1100 s. (a–b) Time-averaged magnetic field
+magnitude along the intersections of the black and grey cuts with the Y
+= 0 plane, respectively.
+(c–d) Same for the time-averaged proton number density. (e–f) Same for the time-averaged lo-
+cal soft X-ray emissivity. (g–h) Soft X-ray emission values along the black and grey cuts in the
+300 s-integrated images shown in Fig. 4. The last two panels show the same data as Fig. 5b–c.
+ton density increases. This corresponds to the crossing of the magnetopause. Beyond the
+magnetopause, one can notice fluctuations in both |B| and np, with sizes on the order
+of 1–2 RE along the X coordinate. These fluctuations are more prominent in the north-
+ern (black) cut than in the southern (grey) cut, and there is a fairly clear anticorrela-
+tion in the variations of both parameters. Besides, we can see that the structures asso-
+ciated with these fluctuations appear to drift Earthward by about 1 RE in 40 s, giving
+an estimated speed along the X direction of ∼150 km s−1. Comparing with Fig. 2b, we
+can infer that the structures therefore approximately drift alongside the plasma. The an-
+ticorrelation between magnetic field and density variations and the fact that the waves
+are roughly static in the plasma frame strongly suggest that these waves are indeed mir-
+ror modes.
+–19–
+
+Northern hemisphere
+Southern hemisphere
+60
+60
+800-1100 s
+950-1250 s
+1100-1400 s
+[nT]
+40
+40
+ (a)
+20
+20
+[(b)
+3
+(c)
+(d)
+3
+(np/300s
+2
+2
+1
+1
+0
+0
+1e-10
+1e-10
+3
+3
+(e)
+(f)
+(Qioc/300 s
+2
+2
+[keV cm-
+1
+1
+0
+0
+800
+800
+(g)
+(h)
+600
+600
+400
+400
+18
+20
+22
+24
+26
+28
+18
+20
+22
+24
+26
+28
+[.p
+[。]pmanuscript submitted to Earth and Planetary Physics
+Given the significant propagation of the mirror-mode waves over the span of 40 s
+seen in Fig. 9, one cannot expect that individual wave fronts will be visible in an image
+integrated over 300 s. However, like for FTEs, one can investigate whether mirror-mode
+wave activity in the magnetosheath can nevertheless produce a signature in soft X-ray
+images. Figure 10 presents this analysis along the same cuts as Fig. 9, but this time for
+time-averaged parameters along the cut, again using the same three 300 s time intervals
+as previously. In all panels of this figure, the horizontal axis corresponds to the angu-
+lar distance d to the origin of the azimuth–elevation frame used in soft X-ray images.
+In Figs. 10a–b, we show the time-averaged magnetic field magnitude along the cuts
+during the three intervals. While the magnetosphere and magnetopause values do not
+differ significantly during the three intervals, one can see that the magnetosheath fluc-
+tuations, associated with the mirror-mode waves identified previously, differ slightly from
+each other. The differences appear more clearly in the time-averaged proton density pan-
+els (Figs. 10c–d), especially at d ≈ 21–26◦, and the resulting time-averaged local soft
+X-ray emission values (⟨Qloc⟩300s, Figs. 10e–f) in the Y = 0 plane are consistent with
+average density fluctuations. To assess the effect of these local time-averaged fluctua-
+tions in actual soft X-ray images, Figs. 10g–h show the soft X-ray emissions integrated
+along the line of sight corresponding to the two cuts (same data as in Fig. 5b–c, respec-
+tively). In these panels, we can see that the three lines corresponding to the three time
+intervals do not fully overlap within d ≈ 22–26◦.
+5 Discussion
+First of all, it is worth comparing the soft X-ray emissivity values obtained with
+the Vlasiator simulation to those reported in modeling studies using different models.
+Sun et al. (2019) carried out a series of soft X-ray emission simulations using a MHD model,
+driven by various types of solar wind conditions. Our driving conditions correspond clos-
+est to their first case, with a solar wind density of 5 cm−3, a solar wind speed of 400 km s−1,
+and a purely southward IMF of 5 nT magnitude. They obtained a soft X-ray image from
+a virtual imaging satellite approximately positioned above the north pole (analogous to
+our Fig. 3d), in which the magnetopause signature reached a value of 13 keV cm−2 s−1 sr−1.
+In contrast, with our Vlasiator run, we get a value close to 3 keV cm−2 s−1 sr−1, which
+is almost a factor of 5 lower. This is consistent with our solar wind density being 5 times
+lower than theirs, leading to a magnetosheath density accordingly lower. Our values are
+also in fairly good agreement with those obtained by Connor et al. (2021) using a sim-
+ulation from the OpenGGCM global MHD model and providing soft X-ray images seen
+from the dawnside (analogous to our Fig. 2d).
+While Vlasiator is too computationally expensive to be run over a long time inter-
+val, its unique advantage is that the hybrid-kinetic description of the space plasma en-
+ables the formation of transient features which cannot emerge in MHD simulations. We
+can therefore simulate the possible effect of such phenomena in soft X-ray images of the
+near-Earth space as will be obtained by SMILE and LEXI. Considering first the case of
+FTEs, if we focus on the largest peak-to-trough variation of Qint 300s in Fig. 8c, corre-
+sponding to the 1100–1400 s time interval and θESL angles between −20◦ and 0◦, we ob-
+tain ∆Qint 300s ≈ 90 keV cm−2 sr−1, amounting to about 12% of the background value
+along the dayside magnetopause. The corresponding spatial variations in proton den-
+sity in the Y = 0 plane amount to 0.81 cm−3, on the order of 30% of the background
+value. This can provide quantitative elements to assess whether a soft X-ray imager might
+be able to detect signatures associated with FTEs, depending on the specifications of the
+soft X-ray sensor.
+FTEs have been found to form at the dayside magnetopause quasi-periodically ap-
+pearing approximately once every 8 minutes (Rijnbeek et al., 1984) preferably forming
+during southward IMF conditions (e.g., Berchem & Russell, 1984). A statistical study
+–20–
+
+manuscript submitted to Earth and Planetary Physics
+of 55 FTEs using MMS observations found an average size of a subsolar FTE to be 1700 ± 400 km
+(Akhavan-Tafti et al., 2018), which is 3 to 7 times smaller than the FTEs observed at
+the higher latitudes (Akhavan-Tafti et al., 2018; Fermo et al., 2011; Y. L. Wang et al.,
+2005). However, the scale sizes of the observed FTEs vary from ion-scale flux ropes (e.g.,
+Eastwood et al., 2016; Dong et al., 2017) to FTEs with a diameter up to 1–2 RE (Rijnbeek
+et al., 1984; Fear et al., 2007; Hasegawa et al., 2006), while their axial length can be much
+longer (e.g., Fear et al., 2008).
+A typical FTE has an internal structure with twisted field lines and axial magnetic
+field leading to an increase in total magnetic field magnitude and a decrease in the den-
+sity in the core of the FTE (e.g., Akhavan-Tafti et al., 2018; Zhang et al., 2010). How-
+ever, “crater” FTEs with a decrease in total magnetic field with a higher density in the
+middle have been observed (Zhang et al., 2010). A statistical study of 18 typical and 14 crater
+FTEs comparing FTE densities (nFTE) and magnetic field strength (BFTE) to the mag-
+netosheath density (nMS) and magnetic field (BMS) showed that 75% of typical FTEs
+have the density ratio nFTE/nMS < 0.5 and 80% of the FTEs have BFTE > BMS, while
+90% of the crater FTEs had nFTE/nMS > 0.5 (and approximately 50% nFTE > nMS) and
+approximately 95% had smaller magnetic field magnitude than the surrounding magne-
+tosheath (Zhang et al., 2010). In this simulation, the identified FTEs have a higher-than-
+background proton density in their core, which makes them analogous to crater FTEs.
+We can therefore speculate that soft X-ray images might be able to capture signatures
+of crater FTEs, even more so if their peak-to-trough density variation is similar to or greater
+than that of FTEs identified in this simulation.
+It is however clear from the presented results that FTEs forming in this Vlasiator
+run move across a 300 s integrated soft X-ray image too fast for them to be individu-
+ally visible in the image. Rather, the local soft X-ray emission enhancements in some
+parts of the magnetopause signature are likely due to the cumulative effect of several FTEs,
+especially if these tend to remain longer at a given latitude. Hoilijoki et al. (2019) showed
+that the speed at which FTEs propagate toward higher latitudes along the magnetopause
+depends on the driving conditions. We can therefore speculate that, in some cases where
+the FTE motion is slower than in this simulation, their signatures in soft X-ray images
+might be more prominent than in the results obtained here.
+Regarding mirror-mode waves in the magnetosheath, we can estimate the effect of
+the largest density enhancement visible in Fig. 10c. Considering the difference between
+the ⟨np⟩300s values during the first and third time intervals at d = 24◦, we find that it
+consists of a local density enhancement by 0.4 cm−3 (∼14%) in the Y
+= 0 plane as-
+sociated with mirror-mode waves structures. The corresponding difference in time-integrated
+soft X-ray emissions amounts to about 20 keV cm−2 sr−1 (∼4%).
+Mirror-mode structures are large-scale compressional waves that are non-propagating
+in the plasma frame. They are often observed in the Earth’s magnetosheath especially
+behind the quasi-perpendicular bow shock. They may appear as quasi-sinusoidal oscil-
+lations in the magnetic field but mirror-mode structures are often observed as peaks in
+the mirror-unstable plasma and dips in mirror-stable conditions near the magnetopause
+(Soucek et al., 2008). The depth of the mirror structures has been found to increase with
+the decreasing distance to the magnetopause (Soucek et al., 2008). The analysis of 2 months
+of Cluster data by Soucek et al. (2008) showed that the average period of the mirror-
+mode waves was approximately 12 s (with 98% of the observed structures being between
+4 to 24 s). A statistical study by Lucek et al. (2001) showed that the scale sizes of the
+mirror structures can vary from less than 600 km along the local magnetopause normal,
+750–1000 km along the maximum variance direction to 1500–3000 km along the flow di-
+rection. The amplitude of the magnetic field variations associated with mirror-mode waves
+is typically on the order of several 10% of the background field, and the associated pro-
+ton density variations can lead to up to multi-fold local enhancements (e.g., Chandler
+et al., 2021).
+–21–
+
+manuscript submitted to Earth and Planetary Physics
+The differences in Qint 300s values obtained above are estimates under the driving
+conditions and the setup used in the Vlasiator run on which this study is based. These
+choices and constraints have several consequences on the properties of the mirror-mode
+waves. First of all, in our simulation, the spatial resolution in the magnetosheath (2000 km)
+likely leads to mirror-mode structures being larger than in reality. Therefore, one may
+speculate that SMILE and LEXI observations of mirror-mode structures would likely be
+smaller and might hence be blurred out more easily than in this simulation.
+Besides, Dubart et al. (2020) showed that, when the spatial resolution of a Vlasi-
+ator simulation is too coarse, the unstable proton distributions in the magnetosheath en-
+tirely pump their free energy into the growth of mirror-mode structures, instead of trans-
+ferring part of it to electromagnetic ion cyclotron waves, which require a finer spatial res-
+olution to emerge. It is worth noting, however, that the solar wind driving conditions
+used in the run are not the most typical ones. Indeed, the solar wind density is notably
+lower than average, whereas the solar wind velocity and temperatures are higher than
+average. This suggests that, under solar wind conditions closer to average values than
+in this Vlasiator run, such as those from Sun et al. (2019) discussed above, larger sig-
+nal than what we obtained can be expected. This, combined with the fact that density
+fluctuations associated with mirror-mode waves can be greater than in our simulation
+(factor of 3 in the event analyzed by Chandler et al. (2021)), suggests that the constraints
+on FTEs and mirror-mode waves for them to produce signatures in the soft X-ray ob-
+servations of LEXI and SMILE might actually be less strict than suggested by our re-
+sults. Additional simulations with a finer spatial resolution and different solar wind driv-
+ing are needed to test this hypothesis.
+Another point worth discussing is that the Vlasiator run duration analyzed for this
+study consists of a 700 s time interval only. In reality, both LEXI and SMILE SXI in-
+struments will observe processes over significantly longer time scales, which could open
+more possibilities for the data analysis. For instance, calculating the discrepancy between
+two time-integrated images separated by a few minutes could enhance contrast in regions
+where the plasma conditions have changed during that time frame. While this would likely
+still not enable the detection and monitoring of individual structures given the required
+image acquisition time, this could provide additional approaches to indirectly identify
+such processes in near-Earth space, as well as context for in-situ observations.
+Finally, one should point out the fact that, in this Vlasiator run, the driving con-
+ditions are steady. As a consequence, one may expect that the large-scale features such
+as the boundary locations (bow shock, magnetopause, polar cusps) do not change much
+during the studied time interval. Nevertheless, Figs. 5b–e indicate that the magnetopause
+X-ray signature exhibits a slight outward motion with time while the peaks in the cusp
+signatures move toward higher latitudes. This suggests that the analysis of successive
+soft X-ray images might enable the tracking of boundary motions. To confirm this, it
+would therefore prove interesting to monitor the motion of these boundaries in a Vlasi-
+ator run in which solar wind conditions are varying. This is a future avenue for an up-
+coming study which could be made possible once a Vlasiator run with time-varying driv-
+ing conditions is available.
+6 Conclusions
+In this paper, we presented the first estimates of SWCX soft X-ray emission imag-
+ing of the dayside near-Earth space based on a hybrid-Vlasov simulation. We used ∼700 s
+of a Vlasiator run driven by constant solar wind of low density and high velocity and with
+purely southward IMF with Bz = −5 nT. From the Vlasiator simulation outputs, we
+calculated at every second the local soft X-ray emissivity based on the plasma param-
+eters and produced line-of-sight-integrated instantaneous soft X-ray images as would be
+obtained from a virtual imaging spacecraft placed at a distance of 30 RE from the Earth’s
+–22–
+
+manuscript submitted to Earth and Planetary Physics
+center, providing the polar and side views of the Earth’s magnetosphere similar to the
+SMILE and LEXI views, respectively. We then integrated those instantaneous images
+over 300 s intervals to reproduce anticipated observational requirements for the SMILE
+and LEXI SXI instruments. The main results of this study are as follows:
+1. The most prominent features in soft X-ray images obtained from above the pole
+or on the dawnside consist of the magnetopause and polar cusps (the latter only
+when observing from the dawnside). The soft X-ray emissivity values obtained with
+Vlasiator are consistent with earlier MHD results, taking into account differences
+in the solar wind driving conditions.
+2. Despite the 300 s integration time, the obtained soft X-ray images exhibit smaller-
+scale features such as the brightening of localized areas at the dayside magnetopause
+and wave-like patterns in the magnetosheath signature. These features are cre-
+ated by transient phenomena, namely FTEs and mirror-mode waves, respectively.
+3. Based on the simulated soft X-ray images, one can anticipate that FTE signatures
+could amount to 12% of the background signal if FTEs occurring during the 300 s
+interval cumulatively lead to local proton density enhancements by about 30%.
+This result holds for “crater” FTEs with enhanced proton density in their core
+(Zhang et al., 2010); possible soft X-ray signatures associated with typical FTEs
+(which generally exhibit a decrease in proton density in their core) need further
+investigation to be characterized.
+4. Correspondingly, soft X-ray images of SMILE and LEXI might reveal the pres-
+ence of mirror-mode waves, as in the simulation the associated density structures
+in the magnetosheath cumulatively led to an enhancement by 14% in the proton
+density, locally, and produced an increase in soft X-ray emission signal by 4%. De-
+pending on the scale size, propagation speed and amplitude of the variations in
+the proton density, mirror-mode structures may lead to more pronounced or on
+the contrary more blurred out signatures in soft X-ray images.
+5. The above values are likely conservative estimates, given that the solar wind con-
+ditions in this Vlasiator run are representative of solar wind high-speed streams
+rather than the more common average solar wind driving which typically has larger
+proton number density.
+The results of this study contribute to the SMILE Modeling Working Group ob-
+jectives by comparing simulation results using a hybrid-Vlasov code with other works
+based on MHD. Moreover, the hybrid-kinetic description of the space plasma provides
+a unique opportunity to anticipate under what conditions transient processes in the day-
+side near-Earth space might be observed by soft X-ray imagers on SMILE and LEXI.
+Future plans could include the monitoring of the motion of the large-scale space plasma
+boundaries (magnetopause, polar cusps, bow shock) in a different run with time-varying
+upstream conditions.
+Acknowledgments
+We acknowledge the European Research Council for starting grant 200141-QuESpace,
+with which the Vlasiator model was developed, and consolidator grant 682068-PRESTISSIMO
+awarded for further development of Vlasiator and its use in scientific investigations. We
+gratefully acknowledge Academy of Finland grant numbers 338629-AERGELC’H, 339756-
+KIMCHI, 336805-FORESAIL, and 335554-ICT-SUNVAC. The Academy of Finland also
+supported this work through the PROFI4 grant (grant number 3189131). Hyunju K. Con-
+nor gratefully acknowledges support from the NASA grants, 80NSSC20K1670 and 80MSFC20C0019,
+and the NASA GSFC FY23 IRAD and HIF funds. The CSC – IT Center for Science and
+the PRACE Tier-0 supercomputer infrastructure in HLRS Stuttgart (grant number 2019204998)
+are acknowledged as they made these results possible. The authors wish to thank the
+–23–
+
+manuscript submitted to Earth and Planetary Physics
+Finnish Grid and Cloud Infrastructure (FGCI) for supporting this project with compu-
+tational and data storage resources.
+The authors declare that they have no conflict of interest.
+The Vlasiator simulation data used in the study amount to 18 TB of disk space;
+access to the raw simulation data can be granted by following the Vlasiator data access
+policy (see https://www2.helsinki.fi/en/researchgroups/vlasiator/rules-of-the
+-road). The Vlasiator code is preserved at https://zenodo.org/record/3640593 (Pfau-
+Kempf et al., 2022), available in open access. It is developed openly at https://github
+.com/fmihpc/vlasiator.
+References
+Akhavan-Tafti, M., Slavin, J. A., Le, G., Eastwood, J. P., Strangeway, R. J., Russell,
+C. T., . . . Burch, J. L.
+(2018).
+Mms examination of ftes at the earth’s sub-
+solar magnetopause.
+Journal of Geophysical Research: Space Physics, 123(2),
+1224-1241. doi: 10.1002/2017JA024681
+Ala-Lahti, M., Pulkkinen, T. I., Pfau-Kempf, Y., Grandin, M., & Palmroth, M.
+(2022).
+Energy Flux Through the Magnetopause During Flux Transfer Events
+in Hybrid-Vlasov 2D Simulations.
+Geophysical Research Letters, 49(19),
+e2022GL100079. doi: 10.1029/2022GL100079
+Atz, E. A., Walsh, B. M., Broll, J. M., & Zou, Y. (2022). The spatial extent of mag-
+netopause magnetic reconnection from in situ themis measurements. Journal of
+Geophysical Research: Space Physics, 127(12), e2022JA030894.
+doi: 10.1029/
+2022JA030894
+Battarbee, M., Blanco-Cano, X., Turc, L., Kajdiˇc, P., Johlander, A., Tarvus,
+V., . . . Palmroth, M.
+(2020).
+Helium in the Earth’s foreshock: a global
+Vlasiator survey.
+Annales Geophysicae, 38(5), 1081-1099.
+doi: 10.5194/
+angeo-38-1081-2020
+Battarbee, M., Ganse, U., Pfau-Kempf, Y., Turc, L., Brito, T., Grandin, M., . . .
+Palmroth, M.
+(2020).
+Non-locality of Earth’s quasi-parallel bow shock: injec-
+tion of thermal protons in a hybrid-Vlasov simulation.
+Annales Geophysicae,
+38(3), 625-643. doi: 10.5194/angeo-38-625-2020
+Berchem, J., & Russell, C. T.
+(1984).
+Flux transfer events on the magnetopause -
+Spatial distribution and controlling factors.
+Journal of Geophysical Research,
+89, 6689-6703. doi: 10.1029/JA089iA08p06689
+Branduardi-Raymont, G., Wang, C., Escoubet, C. P., Adamovic, M., Agnolon,
+D., Berthomier, M., . . . Zhu, Z.
+(2018).
+SMILE definition study re-
+port (Red Book).
+European Space Agency, ESA/SCI, 1.
+doi: 10.5270/
+esa.smile.definition study report-2018-12
+Carter, J. A., Sembay, S., & Read, A. M.
+(2010).
+A high charge state coronal mass
+ejection seen through solar wind charge exchange emission as detected by
+XMM-Newton.
+Monthly Notices of the Royal Astronomical Society, 402(2),
+867-878. doi: 10.1111/j.1365-2966.2009.15985.x10.48550/arXiv.0911.0897
+Carter, J. A., Sembay, S., & Read, A. M.
+(2011).
+Identifying XMM-Newton ob-
+servations affected by solar wind charge exchange - Part II.
+Astronomy & As-
+trophysics, 527, A115.
+doi: 10.1051/0004-6361/20101581710.48550/arXiv.1101
+.1848
+Chandler, M. O., Schwartz, S. J., Avanov, L. A., Coffey, V. N., Giles, B. L., Moore,
+T. E., . . . Torbert, R. B.
+(2021).
+Observations of Mirror Mode Structures
+in the Dawn Side Magnetosphere.
+Journal of Geophysical Research (Space
+Physics), 126(2), e28649. doi: 10.1029/2020JA028649
+Collier, M. R., & Connor, H. K.
+(2018).
+Magnetopause Surface Reconstruction
+From Tangent Vector Observations.
+Journal of Geophysical Research (Space
+Physics), 123(12), 10,189-10,199. doi: 10.1029/2018JA025763
+–24–
+
+manuscript submitted to Earth and Planetary Physics
+Connor, H. K., & Carter, J. A. (2019). Exospheric Neutral Hydrogen Density at the
+Nominal 10 RE Subsolar Point Deduced From XMM-Newton X-Ray Obser-
+vations.
+Journal of Geophysical Research (Space Physics), 124(3), 1612-1624.
+doi: 10.1029/2018JA026187
+Connor, H. K., Sibeck, D. G., Collier, M. R., Baliukin, I. I., Branduardi-Raymont,
+G., Brandt, P. C., . . . Zoennchen, J. H. (2021). Soft X ray and ENA Imaging
+of the Earth’s Dayside Magnetosphere. Journal of Geophysical Research (Space
+Physics), 126(3), e28816. doi: 10.1029/2020JA028816
+Cox, D. P. (1998). Modeling the local bubble. In D. Breitschweidt, M. J. Freyberg,
+& J. Tr¨umper (Eds.), The local bubble and beyond (p. 121).
+Springer-Verlag,
+New York.
+Cravens, T. E., Robertson, I. P., & Snowden, S. L.
+(2001).
+Temporal variations
+of geocoronal and heliospheric X-ray emission associated with the solar wind
+interaction with neutrals.
+Journal of Geophysical Research, 106(A11), 24883-
+24892. doi: 10.1029/2000JA000461
+Dong, X.-C., Dunlop, M. W., Trattner, K. J., Phan, T. D., Fu, H.-S., Cao, J.-B., . . .
+Burch, J. L.
+(2017).
+Structure and evolution of flux transfer events near day-
+side magnetic reconnection dissipation region: MMS observations.
+Geophysical
+Research Letters, 44(12), 5951-5959. doi: 10.1002/2017GL073411
+Dubart, M., Ganse, U., Osmane, A., Johlander, A., Battarbee, M., Grandin, M.,
+. . . Palmroth, M.
+(2020).
+Resolution dependence of magnetosheath waves in
+global hybrid-Vlasov simulations. Annales Geophysicae, 38(6), 1283-1298. doi:
+10.5194/angeo-38-1283-2020
+Eastwood, J. P., Phan, T. D., Cassak, P. A., Gershman, D. J., Haggerty, C.,
+Malakit, K., . . . Wang, S.
+(2016).
+Ion-scale secondary flux ropes generated
+by magnetopause reconnection as resolved by MMS.
+Geophysical Research
+Letters, 43(10), 4716–4724. doi: 10.1002/2016GL068747
+Fear, R. C., Milan, S. E., Fazakerley, A. N., Lucek, E. A., Cowley, S. W. H., & Dan-
+douras, I.
+(2008, August).
+The azimuthal extent of three flux transfer events.
+Annales Geophysicae, 26, 2353-2369. doi: 10.5194/angeo-26-2353-2008
+Fear, R. C., Milan, S. E., Fazakerley, A. N., Owen, C. J., Asikainen, T., Taylor,
+M. G. G. T., . . . Daly, P. W.
+(2007).
+Motion of flux transfer events: a
+test of the Cooling model.
+Annales Geophysicae, 25(7), 1669–1690.
+doi:
+10.5194/angeo-25-1669-2007
+Fermo, R. L., Drake, J. F., Swisdak, M., & Hwang, K.-J.
+(2011).
+Comparison of a
+statistical model for magnetic islands in large current layers with Hall MHD
+simulations and Cluster FTE observations.
+Journal of Geophysical Research:
+Space Physics, 116(A9). doi: 10.1029/2010JA016271
+Grandin, M., Battarbee, M., Osmane, A., Ganse, U., Pfau-Kempf, Y., Turc, L., . . .
+Palmroth, M. (2019). Hybrid-Vlasov modelling of nightside auroral proton pre-
+cipitation during southward interplanetary magnetic field conditions.
+Annales
+Geophysicae, 37, 791–806. doi: 10.5194/angeo-37-791-2019
+Grandin, M., Turc, L., Battarbee, M., Ganse, U., Johlander, A., Pfau-Kempf, Y.,
+. . . Palmroth, M.
+(2020).
+Hybrid-Vlasov simulation of auroral proton precip-
+itation in the cusps: Comparison of northward and southward interplanetary
+magnetic field driving.
+Journal of Space Weather and Space Climate, 10, 51.
+doi: 10.1051/swsc/2020053
+Hasegawa, A. (1969). Drift mirror instability of the magnetosphere. Physics of Flu-
+ids, 12, 2642-2650. doi: 10.1063/1.1692407
+Hasegawa, H., Sonnerup, B. U. O., Owen, C. J., Klecker, B., Paschmann, G.,
+Balogh, A., & R`eme, H.
+(2006).
+The structure of flux transfer events
+recovered from cluster data.
+Annales Geophysicae, 24(2), 603–618.
+Re-
+trieved from https://angeo.copernicus.org/articles/24/603/2006/
+doi:
+10.5194/angeo-24-603-2006
+Hoilijoki, S., Ganse, U., Pfau-Kempf, Y., Cassak, P. A., Walsh, B. M., Hietala, H.,
+–25–
+
+manuscript submitted to Earth and Planetary Physics
+. . . Palmroth, M. (2017). Reconnection rates and X line motion at the magne-
+topause: Global 2D-3V hybrid-Vlasov simulation results.
+Journal of Geophysi-
+cal Research (Space Physics), 122(3), 2877-2888. doi: 10.1002/2016JA023709
+Hoilijoki, S., Ganse, U., Sibeck, D. G., Cassak, P. A., Turc, L., Battarbee, M., . . .
+Palmroth, M.
+(2019).
+Properties of Magnetic Reconnection and FTEs on the
+Dayside Magnetopause With and Without Positive IMF Bx Component Dur-
+ing Southward IMF. Journal of Geophysical Research (Space Physics), 124(6),
+4037-4048. doi: 10.1029/2019JA026821
+Hoilijoki, S., Palmroth, M., Walsh, B. M., Pfau-Kempf, Y., vonˆA Alfthan, S., Ganse,
+U., . . . Vainio, R.
+(2016).
+Mirror modes in the Earth’s magnetosheath: Re-
+sults from a global hybrid-Vlasov simulation.
+Journal of Geophysical Research
+(Space Physics), 121(5), 4191-4204. doi: 10.1002/2015JA022026
+Jung, J., Connor, H. K., Carter, J. A., Koutroumpa, D., Pagani, C., & Kuntz, K. D.
+(2022).
+Solar Minimum Exospheric Neutral Density Near the Subsolar Magne-
+topause Estimated From the XMM Soft X-Ray Observations on 12 November
+2008.
+Journal of Geophysical Research (Space Physics), 127(3), e29676.
+doi:
+10.1029/2021JA029676
+Juusola, L., Pfau-Kempf, Y., Ganse, U., Battarbee, M., Brito, T., Grandin, M.,
+. . . Palmroth, M.
+(2018).
+A possible source mechanism for magneto-
+tail current sheet flapping.
+Annales Geophysicae, 36, 1027-1035.
+doi:
+10.5194/angeo-36-1027-2018
+Londrillo, P., & del Zanna, L. (2004). On the divergence-free condition in Godunov-
+type schemes for ideal magnetohydrodynamics: the upwind constrained
+transport method.
+Journal of Computational Physics, 195(1), 17-48.
+doi:
+10.1016/j.jcp.2003.09.016
+Lucek, E. A., Dunlop, M. W., Horbury, T. S., Balogh, A., Brown, P., Cargill, P., . . .
+Oddy, T.
+(2001, October).
+Cluster magnetic field observations in the magne-
+tosheath: four-point measurements of mirror structures.
+Annales Geophysicae,
+19, 1421-1428. doi: 10.5194/angeo-19-1421-2001
+Palmroth, M., Ganse, U., Pfau-Kempf, Y., Battarbee, M., Turc, L., Brito, T.,
+. . . von Alfthan, S.
+(2018).
+Vlasov methods in space physics and as-
+trophysics.
+Living Reviews in Computational Astrophysics, 4, 1.
+doi:
+10.1007/s41115-018-0003-2
+Papadakis, K., Pfau-Kempf, Y., Ganse, U., Battarbee, M., Alho, M., Grandin, M.,
+. . . Palmroth, M. (2022). Spatial filtering in a 6d hybrid-vlasov scheme to alle-
+viate adaptive mesh refinement artifacts: a case study with vlasiator (versions
+5.0, 5.1, and 5.2.1). Geoscientific Model Development, 15(20), 7903–7912. doi:
+10.5194/gmd-15-7903-2022
+Pfau-Kempf, Y., Palmroth, M., Johlander, A., Turc, L., Alho, M., Battarbee, M.,
+. . . Ganse, U.
+(2020).
+Hybrid-Vlasov modeling of three-dimensional day-
+side magnetopause reconnection.
+Physics of Plasmas, 27(9), 092903.
+doi:
+10.1063/5.0020685
+Pfau-Kempf, Y., von Alfthan, S., Sandroos, A., Ganse, U., Koskela, T., Bat-
+tarbee, M., . . . Alho, M.
+(2022).
+fmihpc/vlasiator: Vlasiator [dataset].
+Zenodo.
+Retrieved from https://zenodo.org/record/3640593
+doi:
+10.5281/ZENODO.3640593
+Rijnbeek, R. P., Cowley, S. W. H., Southwood, D. J., & Russell, C. T.
+(1984).
+A survey of dayside flux transfer events observed by ISEE 1 and 2 mag-
+netometers.
+Journal of Geophysical Research, 89, 786-800.
+doi: 10.1029/
+JA089iA02p00786
+Robertson, I. P., Collier, M. R., Cravens, T. E., & Fok, M. C.
+(2006).
+X-ray
+emission from the terrestrial magnetosheath including the cusps.
+Jour-
+nal of Geophysical Research (Space Physics), 111(A12), A12105.
+doi:
+10.1029/2006JA011672
+Robertson, I. P., & Cravens, T. E.
+(2003a).
+Spatial maps of heliospheric and geo-
+–26–
+
+manuscript submitted to Earth and Planetary Physics
+coronal X-ray intensities due to the charge exchange of the solar wind with
+neutrals.
+Journal of Geophysical Research (Space Physics), 108(A10), 8031.
+doi: 10.1029/2003JA009873
+Robertson, I. P., & Cravens, T. E.
+(2003b).
+X-ray emission from the terrestrial
+magnetosheath.
+Geophysical Research Letters, 30(8), 1439.
+doi: 10.1029/
+2002GL016740
+Russell, C. T., & Elphic, R. C. (1978). Initial ISEE magnetometer results - Magne-
+topause observations. Space Sci. Rev., 22, 681-715. doi: 10.1007/BF00212619
+Sibeck, D. G., Allen, R., Aryan, H., Bodewits, D., Brandt, P., Branduardi-Raymont,
+G., . . . Wing, S.
+(2018).
+Imaging Plasma Density Structures in the Soft X-
+Rays Generated by Solar Wind Charge Exchange with Neutrals. Space Science
+Reviews, 214(4), 79. doi: 10.1007/s11214-018-0504-7
+Snowden, S. L., Collier, M. R., & Kuntz, K. D.
+(2004).
+XMM-Newton Observation
+of Solar Wind Charge Exchange Emission.
+The Astrophysical Journal, 610(2),
+1182-1190. doi: 10.1086/42184110.48550/arXiv.astro-ph/0404354
+Snowden, S. L., Freyberg, M. J., Plucinsky, P. P., Schmitt, J. H. M. M., Tr¨umper,
+J., Voges, W., . . . Sanders, W. T. (1995). First maps of the soft X-ray diffuse
+background from the ROSAT XRT/PSPC all-sky survey.
+The Astrophysical
+Journal, 454, 643. doi: 10.1086/176517
+Soucek, J., Lucek, E., & Dandouras, I.
+(2008).
+Properties of magnetosheath mir-
+ror modes observed by cluster and their response to changes in plasma pa-
+rameters.
+Journal of Geophysical Research: Space Physics, 113(A4).
+doi:
+10.1029/2007JA012649
+Sun, T., Wang, C., Connor, H. K., Jorgensen, A. M., & Sembay, S. (2020). Deriving
+the Magnetopause Position from the Soft X-Ray Image by Using the Tangent
+Fitting Approach.
+Journal of Geophysical Research (Space Physics), 125(9),
+e28169. doi: 10.1029/2020JA028169
+Sun, T., Wang, C., Sembay, S. F., Lopez, R. E., Escoubet, C. P., Branduardi-
+Raymont, G., . . . Guo, Y. H.
+(2019).
+Soft X-ray Imaging of the Magne-
+tosheath and Cusps Under Different Solar Wind Conditions: MHD Simula-
+tions.
+Journal of Geophysical Research (Space Physics), 124(4), 2435-2450.
+doi: 10.1029/2018JA026093
+Tarvus, V., Turc, L., Battarbee, M., Suni, J., Blanco-Cano, X., Ganse, U., . . . Palm-
+roth, M.
+(2021).
+Foreshock cavitons and spontaneous hot flow anomalies: a
+statistical study with a global hybrid-Vlasov simulation. Annales Geophysicae,
+39(5), 911-928. doi: 10.5194/angeo-39-911-2021
+Turc, L., Ganse, U., Pfau-Kempf, Y., Hoilijoki, S., Battarbee, M., Juusola, L., . . .
+Palmroth, M.
+(2018).
+Foreshock Properties at Typical and Enhanced Inter-
+planetary Magnetic Field Strengths: Results From Hybrid-Vlasov Simulations.
+Journal of Geophysical Research (Space Physics), 123(7), 5476-5493.
+doi:
+10.1029/2018JA025466
+von Alfthan, S., Pokhotelov, D., Kempf, Y., Hoilijoki, S., Honkonen, I., Sandroos,
+A., & Palmroth, M.
+(2014).
+Vlasiator: First global hybrid-Vlasov simula-
+tions of Earth’s foreshock and magnetosheath.
+Journal of Atmospheric and
+Solar-Terrestrial Physics, 120, 24-35. doi: 10.1016/j.jastp.2014.08.012
+Walsh, B. M., Komar, C. M., & Pfau-Kempf, Y.
+(2017).
+Spacecraft measurements
+constraining the spatial extent of a magnetopause reconnection X line.
+Geo-
+physical Research Letters, 44(7), 3038-3046. doi: 10.1002/2017GL073379
+Wang, C., & Sun, T.
+(2022).
+Methods to derive the magnetopause from soft X-ray
+images by the SMILE mission.
+Geoscience Letters, 9(1), 30.
+doi: 10.1186/
+s40562-022-00240-z
+Wang, Y. L., Elphic, R. C., Lavraud, B., Taylor, M. G. G. T., Birn, J., Raeder,
+J., . . . Friedel, R. H.
+(2005).
+Initial results of high-latitude magnetopause
+and low-latitude flank flux transfer events from 3 years of Cluster observa-
+tions.
+Journal of Geophysical Research (Space Physics), 110, A11221.
+doi:
+–27–
+
+manuscript submitted to Earth and Planetary Physics
+10.1029/2005JA011150
+Zhang, H., Kivelson, M. G., Khurana, K. K., McFadden, J., Walker, R. J., An-
+gelopoulos, V., . . . Auster, H. U.
+(2010).
+Evidence that crater flux transfer
+events are initial stages of typical flux transfer events.
+Journal of Geophysical
+Research: Space Physics, 115(A8). doi: 10.1029/2009JA015013
+Zhang, Y., Sun, T., Wang, C., Ji, L., Carter, J. A., Sembay, S., . . . Zhao, X.
+(2022).
+Solar Wind Charge Exchange Soft X-Ray Emissions in the Mag-
+netosphere during an Interplanetary Coronal Mass Ejection Compared to
+Its Driven Sheath.
+The Astrophysical Journal Letters, 932(1), L1.
+doi:
+10.3847/2041-8213/ac7521
+–28–
+
diff --git a/WdFQT4oBgHgl3EQfbzbk/content/tmp_files/load_file.txt b/WdFQT4oBgHgl3EQfbzbk/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e809356713734cac588ca0c7e1aebe39cd817e5b
--- /dev/null
+++ b/WdFQT4oBgHgl3EQfbzbk/content/tmp_files/load_file.txt
@@ -0,0 +1,1501 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf,len=1500
+page_content='manuscript submitted to Earth and Planetary Physics Hybrid-Vlasov simulation of soft X-ray emissions at the Earth’s dayside magnetospheric boundaries M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Grandin1, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Connor2, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hoilijoki1, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Battarbee1, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Pfau-Kempf1, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Ganse1, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Papadakis1, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 1Department of Physics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' University of Helsinki,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Helsinki,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Finland 2NASA Goddard Space Flight Center,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Greenbelt,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' MD,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 20771,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' USA 3Space and Earth Observation Centre,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Finnish Meteorological Institute,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Helsinki,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Finland Key Points: We produce soft X-ray images of near-Earth space with a global hybrid-Vlasov simulation with southward interplanetary magnetic field Flux transfer events can produce X-ray signatures despite being transient phenom- ena if they cumulatively increase the proton density locally Mirror-mode structures in the magnetosheath can also produce soft X-ray signa- tures in time-integrated images Corresponding author: Maxime Grandin,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' maxime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='grandin@helsinki.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='fi –1– arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='13325v1 [physics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='space-ph] 30 Jan 2023 manuscript submitted to Earth and Planetary Physics Abstract Solar wind charge exchange produces emissions in the soft X-ray energy range which can enable the study of near-Earth space regions such as the magnetopause, the magnetosheath and the polar cusps by remote sensing techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Solar wind–Magnetosphere–Ionosphere Link Explorer (SMILE) and Lunar Environment heliospheric X-ray Imager (LEXI) mis- sions aim to obtain soft X-ray images of near-Earth space thanks to their Soft X-ray Im- ager (SXI) instruments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While earlier modeling works have already simulated soft X-ray images as might be obtained by SMILE SXI during its mission, the numerical models used so far are all based on the magnetohydrodynamics description of the space plasma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To investigate the possible signatures of ion-kinetic-scale processes in soft X-ray images, we use for the first time a global hybrid-Vlasov simulation of the geospace from the Vlasi- ator model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The simulation is driven by fast and tenuous solar wind conditions and purely southward interplanetary magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We first produce global X-ray images of the day- side near-Earth space by placing a virtual imaging satellite at two different locations, providing meridional and equatorial views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We then analyze regional features present in the images and show that they correspond to signatures in soft X-ray emissions of mirror- mode wave structures in the magnetosheath and flux transfer events (FTEs) at the mag- netopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Our results suggest that, although the time scales associated with the mo- tion of those transient phenomena will likely be significantly smaller than the integra- tion time of of the SMILE and LEXI imagers, mirror-mode structures and FTEs can cu- mulatively produce detectable signatures in the soft X-ray images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' For instance, a lo- cal increase by 30% in the proton density at the dayside magnetopause resulting from the transit of multiple FTEs leads to a 12% enhancement in the line-of-sight- and time- integrated soft X-ray emissivity originating from this region.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Likewise, a proton density increase by 14% in the magnetosheath associated with mirror-mode structures can re- sult in an enhancement in the soft X-ray signal by 4%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These are likely conservative es- timates, given that the solar wind conditions used in the Vlasiator run can be expected to generate weaker soft X-ray emissions than the more common denser solar wind.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These results will contribute to the preparatory work for the SMILE and LEXI missions by pro- viding the community with quantitative estimates of the effects of small-scale, transient phenomena occurring on the dayside.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 1 Introduction Over the past two decades, interest has grown in studying near-Earth space by ob- serving the soft X-ray emissions created by charge-exchange interactions between heavy, multiply charged ions and neutral species.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' For example, charge-exchange interactions between O7+ or O8+ ions present in the solar wind and neutral hydrogen atoms from the Earth’s exosphere are known to lead to photon emissions through the de-excitation of the product ion species, which include photon energies in the soft X-ray range (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5– 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 keV).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This process is known as solar wind charge exchange (SWCX).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A review by Sibeck et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018) provides in-depth details on the processes at play as well as pioneer- ing works on the study of terrestrial and planetary space through the observation of soft X-ray emissions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Currently, several satellite missions aim at imaging the geocorona in soft X-rays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Two major upcoming missions making use of SWCX soft X-ray imaging to reseach near- Earth space are the Lunar Environment heliospheric X-ray Imager (LEXI), led by the Boston University in collaboration with NASA GSFC and several universities (http:// sites.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='bu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='edu/lexi), and the Solar wind–Magnetosphere–Ionosphere Link Explorer (SMILE) mission, designed jointly by the European Space Agency and the Chinese Academy of Sciences (Branduardi-Raymont et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Both of these missions will provide soft X- ray images of the polar cusps and the magnetosheath, which are known to be the most prominent sources of SWCX soft X-ray emissions in near-Earth space, with a goal of un- derstanding global interaction between the solar wind and the Earth’s magnetosphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –2– manuscript submitted to Earth and Planetary Physics After its expected launch in 2024, LEXI will observe the dayside magnetosheath from the lunar surface for up to two weeks of a short mission period due to the harsh lunar environment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' SMILE will be launched in 2025 into a highly elliptical polar orbit with an apogee of ∼20 Earth radii (RE), observing the dayside magnetosheath and cusps for up to 40 continuous hours per orbit during three years of its mission period.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While there have been no wide-field-of-view soft X-ray observations of the geospace, the existence of near-Earth SWCX soft X-ray emissions has been studied in various works using data from the XMM and ROSAT astrophysics missions (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Carter et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2010, 2011;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Cravens et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2001;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Snowden et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2004;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Connor & Carter, 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Jung et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Subsequently, modeling efforts were undertaken to understand the soft X-ray emissions detected by the ROSAT mission (Snowden et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 1995) and to quan- tify the contributions of various source mechanisms (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cox, 1998;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Cravens et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2001).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Robertson and Cravens (2003b) were the first to produce images of the dayside magne- topause and magnetosheath in soft X-ray emissions through the SWCX mechanism us- ing a numerical model, and based on their results they suggested that it might be pos- sible to make use of those emissions to monitor the solar wind–magnetosphere interac- tions through remote sensing observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The same authors also produced simulated images in soft X-rays for a virtual instrument placed on the Earth’s surface and observ- ing various directions in the sky (Robertson & Cravens, 2003a), and investigated the po- lar cusps’ signatures in soft X-ray images in further simulations (Robertson et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' More recent studies have taken up the modeling efforts as part of the preparatory phases of the LEXI and SMILE missions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Collier and Connor (2018) determined that the surface of the magnetopause could in principle be determined by soft X-ray imag- ing of near-Earth space, as the line-of-sight-integrated soft X-ray emissions maximize for observations tangential to the magnetopause surface.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Based on these results, Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020) developed a tangent-fitting method to derive the magnetopause position in an- ticipated images from the Soft X-ray Imager (SXI) instrument onboard SMILE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Alter- native methods to determine the magnetopause shape and position from SXI observa- tions have been devised by C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Wang and Sun (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Among the processes taking place in near-Earth space and which could be studied using soft X-ray imaging, Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019) showed that SXI will be able to monitor the motions of the dayside magnetopause and of the polar cusps in response to changes in the solar wind driving conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Connor et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021) also addressed similar research questions by simulating soft X-ray images which could be obtained during the SMILE and LEXI missions in an event when the in- terplanetary magnetic field (IMF) turns southward and an event when the solar wind density abruptly increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' They demonstrated that it might be possible to get infor- mation on the time variations in the magnetic reconnection rate at the dayside magne- topause using such remote-sensing techniques, and they argued that these missions could enable the tracing of the energy flow from the solar wind all the way to the cusps with the same instrument.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The above-mentioned modeling studies all used numerical models of space plasma based on the magnetohydrodynamics (MHD) paradigm, which treats the plasma as a fluid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To complement the results obtained with this approach, it would be interesting to carry out modeling studies of SWCX soft X-ray imaging based on a global model of near-Earth space relying on a kinetic description of the plasma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Indeed, a certain num- ber of space plasma processes are intrinsically of kinetic origin, and therefore cannot emerge in MHD simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Examples of such processes include the growth of instabilities, wave– particle interactions as well as ion-scale physics associated with magnetic reconnection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this paper, we will for the first time carry out a study of soft X-ray emissions relying on a 3D global hybrid-Vlasov simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We will make use of the kinetic nature of the model to investigate to what extent near-Earth space processes associated with ion-scale physics might produce signatures in soft X-ray images as will be obtained by the SMILE SXI instrument.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To that aim, we will make use of Vlasiator (Palmroth et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', –3– manuscript submitted to Earth and Planetary Physics 2018), which is a global hybrid-Vlasov model of near-Earth space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In particular, we will quantify the signatures in terms of line-of-sight- and time-integrated soft X-ray emissions of two transient phenomena occurring on the dayside: mirror-mode waves in the mag- netosheath and flux transfer events (FTEs) produced by magnetic reconnection at the dayside magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Mirror-mode waves are generated by temperature anisotropy in the magnetosheath and appear as nonpropagating structures (in the plasma frame) with anticorrelated variations in plasma density and magnetic field magnitude (Hasegawa, 1969).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' FTEs are flux-rope-like structures forming along the dayside magnetopause in presence of multiple or bursty reconnection lines associated with southward IMF conditions, which often form near the subsolar point and propagate toward the polar cusps (Russell & El- phic, 1978;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Rijnbeek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 1984;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Berchem & Russell, 1984).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The paper is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Vlasiator model and the methodology to construct soft X-ray images in the utilized simulation run are introduced in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Then, in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3, we present examples of global images of the dayside near-Earth space obtained with virtual imaging spacecraft placed at two locations: on the dawnside (meridional view) and above the north pole (equatorial view).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Section 4 describes the analysis and results on the detectability of transient processes (FTEs and mirror-mode waves) in the sim- ulated soft X-ray images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A discussion of the results is provided in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5, and Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 6 gives a summary of the main conclusions of the paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2 Models and Methods 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 Vlasiator 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 Model Vlasiator (von Alfthan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2014;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018) is a global hybrid-Vlasov code simulating near-Earth plasma at ion-kinetic scales.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the hybrid-Vlasov approach, ions are described through their velocity distribution function (VDF) which is discretized in a 3-dimensional (3D) velocity grid, resulting in ion populations evolving in a 6D phase space (3D in ordinary space + 3D in velocity space).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In practice, Vlasiator solves the Vlasov equation for ions and treats electrons as a massless, charge-neutralizing fluid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While most existing Vlasiator simulation runs consider protons as the sole ion species, heav- ier ions such as He2+ have also been included in a few runs (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, Blanco-Cano, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Electromagnetic fields are propagated in time by solving Maxwell’s equations un- der the Darwin approximation, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', neglecting the displacement current term in Amp`ere’s law.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The equation system is closed through a generalized Ohm’s law including the Hall term and a polytropic (adiabatic) description of the electron pressure gradient term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The model considers an ideal geomagnetic dipole field using a nonscaled strength 8·1015 Tm3 and with a zero tilt between its axis and the Z direction in the Geocentric Solar Eclip- tic (GSE) frame of reference — in the GSE frame, the Earth’s center is at the origin, the X axis points toward the Sun, the Z axis toward the north, and the Y axis completes the orthonormal frame and points toward dusk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The geomagnetic dipole is described through a vector potential, which is scaled to zero at the inflow boundary in order to prevent mag- netic divergence entering the simulation domain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Magnetic and electric fields are prop- agated using a finite difference upwind field solver (Londrillo & del Zanna, 2004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The simulation domain extent varies from one run to another, but it generally in- cludes the dayside magnetosphere, the magnetosheath, the bow shock, the ion foreshock (if driving conditions allow for its existence), and part of the magnetotail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This has al- lowed for a great variety of studies focusing on a wide range of processes, such as fore- shock waves (Turc et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018), foreshock cavitons (Tarvus et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2021), bow shock non- locality (Battarbee, Ganse, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2020), magnetosheath waves (Hoilijoki et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Dubart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2020), energy transfer across the magnetopause (Ala-Lahti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2022), magne- –4– manuscript submitted to Earth and Planetary Physics totail current sheet flapping (Juusola et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018), or auroral proton precipitation (Grandin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2019, 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While those past studies were based on 2D–3V runs (2D in ordinary space, 3D in velocity space), recent code developments have enabled the production of the first Vlasiator 3D–3V runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This study makes use of one such 3D–3V run, described below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 Simulation Run The Vlasiator run used in this study has its simulation domain extending from −110 RE to 50 RE in the X direction (with RE the Earth’s radius;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' RE = 6371 km) and confined within |Y | < 58 RE and |Z| < 58 RE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The inner boundary lies at 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 RE from the ori- gin;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' it is implemented as a perfectly conducting sphere on which VDFs are fixed Maxwellian distributions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' External boundaries have Neumann boundary conditions, except for the +X wall from which the solar wind and IMF enter the simulation domain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Driving conditions in this 3D–3V run are as follows: purely southward IMF with Bx = By = 0 and Bz = −5 nT, solar wind with proton number density of 1 cm−3, speed along the −X direction at 750 km s−1, temperature of 500 kK.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Solar wind con- ditions are homogeneous and constant throughout the simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Protons are the sole ion species in this run.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Initial conditions are such that the whole simulation domain is filled with solar-wind-like Maxwellian VDFs and the superposition of the dipole geomag- netic field with the IMF, and the near-Earth space regions thus form self-consistently during the first few hundred seconds of the simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this study, we will focus on the time interval starting at t = 800 s, when the dayside magnetosphere is well formed and lasting until the end of the simulation at t = 1506 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this run, an output file was written at a cadence of 1 s, which contains the plasma bulk parameters as well as elec- tromagnetic field components in every simulation cell.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' One development which made 3D–3V runs possible was the implementation of adap- tive mesh refinement (AMR) for the ordinary-space mesh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this run, least-refined re- gions have a base grid with 8000 km resolution, and there are three refinement levels at 4000, 2000 and 1000 km resolution to improve the description of regions of interest where ion-scale kinetic processes are important (bow shock, magnetosheath, magnetopause, mag- netotail current sheet).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The velocity space is a uniform 3D Cartesian grid with a reso- lution of 40 km s−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The electric and magnetic fields are solved on a uniform Cartesian grid at constant resolution of 1000 km in a process described by Papadakis et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 1 gives an overview of the simulation domain at t = 1100 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It consists of three slices in the X = 0, Y = 0 and Z = 0 planes showing the proton number den- sity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While the chosen viewing angle and slice extents enable seeing dayside structures and processes relevant to this study, the actual simulation domain extends beyond this figure (see above).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the figure, one can identify the bow shock, the magnetosheath ex- hibiting wave-like structures, the magnetopause as well as the northern polar cusp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The magnetotail, partly hidden behind the X = 0 slice, will not be the focus of this study but exhibits complex dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 Soft X-ray Image Generation 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 Analytical Expression of Soft X-ray Emissivity In this study, the method to derive the local soft X-ray emissivity within the sim- ulation will follow the same approach as past studies (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Connor et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is calculated at position r in the simulation domain with the following ex- pression Qloc(r) = αX 4π np(r)nH(r)Veff(r), (1) –5– manuscript submitted to Earth and Planetary Physics Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Proton number density in the X = 0, Y = 0 and Z = 0 planes of the simulation domain at t = 1100 s in the Vlasiator simulation, wherein the bow shock, the magnetopause and the northern polar cusp are prominent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' One can also identify wave-like structures in the dayside magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –6– t=1100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 s - origin at (0,0,0)[RE) Tick every 10 Re nproton [cm-3i 5 Z[RE] 4 3 2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 X [RE] 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 Y [RE] 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1manuscript submitted to Earth and Planetary Physics where αX is the interaction efficiency factor, np is the proton number density, nH is the neutral hydrogen atom density, and Veff is the so-called “effective velocity”, defined as Veff(r) = � Vp(r)2 + 5 3 kBT(r) mp .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2) In Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2), Vp is the proton bulk velocity, kB is Boltzmann’s constant, T is the plasma temperature, and mp is the proton mass.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This equation expresses Veff in m s−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Note that protons do not emit soft X-rays.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The highly charged, heavy solar wind ions like C6+, N6+, N7+, Ne9+, S10+, O7+, and O8+ emit soft X-rays through SWCX (Sibeck et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' By multiplying the interaction efficiency factor αX, the proton-based quantity in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1) is transformed into the soft X-ray emissivity caused by the source ions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Since Vlasiator does not simulate the exosphere, we use the analytical model from Cravens et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2001) for the neutral density, given as nH(r) = 25 �10 RE r �3 , (3) with r the distance of the considered location to the Earth’s center.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The above expres- sion gives nH in cm−3, and in the continuation of the study, we will express soft X-ray emissivity quantities using centimeter as the unit of length, following the common us- age for this specific application.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' For instance, Qloc will be expressed in keV cm−3 s−1 sr−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this study, we will use an interaction efficiency factor value of αX = 1×10−15 eV cm2, following Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019) and references therein.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 Line-of-Sight Integration from a Virtual Spacecraft In order to simulate soft X-ray images close to as they would be obtained by a space- craft such as LEXI and SMILE, we place a virtual spacecraft in the Vlasiator simula- tion domain and calculate the line-of-sight-integrated value of the local soft X-ray emis- sivity along multiple viewing directions within the instrument field-of-view, which cor- responds to many pixels in the images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We define this quantity Qint(ϕ, λ) as a function of the azimuth ϕ and elevation λ of a given line of sight.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Qint(ϕ, λ) = � Qloc(lϕ,λ) dlϕ,λ, (4) with lϕ,λ the distance from the spacecraft along the line of sight associated with the (ϕ, λ) azimuth–elevation pair.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' By convention, we will have the (0, 0) pair corresponding to the direction toward the Earth’s center.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The obtained instantaneous line-of-sight emissiv- ity Qint will be given in keV cm−2 s−1 sr−1 in this study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Soft X-ray images simulated using this Vlasiator run will have an angular resolu- tion of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='33◦ in both azimuth and elevation, and the line-of-sight integration will take place between the virtual satellite location and the outer boundary of the simulation do- main.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' If a given line of sight intersects the inner boundary, Qint will not be calculated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The virtual spacecraft will be placed either at (0, −30 RE, 0) in GSE coordinates (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', observing from the dawnside) or at (0, 0, 30 RE) (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', observing from above the north pole).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The two virtual satellites are selected to provide the side and polar views of the Earth’s magnetosphere similar to LEXI and SMILE, respectively, while keeping the ra- dial distances fixed at 30 RE between the apogees of LEXI and SMILE for easy compar- ison of soft X-ray signatures in polar and side views.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Both instantaneous values of Qint and time-integrated ones over 300 s, Qint 300s, will be shown, as the latter correspond to the integration time that SMILE and LEXI soft X-ray imagers require for good signal- to-noise ratios (Branduardi-Raymont et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Connor et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –7– manuscript submitted to Earth and Planetary Physics 3 Simulated Soft X-ray Images In this section, we present the simulated soft X-ray images obtained from the Vlasi- ator run when the virtual imaging spacecraft is viewing the dayside magnetosphere ei- ther from the dawnside or from above the north pole.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We first show instantaneous im- ages alongside relevant plasma parameters in the noon–midnight meridional plane or equa- torial plane, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Animations showing such instantanous images every second from t = 800 s to t = 1506 s are provided in supplementary material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We then time- integrate these instantaneous images during three 300 s time intervals and discuss the main features that can be observed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 Instantaneous Simulated Soft X-ray Images 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 View from Dawn Figure 2 shows plasma parameters in the noon–midnight meridional plane (Y = 0) at t = 1100 s in the simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 2a displays the proton number density, in which the magnetosheath and the polar cusps stand out.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' One can identify the bow shock, whose subsolar point lies approximately at 15 RE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Given the purely southward IMF orienta- tion, the bow shock is essentially quasi-perpendicular, which is why no ion foreshock is present.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Within the magnetosheath, density irregularities are particularly prominent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We will show in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 that these correspond to mirror-mode waves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The magnetopause subsolar point lies at about 10 RE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 2b gives the proton bulk velocity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' When crossing the bow shock, the solar wind slows down from its inflow speed of 750 km s−1 to velocities on the order of 200– 300 km s−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We can see that plasma convection at the dayside magnetopause gets faster at higher latitudes, before slowing down again when the plasma reaches the polar cusps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 2c presents the local soft X-ray emissivity in the plane, Qloc (calculated with Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 1), which depends not only on the previous two parameters (np and Vp), but also on the plasma temperature and the neutral density, not shown in the figure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The most prominent features in Qloc are the polar cusps and the magnetosheath, with values reach- ing 4 × 10−10 keV cm−3 s−1 sr−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the magnetosheath, the wave field is particularly visible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Brighter areas along the dayside magnetopause can also be identified;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' we will show in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 that these are flux transfer events (FTEs).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 2d shows the instantaneous image of line-of-sight-integrated soft X-ray emis- sions, Qint, viewed from a virtual imaging satellite placed at (0, −30 RE, 0), at this same time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The field of view associated with this image intersects the noon–midnight merid- ional plane within the red trapezoid in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Elevation is the angle along the Z di- rection, whereas azimuth is along X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this instantaneous image, the brightest areas correspond to the cusps and the magnetopause, with values of Qint on the order of 3 keV cm−2 s−1 sr−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It was shown in previous studies that lines-of-sight with maximum brightness correspond to directions tangential to the magnetopause (Collier & Connor, 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The very bright area close to the inner boundary is caused by boundary effects, as proton density is el- evated at low latitudes near the inner boundary (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2a), due to leakage of cold plasma resulting from the boundary condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' However, the bright SWCX soft X-ray emission near the inner boundary does not exist in reality and is an artificial effect by calculat- ing soft X-ray emissivity based on proton parameters (Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the instantaneous image of Qint, one can notice signatures likely associated with the wave field in the magnetosheath, as well as brighter spots along the dayside mag- netopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This means that, despite the line-of-sight integration and parallax effects, some of the structures identified in Qloc (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2c) also show in Qint.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The possible relation be- tween such structures and transient processes will be investigated in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –8– manuscript submitted to Earth and Planetary Physics Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a) Proton number density, (b) proton bulk velocity, and (c) local soft X-ray emis- sivity in the Y = 0 plane at t = 1100 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (d) Instantaneous soft X-ray image with 1 s integration time from a virtual spacecraft placed at (0, −30 RE, 0) at t = 1100 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The red trapezoid in panel (c) indicates the intersection of the instrument’s field of view shown in panel (d) with the Y = 0 plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –9– np [cm-3] Vp [kms-1] 10 900 (a) (b) 20 20 750 10 5 10 600 [Re] [Re] 0 D 0 450 Z Z 300 10 2 10 150 20 20 0 0 10 20 0 10 20 X[Re] X[Re] t = 1100 s 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 (C) (d) 30 20 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 20 10 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 10 Elevation [Re] 0 D 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0 D 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 Z 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 10 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 10 20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 20 30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0 10 20 0 10 20 30 X[RE] Azimuth [omanuscript submitted to Earth and Planetary Physics Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a) Proton number density, (b) proton bulk velocity, and (c) local soft X-ray emis- sivity in the Z = 0 plane at t = 1100 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (d) Instantaneous soft X-ray image with 1 s integration time from a virtual spacecraft placed at (0, 0, 30 RE) at t = 1100 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The red trapezoid in panel (c) indicates the intersection of the instrument’s field of view shown in panel (d) with the Z = 0 plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' An animated version of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2 is provided in the supplementary material as Movie S1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this animation, we can especially visualize how the magnetosheath wave signatures and dayside magnetopause bright spots move as a function of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The fact that the latter are colocated with confined regions of increased proton density forming near the subsolar point and transiting along the magnetopause toward the polar cusps strongly suggests that these are signatures of FTEs (see Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 and discussion in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 View from North Figure 3 is analogous to Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2, but this time the virtual imaging spacecraft has been placed above the north pole, at (0, 0, 30 RE) in GSE coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The chosen time step for this instantaneous snapshot of plasma parameters and soft X-ray emissions is –10– np [cm-3] Vp [kms-1] 10 900 (a) (b) 20 20 750 10 5 10 600 [Re] [Re] 0 450 Y Y 300 10 2 10 150 20 20 0 0 10 20 0 10 20 X[Re] X [Re] t = 1100 s 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 (C) (d) 30 20 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 20 10 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 10 Elevation [Re] D 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 Y 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 10 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 10 20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 20 30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0 10 20 0 10 20 30 X[RE] Azimuth [omanuscript submitted to Earth and Planetary Physics Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Soft X-ray image obtained from a virtual spacecraft located at (0, −30 RE, 0) with 300 s integration time starting at (a) t = 800 s, (b) t = 950 s, and (c) t = 1100 s in the simula- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' t = 1100 s, like previously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figures 3a–c show parameters in the equatorial plane (Z=0), and Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3d shows the line-of-sight integrated soft X-ray emissions as a function of az- imuth (along the X direction) and elevation (this time along Y ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The bow shock and dayside magnetopause are again prominent in the proton den- sity panel (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3a), as are the magnetosheath waves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The proton bulk velocity panel in- dicates the abrupt decrease in velocity across the bow shock, while the magnetosheath plasma accelerates once it reaches the flanks (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In local soft X-ray emissivity (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3c), it is again the magnetosheath that shows up the most (ignoring the artifacts near the inner boundary).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the instantaneous soft X-ray emission image (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3d), a bright arc is visible across elevations, corresponding to the dayside magnetopause observed tangentially.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Beyond the magnetopause signature, one can also see elongated structures originating from the magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Due to the viewing angle, polar cusps are not visible in this image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' An animated version of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3 is provided in the supplementary material as Movie S2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this animation, we can see how the elongated magnetosheath structures drift Earthward until they merge with the magnetopause signature, which occasionally brightens and ex- hibits undulations (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', around t = 1000 s and t = 1440 s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 Time-Integrated Soft X-ray Images 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 View from Dawn While the instantaneous soft X-ray images presented above show many interest- ing features, they do not correspond to images which could be obtained by SMILE and LEXI, as the instruments will need an integration time on the order of 300 s to obtain a sufficient number of counts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We therefore produce time-integrated images over three 300 s time intervals during the studied part of the simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We first consider the view from a virtual imaging spacecraft in the dawn sector;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 4 shows time-integrated soft X-ray images obtained during t = 800–1100 s (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4a), t = 950–1250 s (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4b), and –11– t = 800-1100 s t = 950-1250 s t = 1100-1400 s 820 (a) (b) (c) 30 30 30 700 20 20 20 sr-11 580 10 10 10 N (along 460 0 0 0 10 10 10 340 20 20 20 220 30 30 30 100 0 20 0 20 0 20 Azimuth[°] Azimuth[°] Azimuth[°] (along X) (along X) (along X)manuscript submitted to Earth and Planetary Physics Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a) Time-integrated soft X-ray image from t = 950 to t = 1250 s obtained from a virtual spacecraft located at (0, −30 RE, 0) (same as Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4b), with several cuts along and across boundary region signatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (b) Soft X-ray emission along the black cut through the northern- hemisphere magnetopause and magnetosheath signatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (c) Same along the grey cut through the southern-hemisphere magnetopause and magnetosheath signatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (d) Same along the red cut across the northern cusp signature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (e) Same along the orange cut across the southern cusp signature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (f) Same along the magenta cut following the dayside magnetopause signature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In pan- els b–f, the three lines correspond to the three studied 300 s time intervals for image acquisition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' t = 1100–1400 s (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In these images, the brightest structures correspond to the cusps as well as the dayside magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is clear from the figures that the time integration blurs out most small-scale struc- tures that are associated with transient phenomena at the magnetopause and in the mag- netosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' However, one can see some irregularity in the shape and brightness of the magnetopause signature, such as a dimmer and thinner region at elevations near −3◦ in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4a, and also at elevations near −7◦ in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4c, for instance (indicated with black ar- rows).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Conversely, parts of the dayside magnetopause signature appear brighter, like re- gions near +3◦ and −9◦ elevations in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4a (indicated with white arrows).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We will in- vestigate the possible connection between these brighter Qint 300s values at the dayside magnetopause and FTEs in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Besides, Figs 4b–c exhibit brighter stripes in their magnetosheath signatures (in- dicated with magenta arrows).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Such stripes have oblique orientations with respect to the Earth–Sun line and can be seen in both the northern and the southern part of the do- main.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2, we will show that these signatures are likely the result of mirror-mode waves in the magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In order to better visualize differences between the three images, we will look at Qint 300s values along a few selected cuts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a reproduces the time-integrated image within t = 950–1250 s (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4b) and indicates where five cuts have been performed, along which the Qint 300s values are extracted and displayed in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b–f for the three time intervals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b and 5c show time-integrated soft X-ray emission values along the black and grey straight lines, respectively, which both cut across the magnetopause and through the magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The horizontal axis is the angular distance, noted d, of a point on –12– t = 950-1250 s 820 800 (a) 800 800-1100 s C 30 600 950-1250 s 600 [keV cm-} 1100-1400 400 400 700 200 200 20 18 21 24 27 30 33 36 18 21 24 27 30 33 36 d[°] [。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content="]p b 580 10 levation[ N Qint_300s (d) (e) 700 700 (along ) 2 460 [keV cm' 0 600 600 500 500 E 10 50 55 60 65 65 60 55 50 340 ESL[°] ESL[°] 20 Qint_300s 800 220 (f) 700 30 600 100 500 0 10 20 30 70 50 30 10 10 30 50 70 EsL[°] Azimuth[°] (along X)manuscript submitted to Earth and Planetary Physics Figure 6." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Same as Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4 but with the virtual spacecraft placed at (0, 0, 30 RE).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' the line from the (0, 0) viewing direction (calculated as � ϕ2 + λ2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In both panels, Qint 300s peaks at d ≈ 21◦, which corresponds to the tangent direction to the magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While the curves corresponding to the three integrating time intervals are almost perfectly su- perimposed on top of each other in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b, one can see that the peak was slightly re- duced and drifted to larger d values with time in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Around d ≈ 24◦, there are small-amplitude fluctuations visible in the violet and orange curves, in both panels, cor- responding to the stripes identified in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4b–c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A last observation that can be made from these two panels is a change in the slope of Qint 300s occurring at d ≈ 36◦, indi- cated with black arrows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This change of slope is likely associated with the line of sight tangent to the bow shock, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4 of Connor et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The correspond- ing locations in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a are indicated with white arrows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Values of Qint 300s along the red and orange curved lines, cutting through the north- ern and southern cusp, respectively, are shown in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5d and 5e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The horizontal axis is the angle between the Earth–point direction and the Earth–Sun line, noted θESL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In both hemispheres, the cusp signature drifted slightly poleward with time, although the differences are relatively small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 5f give the Qint 300s values along the magenta curve, which essentially fol- lows the magnetopause signature in the images, as a function of θESL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It shows quite clearly the variations in Qint 300s previously identified in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4 and enables estimating the am- plitude of those variations, which can reach up to 90 keV cm−2 sr−1 (orange curve peak- to-trough amplitude at θESL within [−20◦, −10◦]), i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' representing about 10% of the average value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Other peaks can be identified at θESL ≈ ±55◦, corresponding to loca- tions in the high-altitude polar cusps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 View from North By applying the same methodology as above, we can also produce time-integrated soft X-ray images as would be obtained from the virtual imaging spacecraft placed above the north pole.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 6 presents the three images corresponding to the same integra- tion intervals (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e, t = 800–1100 s, t = 950–1250 s, and t = 1100–1400 s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the three panels,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' the brightest signature comes from the dayside magnetopause,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' with brightness maximizing at locations corresponding to low nose angles and signatures –13– t = 800-1100 s t = 950-1250 s t = 1100-1400 s 820 (a) (b) (c) 30 30 30 700 20 20 20 sr-11 580 10 10 10 (along ) 0 0 0 460 10 10 10 340 20 20 20 220 30 30 30 100 0 20 0 20 0 20 Azimuth[°] Azimuth[°] Azimuth[°] (along X) (along X) (along X)manuscript submitted to Earth and Planetary Physics Figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a) Time-integrated soft X-ray image from t = 950 to t = 1250 s obtained from a virtual spacecraft located at (0, 0, 30 RE) (same as Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 6b), with several cuts along and across boundary region signatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (b) Soft X-ray emission along the black cut through the post-noon magnetopause and magnetosheath signatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (c) Same along the grey cut through the pre-noon magnetopause and magnetosheath signatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (d) Same along the magenta cut following the day- side magnetopause signature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In panels b–d, the three lines correspond to the three studied 300 s time intervals for image acquisition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' becoming fainter toward the flanks of the magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Magnetosheath stripes also show from this perspective, and are visible in all three panels (indicated with magenta arrows), with mostly oblique orientations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Beyond the magnetopause signature near the flanks, a second, fainter line parallel to it can be seen, especially in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 6b–c (indicated with black arrows).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This parallel line might be associated with the same processes as the stripes closer to the subsolar point, as they appear similar in terms of brightness and orienta- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 7 shows the values of Qint 300s along a few selected cuts, in a same way as in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 7a reproduces Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b and indicates where three cuts are considered to study the variations of Qint 300s along the magnetopause X-ray signature as well as across the magnetosheath in the pre-noon and post-noon sectors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We can see from Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 7b–c that, along the cuts crossing the magnetosheath, the integrated soft X-ray emissivity peaks at d ≈ 21◦ in both the post-noon and pre-noon sectors, respectively, with a slight reduction in the peak value and very slight outward shift with time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This is consistent with the trend observed in the virtual images obtained with the view from dawn, and can be interpreted as a slight sunward motion of the day- side magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Sunward from the peak (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' at larger d values), fluctuations in Qint 300s can be seen, almost reaching a secondary peak at d ≈ 23◦ (more visible in the dawnside cut, Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 7c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This corresponds to the oblique stripes indicated with magenta arrows in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Like in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b–c, we can notice a change in the slope of the curves in Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 7b– c, around d ≈ 33◦ (indicated with black arrows).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The corresponding locations in the soft X-ray image (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 7a) are indicated with white arrows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' As mentioned previously, this slope change is likely associated with lines of sight tangent to the bow shock (Connor et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –14– t = 950-1250 s 820 (a) 800 (b) 800-1100 s 800 (c) 30 sr-1] 950-1250 s 1100-1400 s Qint_300s 600 600 [keV cm-2 s 700 20 400 400 580 200 10 200 。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='. levation[ 18 21 24 27 30 33 36 18 21 24 27 30 33 36 460 [。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=']p [。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=']p 0 E (d) 10 340 800 Qint_300s 700 20 220 600 30 100 500 50 0 10 20 30 70 30 10 10 30 50 70 EsL[°] Azimuth[°] (along X)manuscript submitted to Earth and Planetary Physics Finally, looking at Qint 300s values along the magenta cut following the magnetopause signature (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 7d), we note that the signal roughly follows linear slopes in the flanks (|θESL| > 30◦), with little to no evolution with time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' On the other hand, the section corresponding to the nose of the magnetopause exhibits fluctuations and variability amounting to about 5% of the Qint 300s values (up to 40 keV cm−2 sr−1 between the violet and the orange lines at θESL ≈ −20◦).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In a given image (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', focusing on a single line in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 7d), fluctua- tions can lead to the appearance of several local maxima in Qint 300s, denoting the ex- istence of structures near the magnetopause nose.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These structures might be associated with the nature of the reconnection taking place at the dayside magnetopause, which can either be over an extended X-line or on the contrary more patchy (Atz et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Walsh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4 Signatures of Transient Processes in Soft X-ray Images In this section, we will investigate to what extent transient processes taking place in the magnetosheath and at the dayside magnetopause can induce signatures in soft X- ray images as will be measured by SMILE and LEXI SXI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We will focus on two processes: flux transfer events forming by magnetic reconnection at the dayside magnetopause and mirror-mode waves developing in the magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 Flux Transfer Events at the Magnetopause In Vlasiator 2D–3V runs driven by IMF with a southward Bz component, flux trans- fer events regularly form at the magnetopause near the subsolar point (Hoilijoki et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2017, 2019), and they propagate poleward toward the cusps, with consequences in terms of energy transfer into the magnetosphere (Ala-Lahti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2022) and proton precipi- tation into the cusps (Grandin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this 3D–3V run, FTEs also form near the subsolar point and gradually follow the magnetopause toward the cusps, in both hemispheres.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This can be seen in Supple- mentary Movie S3, where FTEs appear as regions of increased proton density enclosed within a magnetic island.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To investigate whether these FTEs can have a signature on soft X-ray images, we first find a simple way to detect and track them along the day- side magnetopause, and we then look at their time-integrated parameters over 300 s win- dows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 8a shows (background color), as a function of time in the simulation, the magnetic field component along the radial direction (away from the Earth), Br, along the magenta cut following the magnetopause signature shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' More precisely, the values of Br are taken along the intersection of the lines of sight forming this cut with the Y = 0 plane (noon–midnight meridional plane), and they approximately correspond to the magnetic field component normal to the magnetopause surface.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' One can see that, at a given value of θESL along the intersected cut, Br exhibits pseudo-oscillations with amplitudes up to ±10 nT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These correspond to bipolar signatures as FTEs transit across the location.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In the northern hemisphere, when the leading edge of a FTE approaches a given location at the dayside magnetopause, Br is increased (red values in the plot), and as the trailing edge of the FTE passes the location, Br shows a negative deflection (blue values).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Due to the geometry, the situation is opposite in the southern hemisphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hence, we can see how FTEs tend to form within a few degrees from the subsolar point at the dayside magnetopause, and how they propagate toward one of the polar cusps in ∼150 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is noteworthy that the propagation time of the identified FTEs is shorter than the image acquisition time which is retained in this study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' However, one may want to determine whether FTEs can still create a signature in soft X-ray images integrated over 300 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –15– manuscript submitted to Earth and Planetary Physics Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a) Background color: Time evolution of the radial (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', nearly normal to the mag- netopause) component of the magnetic field along the intersection of the magenta cut following the magnetopause signature in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a with the Y = 0 plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' FTEs generated near the subsolar point create a bipolar signature moving toward higher magnetopause angles as a function of time, with opposite polarity in opposite hemispheres.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Isocontours: Time evolution of the line-of-sight- integrated proton number density along the magenta cut in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (b) Proton number density along the intersection of the cut with the Y = 0 plane, averaged over 300 s for the three studied time intervals (starting at t = 800 s, t = 950 s and t = 1100 s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (c) Value of the 300 s-integrated soft X-ray emission along the cut in the corresponding three images shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4 (same data as Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5f).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –16– 800-1100 s 950-1250 s Br [nT] (a) 1100-1400 s 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3 (b) (c) 40 40 40 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 20 20 20 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='7 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0 0 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='4 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 20 20 20 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 0 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 40 : 40 40 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='8 800 900 1000 1100 1200 1300 1400 1500 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0 700 750 800 [np[1010cm-2] Time [s] (np)300s [cm-3] Qint_300s [keV cm-2 sr-1]manuscript submitted to Earth and Planetary Physics In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 8b, we show the time-averaged value of the proton number density, ⟨np⟩300s, along the intersection of the cut with the Y = 0 plane (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', at the locations where Br is monitored).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The time averaging is done over 300 s intervals corresponding to the in- tegration time of the three images shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It can be seen that (i) the time-averaged density along the magnetopause exhibits some variations as a function of θESL, and (ii) these variations change from one 300 s interval to another.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The largest differences occur at θESL values comprised between 0◦ and −20◦ during the third time interval (1100–1400 s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Dur- ing this time interval, three bipolar signatures can be identified in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 8a at those θESL values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We can see in Movie S3 that these FTEs are indeed associated with higher-than- background proton number densities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 8c reproduces the corresponding values of Qint 300s during the three time intervals along the cut in the soft X-ray images (same data as in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5f).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While the match between the spatial and temporal variations in ⟨np⟩300s and Qint 300s is not perfect, the corresponding curves do nonetheless exhibit striking resemblance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This suggests that the presence of FTEs at the dayside magnetopause during the image acquisition time can affect the average proton density along corresponding lines of sight and induce a signa- ture in soft X-ray emission values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The imperfect match can be expected given that Qint 300s values result from a line-of-sight integration, whereas the ⟨np⟩300s values are taken lo- cally in the Y = 0 plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To investigate this further, overlaid with the Br data in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 8a are isocontours of the proton number density integrated along the lines of sight forming the magenta cut in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' As for the Qint 300s shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 8c, this line-of-sight-integrated density likely contains effects of variations occurring outside of the Y = 0 plane;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' however, one can see a tendency for density enhancements to follow the pattern formed by the bipolar sig- natures associated with FTEs in the Y = 0 plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This suggests that the proton den- sity enhancements occurring in the core of the simulated FTEs can produce soft X-ray signatures even in line-of-sight and time-integrated images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is worth mentioning that, in observations, FTEs often exhibit a density decrease in their core (Akhavan-Tafti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Yet, some observed FTEs do show a proton density increase and might there- fore produce signatures in soft X-ray images similar to those described here (see Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2 Mirror-Mode Waves in the Magnetosheath In Movie S3, one can identify wave-like structures in the magnetosheath consist- ing of patches of increased proton number density appearing in the central part of the magnetosheath (Earthward from the bow shock) and slowly drifting toward the mag- netopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In 2D–3V Vlasiator simulations, mirror-mode waves have been identified in the magnetosheath (Hoilijoki et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Dubart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In a nearly 3D–3V setup wherein a noon–midnight meridional-plane slice with a thickness of 7RE along the dawn– dusk dimension was simulated with Vlasiator, Pfau-Kempf et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020) identified an- ticorrelated magnetic field and proton density fluctuations in the magnetosheath, attributed to mirror-mode waves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We will first check whether in this 3D–3V run those waves ex- hibit similar properties as mirror-mode waves, after which we will investigate to what extent they can create signatures in soft X-ray images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 9 presents the magnetic field magnitude (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 9a–b), proton number den- sity (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 9c–d) and local soft X-ray emissivity (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 9e–f) along the intersections of the black and grey cuts in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a with the Y = 0 plane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These data are shown at four time steps in the simulation: t = 1000, 1010, 1020, and 1030 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Here, the horizontal axis cor- responds to the X coordinate of the cut intersection points in the Y = 0 plane to give a better intuition of spatial scales.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' At X < 9 RE, the magnetic field magnitude decreases steadily and the proton den- sity is on the order of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 cm−3, which corresponds to magnetosphere field and plasma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' As X increases, the magnetic field magnitude drops more drastically, whereas the pro- –17– manuscript submitted to Earth and Planetary Physics Figure 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Investigation of the magnetosheath waves signatures along the black and grey cuts in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The shown variables are taken along the intersections of the cuts with the Y = 0 plane at four time steps in the simulation (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', these are instantaneous values).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a–b) Magnetic field magnitude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (c–d) Proton number density.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (e–f) Local soft X-ray emissivity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –18– Northern hemisphere Southern hemisphere 60 t = 1000 s 60 [BI [nT] t = 1010 s t = 1020 s 40 40 t = 1030 s 20 20 (a) (b) 4 4 (d) C 3 3 2 2 du 1 1 0 0 1e-10 1e-10 4 4 (e) (f) 3 2 2 [keV cm- 1 1 0 0 8 9 10 11 12 8 9 10 11 12 X[Re] X[Re]manuscript submitted to Earth and Planetary Physics Figure 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Time-averaged values of the magnetosheath parameters shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 9 over 300 s intervals starting at t = 800 s, t = 950 s, and t = 1100 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (a–b) Time-averaged magnetic field magnitude along the intersections of the black and grey cuts with the Y = 0 plane, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (c–d) Same for the time-averaged proton number density.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (e–f) Same for the time-averaged lo- cal soft X-ray emissivity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (g–h) Soft X-ray emission values along the black and grey cuts in the 300 s-integrated images shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The last two panels show the same data as Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b–c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' ton density increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This corresponds to the crossing of the magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Beyond the magnetopause, one can notice fluctuations in both |B| and np, with sizes on the order of 1–2 RE along the X coordinate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These fluctuations are more prominent in the north- ern (black) cut than in the southern (grey) cut, and there is a fairly clear anticorrela- tion in the variations of both parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Besides, we can see that the structures asso- ciated with these fluctuations appear to drift Earthward by about 1 RE in 40 s, giving an estimated speed along the X direction of ∼150 km s−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Comparing with Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2b, we can infer that the structures therefore approximately drift alongside the plasma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The an- ticorrelation between magnetic field and density variations and the fact that the waves are roughly static in the plasma frame strongly suggest that these waves are indeed mir- ror modes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –19– Northern hemisphere Southern hemisphere 60 60 800-1100 s 950-1250 s 1100-1400 s [nT] 40 40 (a) 20 20 [(b) 3 (c) (d) 3 (np/300s 2 2 1 1 0 0 1e-10 1e-10 3 3 (e) (f) (Qioc/300 s 2 2 [keV cm- 1 1 0 0 800 800 (g) (h) 600 600 400 400 18 20 22 24 26 28 18 20 22 24 26 28 [.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='p [。' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=']pmanuscript submitted to Earth and Planetary Physics Given the significant propagation of the mirror-mode waves over the span of 40 s seen in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 9, one cannot expect that individual wave fronts will be visible in an image integrated over 300 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' However, like for FTEs, one can investigate whether mirror-mode wave activity in the magnetosheath can nevertheless produce a signature in soft X-ray images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Figure 10 presents this analysis along the same cuts as Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 9, but this time for time-averaged parameters along the cut, again using the same three 300 s time intervals as previously.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In all panels of this figure, the horizontal axis corresponds to the angu- lar distance d to the origin of the azimuth–elevation frame used in soft X-ray images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 10a–b, we show the time-averaged magnetic field magnitude along the cuts during the three intervals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While the magnetosphere and magnetopause values do not differ significantly during the three intervals, one can see that the magnetosheath fluc- tuations, associated with the mirror-mode waves identified previously, differ slightly from each other.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The differences appear more clearly in the time-averaged proton density pan- els (Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 10c–d), especially at d ≈ 21–26◦, and the resulting time-averaged local soft X-ray emission values (⟨Qloc⟩300s, Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 10e–f) in the Y = 0 plane are consistent with average density fluctuations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To assess the effect of these local time-averaged fluctua- tions in actual soft X-ray images, Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 10g–h show the soft X-ray emissions integrated along the line of sight corresponding to the two cuts (same data as in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b–c, respec- tively).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In these panels, we can see that the three lines corresponding to the three time intervals do not fully overlap within d ≈ 22–26◦.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5 Discussion First of all, it is worth comparing the soft X-ray emissivity values obtained with the Vlasiator simulation to those reported in modeling studies using different models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019) carried out a series of soft X-ray emission simulations using a MHD model, driven by various types of solar wind conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Our driving conditions correspond clos- est to their first case, with a solar wind density of 5 cm−3, a solar wind speed of 400 km s−1, and a purely southward IMF of 5 nT magnitude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' They obtained a soft X-ray image from a virtual imaging satellite approximately positioned above the north pole (analogous to our Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3d), in which the magnetopause signature reached a value of 13 keV cm−2 s−1 sr−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In contrast, with our Vlasiator run, we get a value close to 3 keV cm−2 s−1 sr−1, which is almost a factor of 5 lower.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This is consistent with our solar wind density being 5 times lower than theirs, leading to a magnetosheath density accordingly lower.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Our values are also in fairly good agreement with those obtained by Connor et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021) using a sim- ulation from the OpenGGCM global MHD model and providing soft X-ray images seen from the dawnside (analogous to our Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While Vlasiator is too computationally expensive to be run over a long time inter- val, its unique advantage is that the hybrid-kinetic description of the space plasma en- ables the formation of transient features which cannot emerge in MHD simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We can therefore simulate the possible effect of such phenomena in soft X-ray images of the near-Earth space as will be obtained by SMILE and LEXI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Considering first the case of FTEs, if we focus on the largest peak-to-trough variation of Qint 300s in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 8c, corre- sponding to the 1100–1400 s time interval and θESL angles between −20◦ and 0◦, we ob- tain ∆Qint 300s ≈ 90 keV cm−2 sr−1, amounting to about 12% of the background value along the dayside magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The corresponding spatial variations in proton den- sity in the Y = 0 plane amount to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='81 cm−3, on the order of 30% of the background value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This can provide quantitative elements to assess whether a soft X-ray imager might be able to detect signatures associated with FTEs, depending on the specifications of the soft X-ray sensor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' FTEs have been found to form at the dayside magnetopause quasi-periodically ap- pearing approximately once every 8 minutes (Rijnbeek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 1984) preferably forming during southward IMF conditions (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Berchem & Russell, 1984).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A statistical study –20– manuscript submitted to Earth and Planetary Physics of 55 FTEs using MMS observations found an average size of a subsolar FTE to be 1700 ± 400 km (Akhavan-Tafti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018), which is 3 to 7 times smaller than the FTEs observed at the higher latitudes (Akhavan-Tafti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Fermo et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2011;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2005).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' However, the scale sizes of the observed FTEs vary from ion-scale flux ropes (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Eastwood et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Dong et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2017) to FTEs with a diameter up to 1–2 RE (Rijnbeek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 1984;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Fear et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2007;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hasegawa et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2006), while their axial length can be much longer (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Fear et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A typical FTE has an internal structure with twisted field lines and axial magnetic field leading to an increase in total magnetic field magnitude and a decrease in the den- sity in the core of the FTE (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Akhavan-Tafti et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' How- ever, “crater” FTEs with a decrease in total magnetic field with a higher density in the middle have been observed (Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A statistical study of 18 typical and 14 crater FTEs comparing FTE densities (nFTE) and magnetic field strength (BFTE) to the mag- netosheath density (nMS) and magnetic field (BMS) showed that 75% of typical FTEs have the density ratio nFTE/nMS < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 and 80% of the FTEs have BFTE > BMS, while 90% of the crater FTEs had nFTE/nMS > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5 (and approximately 50% nFTE > nMS) and approximately 95% had smaller magnetic field magnitude than the surrounding magne- tosheath (Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In this simulation, the identified FTEs have a higher-than- background proton density in their core, which makes them analogous to crater FTEs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We can therefore speculate that soft X-ray images might be able to capture signatures of crater FTEs, even more so if their peak-to-trough density variation is similar to or greater than that of FTEs identified in this simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is however clear from the presented results that FTEs forming in this Vlasiator run move across a 300 s integrated soft X-ray image too fast for them to be individu- ally visible in the image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Rather, the local soft X-ray emission enhancements in some parts of the magnetopause signature are likely due to the cumulative effect of several FTEs, especially if these tend to remain longer at a given latitude.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hoilijoki et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019) showed that the speed at which FTEs propagate toward higher latitudes along the magnetopause depends on the driving conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We can therefore speculate that, in some cases where the FTE motion is slower than in this simulation, their signatures in soft X-ray images might be more prominent than in the results obtained here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Regarding mirror-mode waves in the magnetosheath, we can estimate the effect of the largest density enhancement visible in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 10c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Considering the difference between the ⟨np⟩300s values during the first and third time intervals at d = 24◦, we find that it consists of a local density enhancement by 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='4 cm−3 (∼14%) in the Y = 0 plane as- sociated with mirror-mode waves structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The corresponding difference in time-integrated soft X-ray emissions amounts to about 20 keV cm−2 sr−1 (∼4%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Mirror-mode structures are large-scale compressional waves that are non-propagating in the plasma frame.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' They are often observed in the Earth’s magnetosheath especially behind the quasi-perpendicular bow shock.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' They may appear as quasi-sinusoidal oscil- lations in the magnetic field but mirror-mode structures are often observed as peaks in the mirror-unstable plasma and dips in mirror-stable conditions near the magnetopause (Soucek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The depth of the mirror structures has been found to increase with the decreasing distance to the magnetopause (Soucek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The analysis of 2 months of Cluster data by Soucek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2008) showed that the average period of the mirror- mode waves was approximately 12 s (with 98% of the observed structures being between 4 to 24 s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A statistical study by Lucek et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2001) showed that the scale sizes of the mirror structures can vary from less than 600 km along the local magnetopause normal, 750–1000 km along the maximum variance direction to 1500–3000 km along the flow di- rection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The amplitude of the magnetic field variations associated with mirror-mode waves is typically on the order of several 10% of the background field, and the associated pro- ton density variations can lead to up to multi-fold local enhancements (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Chandler et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' –21– manuscript submitted to Earth and Planetary Physics The differences in Qint 300s values obtained above are estimates under the driving conditions and the setup used in the Vlasiator run on which this study is based.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These choices and constraints have several consequences on the properties of the mirror-mode waves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' First of all, in our simulation, the spatial resolution in the magnetosheath (2000 km) likely leads to mirror-mode structures being larger than in reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Therefore, one may speculate that SMILE and LEXI observations of mirror-mode structures would likely be smaller and might hence be blurred out more easily than in this simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Besides, Dubart et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020) showed that, when the spatial resolution of a Vlasi- ator simulation is too coarse, the unstable proton distributions in the magnetosheath en- tirely pump their free energy into the growth of mirror-mode structures, instead of trans- ferring part of it to electromagnetic ion cyclotron waves, which require a finer spatial res- olution to emerge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is worth noting, however, that the solar wind driving conditions used in the run are not the most typical ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Indeed, the solar wind density is notably lower than average, whereas the solar wind velocity and temperatures are higher than average.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This suggests that, under solar wind conditions closer to average values than in this Vlasiator run, such as those from Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019) discussed above, larger sig- nal than what we obtained can be expected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This, combined with the fact that density fluctuations associated with mirror-mode waves can be greater than in our simulation (factor of 3 in the event analyzed by Chandler et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021)), suggests that the constraints on FTEs and mirror-mode waves for them to produce signatures in the soft X-ray ob- servations of LEXI and SMILE might actually be less strict than suggested by our re- sults.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Additional simulations with a finer spatial resolution and different solar wind driv- ing are needed to test this hypothesis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Another point worth discussing is that the Vlasiator run duration analyzed for this study consists of a 700 s time interval only.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In reality, both LEXI and SMILE SXI in- struments will observe processes over significantly longer time scales, which could open more possibilities for the data analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' For instance, calculating the discrepancy between two time-integrated images separated by a few minutes could enhance contrast in regions where the plasma conditions have changed during that time frame.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' While this would likely still not enable the detection and monitoring of individual structures given the required image acquisition time, this could provide additional approaches to indirectly identify such processes in near-Earth space, as well as context for in-situ observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Finally, one should point out the fact that, in this Vlasiator run, the driving con- ditions are steady.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' As a consequence, one may expect that the large-scale features such as the boundary locations (bow shock, magnetopause, polar cusps) do not change much during the studied time interval.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Nevertheless, Figs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5b–e indicate that the magnetopause X-ray signature exhibits a slight outward motion with time while the peaks in the cusp signatures move toward higher latitudes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This suggests that the analysis of successive soft X-ray images might enable the tracking of boundary motions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' To confirm this, it would therefore prove interesting to monitor the motion of these boundaries in a Vlasi- ator run in which solar wind conditions are varying.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This is a future avenue for an up- coming study which could be made possible once a Vlasiator run with time-varying driv- ing conditions is available.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 6 Conclusions In this paper, we presented the first estimates of SWCX soft X-ray emission imag- ing of the dayside near-Earth space based on a hybrid-Vlasov simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We used ∼700 s of a Vlasiator run driven by constant solar wind of low density and high velocity and with purely southward IMF with Bz = −5 nT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' From the Vlasiator simulation outputs, we calculated at every second the local soft X-ray emissivity based on the plasma param- eters and produced line-of-sight-integrated instantaneous soft X-ray images as would be obtained from a virtual imaging spacecraft placed at a distance of 30 RE from the Earth’s –22– manuscript submitted to Earth and Planetary Physics center, providing the polar and side views of the Earth’s magnetosphere similar to the SMILE and LEXI views, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We then integrated those instantaneous images over 300 s intervals to reproduce anticipated observational requirements for the SMILE and LEXI SXI instruments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The main results of this study are as follows: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The most prominent features in soft X-ray images obtained from above the pole or on the dawnside consist of the magnetopause and polar cusps (the latter only when observing from the dawnside).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The soft X-ray emissivity values obtained with Vlasiator are consistent with earlier MHD results, taking into account differences in the solar wind driving conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Despite the 300 s integration time, the obtained soft X-ray images exhibit smaller- scale features such as the brightening of localized areas at the dayside magnetopause and wave-like patterns in the magnetosheath signature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' These features are cre- ated by transient phenomena, namely FTEs and mirror-mode waves, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Based on the simulated soft X-ray images, one can anticipate that FTE signatures could amount to 12% of the background signal if FTEs occurring during the 300 s interval cumulatively lead to local proton density enhancements by about 30%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' This result holds for “crater” FTEs with enhanced proton density in their core (Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2010);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' possible soft X-ray signatures associated with typical FTEs (which generally exhibit a decrease in proton density in their core) need further investigation to be characterized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Correspondingly, soft X-ray images of SMILE and LEXI might reveal the pres- ence of mirror-mode waves, as in the simulation the associated density structures in the magnetosheath cumulatively led to an enhancement by 14% in the proton density, locally, and produced an increase in soft X-ray emission signal by 4%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' De- pending on the scale size, propagation speed and amplitude of the variations in the proton density, mirror-mode structures may lead to more pronounced or on the contrary more blurred out signatures in soft X-ray images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The above values are likely conservative estimates, given that the solar wind con- ditions in this Vlasiator run are representative of solar wind high-speed streams rather than the more common average solar wind driving which typically has larger proton number density.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The results of this study contribute to the SMILE Modeling Working Group ob- jectives by comparing simulation results using a hybrid-Vlasov code with other works based on MHD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Moreover, the hybrid-kinetic description of the space plasma provides a unique opportunity to anticipate under what conditions transient processes in the day- side near-Earth space might be observed by soft X-ray imagers on SMILE and LEXI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Future plans could include the monitoring of the motion of the large-scale space plasma boundaries (magnetopause, polar cusps, bow shock) in a different run with time-varying upstream conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Acknowledgments We acknowledge the European Research Council for starting grant 200141-QuESpace, with which the Vlasiator model was developed, and consolidator grant 682068-PRESTISSIMO awarded for further development of Vlasiator and its use in scientific investigations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' We gratefully acknowledge Academy of Finland grant numbers 338629-AERGELC’H, 339756- KIMCHI, 336805-FORESAIL, and 335554-ICT-SUNVAC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Academy of Finland also supported this work through the PROFI4 grant (grant number 3189131).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hyunju K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Con- nor gratefully acknowledges support from the NASA grants, 80NSSC20K1670 and 80MSFC20C0019, and the NASA GSFC FY23 IRAD and HIF funds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The CSC – IT Center for Science and the PRACE Tier-0 supercomputer infrastructure in HLRS Stuttgart (grant number 2019204998) are acknowledged as they made these results possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The authors wish to thank the –23– manuscript submitted to Earth and Planetary Physics Finnish Grid and Cloud Infrastructure (FGCI) for supporting this project with compu- tational and data storage resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The authors declare that they have no conflict of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Vlasiator simulation data used in the study amount to 18 TB of disk space;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' access to the raw simulation data can be granted by following the Vlasiator data access policy (see https://www2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='helsinki.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='fi/en/researchgroups/vlasiator/rules-of-the road).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Vlasiator code is preserved at https://zenodo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='org/record/3640593 (Pfau- Kempf et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 2022), available in open access.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' It is developed openly at https://github .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='com/fmihpc/vlasiator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' References Akhavan-Tafti, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Slavin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Le, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Eastwood, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Strangeway, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Russell, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Burch, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Mms examination of ftes at the earth’s sub- solar magnetopause.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research: Space Physics, 123(2), 1224-1241.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1002/2017JA024681 Ala-Lahti, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pulkkinen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Energy Flux Through the Magnetopause During Flux Transfer Events in Hybrid-Vlasov 2D Simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geophysical Research Letters, 49(19), e2022GL100079.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2022GL100079 Atz, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Walsh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Broll, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Zou, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The spatial extent of mag- netopause magnetic reconnection from in situ themis measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research: Space Physics, 127(12), e2022JA030894.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/ 2022JA030894 Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Blanco-Cano, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Kajdiˇc, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Johlander, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Tarvus, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Helium in the Earth’s foreshock: a global Vlasiator survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 38(5), 1081-1099.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/ angeo-38-1081-2020 Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Brito, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Non-locality of Earth’s quasi-parallel bow shock: injec- tion of thermal protons in a hybrid-Vlasov simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 38(3), 625-643.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-38-625-2020 Berchem, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Russell, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1984).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Flux transfer events on the magnetopause - Spatial distribution and controlling factors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research, 89, 6689-6703.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/JA089iA08p06689 Branduardi-Raymont, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Escoubet, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Adamovic, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Agnolon, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Berthomier, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Zhu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' SMILE definition study re- port (Red Book).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' European Space Agency, ESA/SCI, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5270/ esa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='smile.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='definition study report-2018-12 Carter, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sembay, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Read, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A high charge state coronal mass ejection seen through solar wind charge exchange emission as detected by XMM-Newton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Monthly Notices of the Royal Astronomical Society, 402(2), 867-878.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1111/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1365-2966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='15985.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='x10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='48550/arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0911.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0897 Carter, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sembay, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Read, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Identifying XMM-Newton ob- servations affected by solar wind charge exchange - Part II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Astronomy & As- trophysics, 527, A115.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1051/0004-6361/20101581710.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='48550/arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1101 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1848 Chandler, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Schwartz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Avanov, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Coffey, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Giles, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Moore, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Torbert, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Observations of Mirror Mode Structures in the Dawn Side Magnetosphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 126(2), e28649.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2020JA028649 Collier, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Connor, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Magnetopause Surface Reconstruction From Tangent Vector Observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 123(12), 10,189-10,199.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2018JA025763 –24– manuscript submitted to Earth and Planetary Physics Connor, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Carter, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Exospheric Neutral Hydrogen Density at the Nominal 10 RE Subsolar Point Deduced From XMM-Newton X-Ray Obser- vations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 124(3), 1612-1624.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2018JA026187 Connor, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sibeck, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Collier, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Baliukin, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Branduardi-Raymont, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Brandt, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Zoennchen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Soft X ray and ENA Imaging of the Earth’s Dayside Magnetosphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 126(3), e28816.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2020JA028816 Cox, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1998).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Modeling the local bubble.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' In D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Breitschweidt, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Freyberg, & J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Tr¨umper (Eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' ), The local bubble and beyond (p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' 121).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Springer-Verlag, New York.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Cravens, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Robertson, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Snowden, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2001).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Temporal variations of geocoronal and heliospheric X-ray emission associated with the solar wind interaction with neutrals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research, 106(A11), 24883- 24892.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2000JA000461 Dong, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Dunlop, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Trattner, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Phan, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Fu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='-B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Burch, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Structure and evolution of flux transfer events near day- side magnetic reconnection dissipation region: MMS observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geophysical Research Letters, 44(12), 5951-5959.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1002/2017GL073411 Dubart, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Osmane, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Johlander, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Resolution dependence of magnetosheath waves in global hybrid-Vlasov simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 38(6), 1283-1298.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-38-1283-2020 Eastwood, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Phan, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cassak, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Gershman, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Haggerty, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Malakit, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Wang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Ion-scale secondary flux ropes generated by magnetopause reconnection as resolved by MMS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geophysical Research Letters, 43(10), 4716–4724.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1002/2016GL068747 Fear, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Milan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Fazakerley, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Lucek, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cowley, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Dan- douras, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2008, August).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The azimuthal extent of three flux transfer events.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 26, 2353-2369.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-26-2353-2008 Fear, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Milan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Fazakerley, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Owen, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Asikainen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Taylor, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Daly, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2007).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Motion of flux transfer events: a test of the Cooling model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 25(7), 1669–1690.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-25-1669-2007 Fermo, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Drake, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Swisdak, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Hwang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Comparison of a statistical model for magnetic islands in large current layers with Hall MHD simulations and Cluster FTE observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research: Space Physics, 116(A9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2010JA016271 Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Osmane, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hybrid-Vlasov modelling of nightside auroral proton pre- cipitation during southward interplanetary magnetic field conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 37, 791–806.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-37-791-2019 Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Johlander, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hybrid-Vlasov simulation of auroral proton precip- itation in the cusps: Comparison of northward and southward interplanetary magnetic field driving.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Space Weather and Space Climate, 10, 51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1051/swsc/2020053 Hasegawa, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1969).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Drift mirror instability of the magnetosphere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Physics of Flu- ids, 12, 2642-2650.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1063/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1692407 Hasegawa, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sonnerup, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Owen, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Klecker, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Paschmann, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Balogh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & R`eme, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The structure of flux transfer events recovered from cluster data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 24(2), 603–618.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Re- trieved from https://angeo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='copernicus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='org/articles/24/603/2006/ doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-24-603-2006 Hoilijoki, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cassak, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Walsh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Hietala, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', –25– manuscript submitted to Earth and Planetary Physics .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Reconnection rates and X line motion at the magne- topause: Global 2D-3V hybrid-Vlasov simulation results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysi- cal Research (Space Physics), 122(3), 2877-2888.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1002/2016JA023709 Hoilijoki, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sibeck, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cassak, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Properties of Magnetic Reconnection and FTEs on the Dayside Magnetopause With and Without Positive IMF Bx Component Dur- ing Southward IMF.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 124(6), 4037-4048.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2019JA026821 Hoilijoki, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Walsh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', vonˆA Alfthan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Vainio, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Mirror modes in the Earth’s magnetosheath: Re- sults from a global hybrid-Vlasov simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 121(5), 4191-4204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1002/2015JA022026 Jung, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Connor, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Carter, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Koutroumpa, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pagani, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Kuntz, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Solar Minimum Exospheric Neutral Density Near the Subsolar Magne- topause Estimated From the XMM Soft X-Ray Observations on 12 November 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 127(3), e29676.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2021JA029676 Juusola, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Brito, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A possible source mechanism for magneto- tail current sheet flapping.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 36, 1027-1035.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-36-1027-2018 Londrillo, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & del Zanna, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' On the divergence-free condition in Godunov- type schemes for ideal magnetohydrodynamics: the upwind constrained transport method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Computational Physics, 195(1), 17-48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='jcp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='09.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='016 Lucek, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Dunlop, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Horbury, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Balogh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Brown, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cargill, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Oddy, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2001, October).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Cluster magnetic field observations in the magne- tosheath: four-point measurements of mirror structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 19, 1421-1428.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-19-1421-2001 Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Brito, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' von Alfthan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Vlasov methods in space physics and as- trophysics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Living Reviews in Computational Astrophysics, 4, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1007/s41115-018-0003-2 Papadakis, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Alho, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Grandin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Spatial filtering in a 6d hybrid-vlasov scheme to alle- viate adaptive mesh refinement artifacts: a case study with vlasiator (versions 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0, 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1, and 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geoscientific Model Development, 15(20), 7903–7912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/gmd-15-7903-2022 Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Johlander, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Alho, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Hybrid-Vlasov modeling of three-dimensional day- side magnetopause reconnection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Physics of Plasmas, 27(9), 092903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1063/5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='0020685 Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', von Alfthan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sandroos, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Koskela, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Bat- tarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Alho, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' fmihpc/vlasiator: Vlasiator [dataset].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Zenodo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Retrieved from https://zenodo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='org/record/3640593 doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5281/ZENODO.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3640593 Rijnbeek, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cowley, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Southwood, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Russell, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1984).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A survey of dayside flux transfer events observed by ISEE 1 and 2 mag- netometers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research, 89, 786-800.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/ JA089iA02p00786 Robertson, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Collier, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Cravens, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Fok, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' X-ray emission from the terrestrial magnetosheath including the cusps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Jour- nal of Geophysical Research (Space Physics), 111(A12), A12105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2006JA011672 Robertson, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Cravens, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2003a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Spatial maps of heliospheric and geo- –26– manuscript submitted to Earth and Planetary Physics coronal X-ray intensities due to the charge exchange of the solar wind with neutrals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 108(A10), 8031.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2003JA009873 Robertson, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Cravens, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2003b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' X-ray emission from the terrestrial magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geophysical Research Letters, 30(8), 1439.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/ 2002GL016740 Russell, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Elphic, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1978).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Initial ISEE magnetometer results - Magne- topause observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Space Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', 22, 681-715.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1007/BF00212619 Sibeck, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Allen, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Aryan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Bodewits, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Brandt, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Branduardi-Raymont, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Wing, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Imaging Plasma Density Structures in the Soft X- Rays Generated by Solar Wind Charge Exchange with Neutrals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Space Science Reviews, 214(4), 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1007/s11214-018-0504-7 Snowden, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Collier, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Kuntz, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' XMM-Newton Observation of Solar Wind Charge Exchange Emission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Astrophysical Journal, 610(2), 1182-1190.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1086/42184110.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='48550/arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='astro-ph/0404354 Snowden, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Freyberg, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Plucinsky, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Schmitt, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Tr¨umper, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Voges, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Sanders, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (1995).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' First maps of the soft X-ray diffuse background from the ROSAT XRT/PSPC all-sky survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Astrophysical Journal, 454, 643.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1086/176517 Soucek, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Lucek, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Dandouras, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Properties of magnetosheath mir- ror modes observed by cluster and their response to changes in plasma pa- rameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research: Space Physics, 113(A4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2007JA012649 Sun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Connor, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Jorgensen, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Sembay, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Deriving the Magnetopause Position from the Soft X-Ray Image by Using the Tangent Fitting Approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 125(9), e28169.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2020JA028169 Sun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sembay, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Lopez, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Escoubet, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Branduardi- Raymont, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Guo, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Soft X-ray Imaging of the Magne- tosheath and Cusps Under Different Solar Wind Conditions: MHD Simula- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 124(4), 2435-2450.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2018JA026093 Tarvus, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Suni, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Blanco-Cano, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palm- roth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Foreshock cavitons and spontaneous hot flow anomalies: a statistical study with a global hybrid-Vlasov simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Annales Geophysicae, 39(5), 911-928.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='5194/angeo-39-911-2021 Turc, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ganse, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Hoilijoki, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Battarbee, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Juusola, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Foreshock Properties at Typical and Enhanced Inter- planetary Magnetic Field Strengths: Results From Hybrid-Vlasov Simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 123(7), 5476-5493.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2018JA025466 von Alfthan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Pokhotelov, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Hoilijoki, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Honkonen, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sandroos, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Palmroth, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Vlasiator: First global hybrid-Vlasov simula- tions of Earth’s foreshock and magnetosheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Atmospheric and Solar-Terrestrial Physics, 120, 24-35.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='jastp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='08.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='012 Walsh, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Komar, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Pfau-Kempf, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Spacecraft measurements constraining the spatial extent of a magnetopause reconnection X line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geo- physical Research Letters, 44(7), 3038-3046.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1002/2017GL073379 Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', & Sun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Methods to derive the magnetopause from soft X-ray images by the SMILE mission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Geoscience Letters, 9(1), 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1186/ s40562-022-00240-z Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Elphic, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Lavraud, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Taylor, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Birn, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Raeder, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Friedel, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2005).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Initial results of high-latitude magnetopause and low-latitude flank flux transfer events from 3 years of Cluster observa- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research (Space Physics), 110, A11221.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: –27– manuscript submitted to Earth and Planetary Physics 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2005JA011150 Zhang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Kivelson, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Khurana, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', McFadden, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Walker, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', An- gelopoulos, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Auster, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Evidence that crater flux transfer events are initial stages of typical flux transfer events.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Journal of Geophysical Research: Space Physics, 115(A8).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='1029/2009JA015013 Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Ji, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Carter, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', Sembay, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=', .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Zhao, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' Solar Wind Charge Exchange Soft X-Ray Emissions in the Mag- netosphere during an Interplanetary Coronal Mass Ejection Compared to Its Driven Sheath.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' The Astrophysical Journal Letters, 932(1), L1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
+page_content='3847/2041-8213/ac7521 –28–' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WdFQT4oBgHgl3EQfbzbk/content/2301.13325v1.pdf'}
diff --git a/WtAyT4oBgHgl3EQf8_rY/vector_store/index.faiss b/WtAyT4oBgHgl3EQf8_rY/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..3db64dd2dae8de822d4712ad7d55dec5bdcbbd1b
--- /dev/null
+++ b/WtAyT4oBgHgl3EQf8_rY/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46478c4154c72c511ba4f18a04957a9e6c4ec411063c5fdf92db71d0327cbe99
+size 3473453
diff --git a/WtE0T4oBgHgl3EQfVwCs/content/tmp_files/2301.02269v1.pdf.txt b/WtE0T4oBgHgl3EQfVwCs/content/tmp_files/2301.02269v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..173f5b7b073cdb3422abae55849a2ab5771f7779
--- /dev/null
+++ b/WtE0T4oBgHgl3EQfVwCs/content/tmp_files/2301.02269v1.pdf.txt
@@ -0,0 +1,1158 @@
+Taming the Rotating Wave Approximation
+Daniel Burgarth1, Paolo Facchi2, Robin Hillier3, and Marilena Ligabò4
+1Center for Engineered Quantum Systems, Macquarie University, 2109 NSW, Australia
+2Dipartimento di Fisica, Università di Bari, I-70126 Bari, Italy, and INFN, Sezione di
+Bari, I-70126 Bari, Italy
+3Department of Mathematics and Statistics, Lancaster University, Lancaster LA1 4YF,
+UK
+4Dipartimento di Matematica, Università di Bari, I-70125 Bari, Italy
+January 9, 2023
+Abstract
+The interaction between light and matter is one of the oldest research areas of quantum
+mechanics, and a field that just keeps on delivering new insights and applications. With
+the arrival of cavity and circuit quantum electrodynamics we can now achieve strong light-
+matter couplings which form the basis of most implementations of quantum technology. But
+quantum information processing also has high demands requiring total error rates of fractions
+of percentage in order to be scalable (fault-tolerant) to useful applications. Since errors can
+also arise from modelling, this has brought into center stage one of the key approximations
+of quantum theory, the Rotating Wave Approximation (RWA) of the quantum Rabi model,
+leading to the Jaynes-Cummings Hamiltonian.
+While the RWA is often very good and
+incredibly useful to understand light-matter interactions, there is also growing experimental
+evidence of regimes where it is a bad approximation. Here, we ask and answer a harder
+question: for which experimental parameters is the RWA, although perhaps qualitatively
+adequate, already not good enough to match the demands of scalable quantum technology?
+For example, when is the error at least, and when at most, 1%? To answer this, we develop
+rigorous non-perturbative bounds taming the RWA.
+We find that these bounds not only depend, as expected, on the ratio of the coupling
+strength and the oscillator frequency, but also on the average number of photons in the
+initial state. This confirms recent experiments on photon-dressed Bloch-Siegert shifts. We
+argue that with experiments reporting controllable cavity states with hundreds of photons
+and with quantum error correcting codes exploring more and more of Fock space, this state-
+dependency of the RWA is increasingly relevant for the field of quantum computation, and
+our results pave the way towards a better understanding of those experiments.
+The Rotating Wave Approximation (RWA) is one of the oldest and most important ap-
+proximations in Quantum Theory. The starting point is at the birthplace of Nuclear Magnetic
+Resonance (NMR) in 1938, when Rabi and co-authors realized that rather than using rotating
+1
+arXiv:2301.02269v1 [quant-ph] 5 Jan 2023
+
+fields, “it is more convenient experimentally to use an oscillating field, in which case the transition
+probability is approximately the same for weak oscillating fields near the resonance frequency” [1].
+This was significant: Rabi had shown earlier that the Schrödinger equation for rotating fields
+is easily solved analytically [2]. This approximation was a crucial step in understanding driven
+quantum dynamics, as the time-dependent Schrödinger equation is notoriously hard to solve.
+Perhaps this is the key reason for the popularity [3] of the RWA: it provides understanding
+and intuition of resonant driving.
+In fact, the importance of these ideas and the resulting
+techniques of NMR led to Rabi being awarded the Nobel Laureate in Physics in 1944. But what
+justified the approximation, and how did Rabi get to it? Primarily reporting an experimental
+finding, Rabi himself does not provide justification, but over the last 80 years many different
+theoretical methods were used to provide justification and deeper understanding of the RWA
+(the literature is extensive, but see for instance [4, 5, 6, 7, 8]).
+Rabi described the atom as a two-level system and the field classically. In the full quantum
+description of light-matter interaction the situation is much more complicated. By the 1960s
+Quantum Electrodynamics was well established, and the electromagnetic field is now itself a
+quantum system described by unbounded operators. Jaynes and Cummings [9] developed the
+full quantum mechanical version of the Rabi model (now called Quantum Rabi Model)
+H = Ω
+2 σz + ωa†a + λσx(a + a†),
+(1)
+and applied the RWA to obtain the Jaynes-Cummings model
+HRWA = Ω
+2 σz + ωa†a + λ(σ+a + σ−a†).
+(2)
+Here, Ω is the energy difference between the two states of the atom, ω the light frequency and λ
+the strength of the light-matter coupling; we always use ℏ = 1.
+Due to its simplicity and wide range of applicability, the Jaynes-Cummings model is the
+main work horse of light-matter interactions and, by extension, quantum technology. For an
+excellent overview of its scope see [10]. While at the time of the original paper the RWA was
+rather natural, given that the bare coupling between matter and light tends to be extremely
+weak, in cavity and circuit QED nowadays it is well understood that the effective coupling can
+be enhanced to a level where the RWA breaks down. This is often referred to as the Ultrastrong
+Coupling regime. For examples of experiments, see [11] and [12], for a recent review see [13].
+While there is no rigorous derivation of the RWA for the Jaynes-Cummings model till date, the
+common lore is that the ratio g ≡ λ/ω between the light-matter coupling and the light frequency
+is the key parameter [10]. This is motivated by perturbative arguments and of course backed
+up by extensive numerical studies and simulations. For a summary of the different regimes see
+Table 1.1 in [10], where it is argued that for g ≈ 0.1 the RWA breaks down. On the other hand,
+this picture changes for high photon numbers. Indeed, Walls showed [14] that the Bloch-Siegert
+shift (taken as a sign of the breakdown of the RWA) scales with the number of photons. This was
+also observed experimentally [15]. See also [16] for a perturbative argument that λ
+�
+⟨a†a⟩ ≪ ω
+is a more relevant condition in that regime.
+2
+
+Figure 1: Light-matter interactions have been a major driver in quantum physics for half a
+decade. Often, atoms are placed into cavities to amplify their effective coupling strength with
+photons. Here, we show that the rotating wave approximation is not only determined by such
+coupling strength and the frequency of the driving, but also by the number of photons (naively
+depicted as golden spheres) in the cavity.
+What this means is that the quality of the RWA does not only depend on the parameters
+of the model, but also on the initial state of the system. See Fig. 2 for a numerical example.
+Indeed, we prove that there are short times [17] t ≤ π/ω for which
+∥e−itH − e−itHRWA∥ ≥ 1
+6
+(3)
+for any parameter value. This should be considered as a big error, because the biggest difference
+between two unitaries is 2 and because modern quantum technology demands errors well below
+1% (see below). Does this mean that the RWA is wrong? No, because we also show that for any
+state ϕ and any time t,
+e−itHϕ − e−itHRWAϕ → 0,
+as g → 0.
+(4)
+This is our main result, providing a rigorous justification to the RWA. It does not contradict
+Eq. (3), but is a typical phenomenon of unbounded Hamiltonians such as H and HRWA: there is
+no norm convergence, only state-dependent convergence. This is one the key technicalities that
+make it hard to apply standard perturbative arguments for the RWA.
+Let us discuss the relevance of this photon-dependence in the context of quantum technology.
+For fault-tolerant quantum computation, very high fidelity with error rates < 10−3 are required
+[18]. Moreover, modern qubit designs such as GKP [19] and CAT qubits use cavity states and
+3
+
+Exact Numerics
+Upper Bound
+Lower Bound
+10
+20
+30
+40
+50
+Photons
+0.02
+0.04
+0.06
+0.08
+0.10
+0.12
+Error
+Figure 2: Bounding the error of the RWA. We consider a Fock state evolving under the quantum
+Rabi and the Jaynes-Cummings model, respectively. We show our analytical upper and lower
+bounds and the exact numerical norm difference between the two models. We see that the error
+grows with the photon number, and that the bounds provide a good understanding of the scaling
+(other parameters here g = λ
+ω =
+1
+100,∆ = 0 , t ≈ 0.04/ω).
+explore high numbers of photons.
+In particular, CAT states have been created with about
+100 photons [20]. It is therefore necessary to have a good handle of the error of the RWA. Since
+quantum algorithms also invoke dynamics, it is not sufficient to simply match spectral properties,
+as it is usually done, but we need to bound the difference in evolution operators. The interesting
+evolution time regime here are short times up to π/ω: already there, the RWA dynamics can
+deviate substantially. We show that the maximal error ϵn that the RWA has for an n−photon
+Fock state in a short time interval up to π/ω is bounded between
+5g
+√
+n + 3 ≥ ϵn ≥ 1
+6 −
+1
+216g2n −
+7
+12n,
+(5)
+proving that the RWA becomes good for small g but bad for large n. Tighter and more general
+bounds and the full proofs of our results are provided in the Appendix.
+See also Fig. 2 for
+numerical examples of these refined bounds. These bounds prove that g√n is the right parameter
+(as anticipated by the perturbative argument [16]) for the validity of the RWA for Fock states.
+For more general states, see the Appendix. These bounds will be useful for experimentalists in
+quantum information to judge if they should apply the RWA or not.
+4
+
+We would now like to explain the idea which allows us to tame the RWA. Although there are
+many different conceptual ideas trying to justify the RWA, almost all of them agree that ‘highly
+oscillatory terms’ in a Hamiltonian may sometimes be discarded to a good approximation. But
+why? Interestingly, some have argued that such terms are not observable, since measurements
+take finite time. This is plausible; however it turns out that even if measurements are instanta-
+neous, the RWA can be taken. Others argue on the basis of first order perturbation theory, when
+the term involves an integral over the Hamiltonian. This gives a good qualitative picture but
+makes it impossible to compute a rigorous and precise picture. In a more recent work [8] a differ-
+ent route was taken: by an integration by part, the difference between two evolutions can indeed
+be written in terms of an integral over the difference of their generating Hamiltonians, where fast
+oscillations average out. This allows one to prove and provide bounds for the RWA, but only in
+the finite dimensional case. Here, we develop an integration by part to unbounded operators. In
+the general case, this is hard, so we are employing several structures of the specific problem of the
+quantum Rabi model to simplify the analysis. First, both H and HRWA are time-independent,
+so we can use the rich theory of semigroups. Secondly, HRWA has many conserved quantities and
+can only increase and decrease the photon number by one. Finally, all involved quantities are
+well-defined on the subspace of rapidly decreasing functions and leave it invariant, which allows
+us to work on that subspace. We refer to the Appendix for the mathematical details.
+To summarize, after decades of work and conjectures around the RWA for the highly relevant
+quantum Rabi model, we have now got a rigorous proof and in addition a complete quantitative
+measure in terms of lower and upper bounds on the error of the approximation. In particular,
+this confirms the experimental and numerical findings that the error becomes large for large
+ratio g between light-matter coupling and light frequency or for large photon numbers and hence
+the dependence on the state of the system.
+In practice, for given fixed photon number and
+given maximally permissible error this tells us how small g has to be in order for the RWA to
+work. Since experiments are working with ever growing systems, our results will be of immediate
+relevance to the understanding and setup of those experiments and further developments in
+quantum technology. We expect that the methods developed for our proof can be applied to
+tame the RWA for other interesting models, such as systems with multiple modes, nonlinearities
+and other descendants of the Jaynes-Cummings model [10].
+Acknowledgements
+DB acknowledges funding by the Australian Research Council (project numbers FT190100106,
+DP210101367, CE170100009).
+PF and ML were partially supported by the Italian National
+Group of Mathematical Physics (GNFM-INdAM), by Istituto Nazionale di Fisica Nucleare
+(INFN) through the project “QUANTUM”, and by Regione Puglia and QuantERA ERA-NET
+Cofund in Quantum Technologies (GA No. 731473), project PACE-IN.
+5
+
+Appendix
+1
+Time evolution of the Rabi and the Jaynes-Cummings models
+We consider the infinite dimensional Hilbert space L2(R), the creation operator a† =
+1
+√
+2
+�
+x − d
+dx
+�
+and the annihilation operator a =
+1
+√
+2
+�
+x + d
+dx
+�
+on Schwartz space S (R). A fundamental feature
+of these two operators is that their commutator is the identity operator, i.e.
+[a, a†] = I.
+(6)
+Now we consider the following two Hamiltonians
+H = Ω
+2 σz ⊗ I + I ⊗ ωa†a + λσx ⊗ (a + a†)
+(7)
+and
+HRWA = Ω
+2 σz ⊗ I + I ⊗ ωa†a + λ(σ+ ⊗ a + σ− ⊗ a†)
+(8)
+on C2 ⊗ S (R), and we also denote their closures by H and HRWA, with suitable dense domains.
+Here λ, ω, Ω ∈ R,
+σx =
+� 0
+1
+1
+0
+�
+,
+σy =
+� 0
+−i
+i
+0
+�
+,
+σz =
+� 1
+0
+0
+−1
+�
+,
+(9)
+are the Pauli matrices and
+σ+ =
+� 0
+1
+0
+0
+�
+,
+σ− =
+� 0
+0
+1
+0
+�
+.
+(10)
+In what follows, we usually work on C2 ⊗ S (R) without saying so explicitly every time, as this
+subspace of C2 ⊗ L2(R) forms a common invariant core of the operators we are studying in this
+section; this can be shown following the standard methods in [22, Sec.X.6]. To simplify the
+notation, we usually use the same notation for an operator on this core and its closure on the
+whole domain.
+1.1
+Interaction picture: time dependent Hamiltonians H1(t) and H2(t)
+We consider the following Hamiltonian
+H0 = ω
+2 σz ⊗ I + I ⊗ ωa†a,
+(11)
+and define for all t ∈ R
+U1(t) = eitH0e−itH,
+U2(t) = eitH0e−itHRWA.
+(12)
+6
+
+We have that for all j ∈ {1, 2}: Uj(0) = I and
+idUj(t)
+dt
+= Hj(t)Uj(t),
+(13)
+where for all t ∈ R:
+H1(t) = eitH0(H − H0)e−itH0,
+H2(t) = eitH0(HRWA − H0)e−itH0,
+(14)
+We define the detuning between field and atom as ∆ = Ω − ω and compute H1(t):
+H1(t)
+=
+eitH0(H − H0)e−itH0
+=
+eitH0
+�∆
+2 σz ⊗ I + λσx ⊗ (a + a†)
+�
+e−itH0
+=
+∆
+2 σz ⊗ I + λ(eitωσz/2σxe−itωσz/2) ⊗ (eitωa†a(a + a†)e−itωa†a)
+(15)
+We get
+eitωσz/2σxe−itωσz/2 = cos (tω) σx − sin (tω) σy,
+(16)
+and
+eitωa†aae−itωa†a = e−itωa,
+eitωa†aa†e−itωa†a = eitωa†.
+(17)
+Therefore,
+H1(t)
+=
+∆
+2 σz ⊗ I + λ (cos (tω) σx − sin (tω) σy) ⊗
+�
+e−itωa + eitωa†�
+=
+∆
+2 σz ⊗ I + λ
+�
+σ+ ⊗ a + σ− ⊗ a† + e2itωσ+ ⊗ a† + e−2itωσ− ⊗ a
+�
+.
+(18)
+In a similar way we compute H2(t):
+H2(t)
+=
+eitH0(HRWA − H0)e−itH0
+=
+eitH0
+�∆
+2 σz ⊗ I + λ(σ+ ⊗ a + σ− ⊗ a†)
+�
+e−itH0
+=
+∆
+2 σz ⊗ I + λ(σ+ ⊗ a + σ− ⊗ a†).
+(19)
+We notice that H2 is time-independent, and again we use the same symbol for its closure.
+1.2
+Computation of U2(t)
+Notice that, since H2 is time-independent, {U2(t)}t∈R is a unitary group: for all t ∈ R,
+U2(t) = eitH0e−itHRWA = e−it( ∆
+2 σz⊗I+λ(σ+⊗a+σ−⊗a†)) = e−itH2.
+(20)
+7
+
+We observe that C2 ⊗ S (R) ⊂ D(H2) is a set of analytic vectors for K. Let
+e1 =
+� 1
+0
+�
+,
+e2 =
+� 0
+1
+�
+(21)
+be the canonical orthonormal basis of C2. Using the following identities
+σ+e1 =
+� 0
+0
+�
+,
+σ+e2 = e1,
+σ−e1 = e2,
+σ−e2 =
+� 0
+0
+�
+,
+(22)
+we compute the even and the odd powers of H2 on vectors e1 ⊗ ψ and e2 ⊗ ψ, with ψ ∈ S (R).
+For all j ∈ N,
+H2j
+2 (e1 ⊗ ψ) =
+j
+�
+ℓ=0
+� j
+ℓ
+�
+λ2ℓ
+�∆
+2
+�2(j−ℓ)
+e1 ⊗ (aa†)ℓψ,
+(23)
+H2j
+2 (e2 ⊗ ψ) =
+j
+�
+ℓ=0
+� j
+ℓ
+�
+λ2ℓ
+�∆
+2
+�2(j−ℓ)
+e2 ⊗ (a†a)ℓψ,
+(24)
+H2j+1
+2
+(e1 ⊗ψ) =
+j
+�
+ℓ=0
+� j
+ℓ
+� �
+λ2ℓ
+�∆
+2
+�2(j−ℓ)+1
+e1 ⊗ (aa†)ℓψ + λ2ℓ+1
+�∆
+2
+�2(j−ℓ)
+e2 ⊗ a†(aa†)ℓψ
+�
+,
+(25)
+H2j+1
+2
+(e2⊗ψ) =
+j
+�
+ℓ=0
+� j
+ℓ
+� �
+−λ2ℓ
+�∆
+2
+�2(j−ℓ)+1
+e2 ⊗ (a†a)ℓψ + λ2ℓ+1
+�∆
+2
+�2(j−ℓ)
+e1 ⊗ a(a†a)ℓψ
+�
+.
+(26)
+Lemma 1.1. U2(t)(C2 ⊗ S (R)) ⊂ C2 ⊗ S (R) for all t ∈ R.
+Proof. By the N-representation theorem for S (R) [21, Thm.V.13], we have that ψ ∈ S (R) if
+and only if for all m ∈ N:
+sup
+n∈N
+|⟨ϕn|ψ⟩|nm < +∞,
+(27)
+where {ϕn}n∈N is the orthonormal eigenbasis of the number operator a†a, i.e. a†aϕn = nϕn for
+all n ∈ N. We have that for all n ∈ N and t ∈ R:
+U2(t)(e1 ⊗ ϕn) = e1 ⊗ an(t)ϕn + e2 ⊗ bn(t)ϕn+1
+(28)
+and
+U2(t)(e2 ⊗ ϕn) = e1 ⊗ cn(t)ϕn−1 + e2 ⊗ dn(t)ϕn
+(29)
+where
+an(t) = cos
+�
+�t
+�
+λ2(n + 1) +
+�∆
+2
+�2
+�
+� − i∆
+2
+sin
+�
+t
+�
+λ2(n + 1) +
+� ∆
+2
+�2
+�
+�
+λ2(n + 1) +
+� ∆
+2
+�2
+,
+(30)
+8
+
+bn(t) = −iλ
+√
+n + 1
+sin
+�
+t
+�
+λ2(n + 2) +
+� ∆
+2
+�2
+�
+�
+λ2(n + 2) +
+� ∆
+2
+�2
+,
+(31)
+cn(t) = −iλ√n
+sin
+�
+t
+�
+λ2n +
+� ∆
+2
+�2
+�
+�
+λ2n +
+� ∆
+2
+�2
+(32)
+and
+dn(t) = cos
+�
+�t
+�
+λ2n +
+�∆
+2
+�2
+�
+� + i∆
+2
+sin
+�
+t
+�
+λ2n +
+� ∆
+2
+�2
+�
+�
+λ2n +
+� ∆
+2
+�2
+.
+(33)
+Let ψ ∈ S (R) and t ∈ R, then
+U2(t)(e1 ⊗ ψ) = e1 ⊗ ψ1 + e2 ⊗ ψ2,
+U2(t)(e2 ⊗ ψ) = e1 ⊗ ψ3 + e2 ⊗ ψ4,
+(34)
+where
+ψ1 =
++∞
+�
+n=0
+⟨ϕn|ψ⟩an(t)ϕn,
+ψ2 =
++∞
+�
+n=0
+⟨ϕn|ψ⟩bn(t)ϕn+1,
+(35)
+and
+ψ3 =
++∞
+�
+n=1
+⟨ϕn|ψ⟩cn(t)ϕn−1,
+ψ4 =
++∞
+�
+n=0
+⟨ϕn|ψ⟩dn(t)ϕn.
+(36)
+Notice that for all m ∈ N and for all j ∈ {1, 2, 3, 4}:
+sup
+n∈N
+|⟨ϕn|ψj⟩|nm < +∞,
+(37)
+hence ψj ∈ S (R) and therefore U2(t)(C ⊗ S (R)) ⊂ C ⊗ S (R).
+Lemma 1.2. For all t ∈ R and Ψ ∈ C2 ⊗ S (R), we have
+� t
+0
+(H2 − H1(s))Ψ ds = −λ sin (tω)
+ω
+�
+eitωσ+ ⊗ a† + e−itωσ− ⊗ a
+�
+Ψ,
+(38)
+and we denote the closure of this operator by S21(t). Moreover:
+• S21(t)(C2 ⊗ S (R)) ⊂ C2 ⊗ S (R) for all t ∈ R;
+• for all Ψ ∈ C2 ⊗ S (R) and for all t ∈ R:
+d
+dtS21(t)Ψ = (H2(t) − H1(t))Ψ.
+9
+
+Proof. On C2 ⊗ S (R), we have
+S21(t)
+:=
+� t
+0
+(H2 − H1(s)) ds
+=
+−λ
+� t
+0
+�
+e2isωσ+ ⊗ a† + e−2isωσ− ⊗ a
+�
+ds
+=
+− λ
+2iω
+�
+e2isω�s=t
+s=0 σ+ ⊗ a† + λ
+2iω
+�
+e−2isω�s=t
+s=0 σ− ⊗ a
+=
+− λ
+2iω
+�
+e2itω − 1
+�
+σ+ ⊗ a† + λ
+2iω
+�
+e−2itω − 1
+�
+σ− ⊗ a
+=
+−λ sin (tω)
+ω
+�
+eitωσ+ ⊗ a† + e−itωσ− ⊗ a
+�
+.
+(39)
+Lemma 1.3. For all Ψ ∈ C2 ⊗ S (R):
+i(U2(t) − U1(t))Ψ
+=
+S21(t)U2(t)Ψ +
+(40)
++i
+� t
+0
+U1(t)U1(s)†(S21(s)H2 − H1(s)S21(s))U2(s)Ψ ds,
+Proof. Let Ψ ∈ C2 ⊗ S (R), by Lemmas 1.1 and 1.2 we have that for all s ∈ R:
+U2(s)Ψ, S21(s)H2U2(s)Ψ, H1(s)S21(s)U2(s)Ψ ∈ C2 ⊗ S (R).
+(41)
+By equation (13) we have that
+i(U2(t) − U1(t))Ψ
+=
+iU1(t)(U1(t)†U2(t) − I)Ψ
+=
+iU1(t)
+�
+U1(s)†U2(s)
+�t
+0 Ψ
+=
+iU1(t)
+� t
+0
+d
+dsU1(s)†U2(s)Ψ ds
+=
+U1(t)
+� t
+0
+U1(s)†(H2 − H1(s))U2(s)Ψ ds.
+(42)
+We observe that for all s ∈ R:
+d
+ds
+�
+U1(s)†S21(s)U2(s)Ψ
+�
+=
+iU1(s)†(H1(s)S21(s) − S21(s)H2)U2(s)Ψ +
++U1(s)†(H2 − H1(s))U2(s)Ψ,
+(43)
+therefore
+i(U2(t) − U1(t))Ψ
+=
+U1(t)
+� t
+0
+U1(s)†(H2 − H1(s))U2(s)Ψ ds
+=
+S21(t)U2(t)Ψ
++i
+� t
+0
+U1(t)U1(s)†(S21(s)H2 − H1(s)S21(s))U2(s)Ψ ds.
+(44)
+10
+
+Notice that a similar Lemma might hold with U1(t) and U2(t) interchanged. This would
+however be much harder to prove, as our current prove relies on the simple structure of the
+Jaynes-Cummings interaction through Lemma 1.1.
+2
+Computation of bounds and the rotating wave approximation
+Without loss of generality, we can assume λ, Ω, ω > 0.
+2.1
+Upper bound for generic vectors
+Theorem 2.1. For all Ψ ∈ C2 ⊗ S (R) and t ∈ R:
+∥(U2(t) − U1(t))Ψ∥ ≤ λ
+ω
+�
+∥(N + 2)1/2Ψ| + |t|
+�
+|∆|∥(N + 2)1/2Ψ∥ + 3λ∥
+�
+(N + 2)(N + 3)
+�1/2Ψ∥
+��
+,
+(45)
+where N = I ⊗ a†a. Moreover for all Ψ ∈ C2 ⊗ L2(R):
+lim
+ω→+∞
+��(e−itHRWA − e−itH)Ψ
+�� = 0,
+(46)
+uniformly for t in compact sets.
+This theorem proves (4). In particular, (46) shows that mathematically the rotating wave
+approximation is correct in the limit ω → ∞. How appropriate the approximation is in practice
+with finite parameters can be computed in (45), which provides a concrete upper bound on the
+norm difference of the time evolution of an initial state under the actual time evolution and
+under the rotating wave approximation.
+Proof. First we prove (45). Let Ψ ∈ C2 ⊗ S (R) and t ∈ R. First of all we observe that
+∥(I ⊗ a)Ψ∥2 = ⟨(I ⊗ a)Ψ|(I ⊗ a)Ψ⟩ = ⟨Ψ|(I ⊗ a†a)Ψ⟩ = ⟨Ψ|NΨ⟩ = ∥N1/2Ψ∥2,
+(47)
+∥(I ⊗ a†)Ψ∥2 = ⟨Ψ|(N + 1)Ψ⟩ = ∥(N + 1)1/2Ψ∥2.
+(48)
+Moreover, the conservation law
+[H2, N] = 0,
+N = P+ + N,
+P+ = σ+σ− ⊗ I,
+(49)
+implies that
+U2(t)†NU2(t)
+=
+U2(t)†NU2(t) − U2(t)†P+U2(t) = N − U2(t)†P+U2(t)
+=
+N +
+�
+P+ − U2(t)†P+U2(t)
+�
+(50)
+≤
+N + 1
+on C2 ⊗ S (R).
+11
+
+Start from the equality
+�
+U2(t) − U1(t)
+�
+Ψ = −iS21(t)U2(t)Ψ +
+� t
+0
+U1(t)U1(s)†�
+S21(s)H2 − H1(s)S21(s)
+�
+U2(s)Ψ ds . (51)
+Let u(t) = U2(t)Ψ. One gets
+∥S21(t)U2(t)Ψ∥2
+=
+⟨S21(t)u(t)|S21(t)u(t)⟩
+=
+�λ sin (tω)
+ω
+�2
+⟨u(t)|
+�
+σ+σ− ⊗ a†a + σ−σ+ ⊗ aa†�
+u(t)⟩
+=
+�λ sin (tω)
+ω
+�2
+∥
+�
+σ+σ− ⊗ n1/2 + σ−σ+ ⊗ (n + 1)1/2�
+u(t)∥2
+≤
+λ2
+ω2 ∥(N + 1)1/2u(t)∥2
+≤
+λ2
+ω2 ∥(N + 2)1/2Ψ∥2.
+(52)
+Moreover, let
+V (t) = H2 − H1(t) = −λ
+�
+e2itωσ+ ⊗ a† + e−2itωσ− ⊗ a
+�
+.
+(53)
+Then
+X(t)
+=
+S21(t)H2 − H1(t)S21(t)
+=
+[S21(t), H2] + V (t)S21(t)
+=
+−λ sin (tω)
+ω
+�
+∆
+�
+−eitωσ+ ⊗ a† + e−itωσ− ⊗ a
+�
++λσ+σ− ⊗
+�
+eitωa†2 − e−itωa2 − eitωa†a
+�
++λσ−σ+ ⊗
+�
+−eitωa†2 + e−itωa2 − e−itωaa†��
+.
+(54)
+We want to estimate ∥X(t)u(t)∥. First we observe that
+∥
+�
+−eitωσ+ ⊗ a† + e−itωσ− ⊗ a
+�
+u(t)∥ ≤ ∥(N + 1)1/2u(t)∥.
+(55)
+Moreover for all ψ ∈ S (R):
+∥
+�
+eitωa†2 − e−itωa2 − eitωa†a
+�
+ψ∥
+≤
+∥a†2ψ∥ + ∥a2ψ∥ + ∥a†aψ∥
+≤
+3∥((a†a + 1)(a†a + 2))1/2ψ∥
+(56)
+and
+∥
+�
+−eitωa†2 + e−itωa2 − e−itωaa†�
+ψ∥
+≤
+∥a†2ψ∥ + ∥a2ψ∥ + ∥aa†aψ∥
+≤
+3∥((a†a + 1)(a†a + 2))1/2ψ∥.
+(57)
+12
+
+Therefore,
+∥X(t)u(t)∥
+≤
+λ
+ω
+�
+|∆|∥(N + 1)1/2u(t)∥ + 3λ∥((N + 1)(N + 2))1/2u(t)∥
+�
+≤
+λ
+ω
+�
+|∆|∥(N + 2)1/2Ψ∥ + 3λ∥((N + 2)(N + 3))1/2Ψ∥
+�
+.
+(58)
+Taking things together we get
+∥(U2(t) − U1(t))Ψ∥ ≤ λ
+ω
+�
+∥(N +2)1/2Ψ∥+|t|
+�
+|∆|∥(N +2)1/2Ψ∥+3λ∥
+�
+(N +2)(N +3)
+�1/2Ψ∥
+��
+.
+(59)
+Therefore
+lim
+ω→+∞
+��(e−itHRWA − e−itH)Ψ
+�� =
+lim
+ω→+∞ ∥(U2(t) − U1(t))Ψ∥ = 0,
+(60)
+for all Ψ ∈ C2 ⊗ S (R), and since the latter is dense in C2 ⊗ L2(R), (46) follows.
+2.2
+Lower bound
+Theorem 2.2. For all Ψ ∈ C2 ⊗ S (R) and for all 0 ≤ t ≤ π/ω:
+∥(U2(t) − U1(t))Ψ∥
+≥
+λ
+ω sin(tω)∥(N − 1)1/2
++ Ψ∥
+− λ
+ω2 (1 − cos(tω))
+�
+|∆|∥(N + 2)1/2Ψ∥ + 3λ∥
+�
+(N + 2)(N + 3)
+�1/2Ψ∥
+�
+,
+where (N − 1)+ denotes the positive part of the operator N − 1.
+This theorem reveals the limitation of the rotating wave approximation in practice: the error
+grows with the photon number, so for larger systems the rotating wave approximation may no
+longer be justified.
+Proof. Start from the equality
+�
+U2(t) − U1(t)
+�
+Ψ = −iS21(t)U2(t)Ψ +
+� t
+0
+U1(t)U1(s)†�
+S21(s)H2 − H1(s)S21(s)
+�
+U2(s)Ψ ds . (61)
+We have that
+∥(U2(t) − U1(t))Ψ∥
+≥
+∥S21(t)Ψ∥ −
+� t
+0
+∥(S21(s)H2 − H1(s)S21(s))U2(s)Ψ∥ ds
+(62)
+We define u(t) = U2(t)Ψ and we have
+∥S21(t)U2(t)Ψ∥2
+=
+�λ sin (tω)
+ω
+�2
+⟨u(t)|
+�
+σ+σ− ⊗ a†a + σ−σ+ ⊗ aa†�
+u(t)⟩
+≥
+�λ sin (tω)
+ω
+�2
+∥N1/2u(t)∥2,
+(63)
+13
+
+moreover, by (50), we have
+U2(t)†NU2(t) ≥ N − 1.
+(64)
+Hence, one can show that
+∥S21(t)U2(t)Ψ∥2 ≥
+�λ sin (tω)
+ω
+�2
+∥(N − 1)1/2
++ Ψ∥2.
+(65)
+Moreover, for 0 ≤ t ≤ π/ω we have
+� t
+0
+∥(S21(s)H2 − H1(s)S21(s))U2(s)Ψ∥ ds
+≤ λ
+ω
+�
+|∆|∥(N + 2)1/2Ψ∥ + 3λ∥((N + 2)(N + 3))1/2Ψ∥
+� � t
+0
+sin(ωs) ds
+= λ(1 − cos(ωt))
+ω2
+�
+|∆|∥(N + 2)1/2Ψ∥ + 3λ∥((N + 2)(N + 3))1/2Ψ∥
+�
+(66)
+and hence, for 0 ≤ t ≤ π/ω, we get
+∥(U2(t) − U1(t))Ψ∥
+≥
+∥S21(t)Ψ∥ −
+� t
+0
+∥(S21(s)H2 − H1(s)S21(s))U2(s)Ψ∥ ds
+≥
+λ
+ω sin(tω)∥(N − 1)1/2
++ Ψ∥
+(67)
+− λ
+ω2 (1 − cos(tω))
+�
+|∆|∥(N + 2)1/2Ψ∥ + 3λ∥
+�
+(N + 2)(N + 3)
+�1/2Ψ∥
+�
+.
+2.3
+Applying the bounds to Fock states
+To understand the scaling better, let us apply the above bounds to (normalised) Fock states
+Φj,n = ej ⊗ ϕn ∈ C2 ⊗ S (R), with j ∈ {1, 2} and n ∈ N, n > 0. For simplicity, we consider the
+case ∆ = 0, but the argument is easily generalised. From the upper bound (45) we obtain
+∥(U2(t) − U1(t))Φj,n∥ ≤ λ(n + 2)1/2
+ω
+�
+1 + 3|t|λ(n + 3)1/2
+�
+.
+(68)
+For the lower bound we have, for 0 ≤ t ≤ π/ω,
+∥(U2(t)−U1(t))Φj,n∥ ≥ λ
+ω sin(tω)(n − 1)1/2 − 3λ2
+ω2 (1 − cos(tω))
+�
+(n + 2)(n + 3)
+�1/2.
+(69)
+We compute that this as a function of t has a maximum at
+t∗ =
+cos−1
+�
+�
+3λ
+�
+9λ2+
+(n−1)ω2
+(n+2)(n+3)
+�
+�
+ω
+(70)
+14
+
+At this time, the right hand side of Eq. (69) evaluates as
+g
+�
+9g2(n + 2)(n + 3) − 3g
+�
+(n + 2)(n + 3) (9g2(n + 2)(n + 3) + n − 1) + n − 1
+�
+�
+9g2(n + 2)(n + 3) + n − 1
+.
+(71)
+Here, we have set g ≡ λ
+ω.
+We can expand this in order of n−1 to obtain a slightly simpler exact lower bound (valid for
+n > 0)
+sup
+t∈[0, π
+ω ]
+∥(U2(t)−U1(t))Φj,n∥ ≥ 1
+6 −
+1
+216g2n −
+7
+12n.
+(72)
+Focussing on the same time interval also for the upper bound and linearising it, we can conclude
+that
+5g
+√
+n + 3 ≥ sup
+t∈[0, π
+ω ]
+∥(U2(t)−U1(t))Φj,n∥ ≥ 1
+6 −
+1
+216g2n −
+7
+12n,
+(73)
+proving (5). This bound is not necessarily a sharp bound but it shows nicely that the short-time
+error becomes small for small g and large for high photon number n, hence it provides us with a
+quantitative condition on g in order to reduce the error below a certain bound for given photon
+number. For high photon numbers n → ∞, there is a time such that the difference becomes
+greater than 1
+6, i.e.,
+∥e−itH − e−itHRWA∥ ≥ 1
+6
+(74)
+by taking the supremum over all states in (72), which means that the rotating wave approximation
+does not work for arbitrarily high photon numbers; this proves (3).
+References
+[1] I. I. Rabi, J. R. Zacharias, S. Millman, and P. Kusch, A New Method of Measuring Nuclear
+Magnetic Moment. Physical Review 53, 318 (1938).
+[2] I. I. Rabi, Space Quantization in a Gyrating Magnetic Field. Physical Review 51, 652 (1937).
+[3] Google Scholar reports close to a million hits.
+[4] F. Bloch and A. Siegert, Magnetic Resonance for Nonrotating Fields. Physical Review 57,
+522 (1940).
+[5] J. H. Shirley, Solution of the schrödinger equation with a hamiltonian periodic in time.
+Physical Review 138, B979 (1965).
+[6] U. Haeberlen and J. S. Waugh, Coherent Averaging Effects in Magnetic Resonance. Physical
+Review 175, 453 (1968).
+15
+
+[7] G. S. Agarwal, Rotating-Wave Approximation and Spontaneous Emission. Physical Review
+A 7, 1195 (1973).
+[8] D. Burgarth, P. Facchi, G. Gramegna, and K. Yuasa, One bound to rule them all: from
+Adiabatic to Zeno. Quantum 6, 737 (2022).
+[9] E. Jaynes and F. Cummings, Comparison of quantum and semiclassical radiation theories
+with application to the beam maser. Proceedings of the IEEE 51, 89 (1963).
+[10] J. Larson and T. Mavrogordatos, The Jaynes-Cummings model and its descendants modern
+research directions. IoP Publishing (2021).
+[11] P. Forn-Díaz, J. Lisenfeld, D. Marcos, J. J. García-Ripoll, E. Solano, C. J. P. M. Harmans,
+and J. E. Mooij, Observation of the Bloch-Siegert Shift in a Qubit-Oscillator System in the
+Ultrastrong Coupling Regime. Physical Review Letters 105, 237001 (2010).
+[12] X. Li, M. Bamba, Q. Zhang, S. Fallahi, G. C. Gardner, W. Gao, M. Lou, K. Yoshioka, M. J.
+Manfra, and J. Kono, Vacuum Bloch-Siegert shift in Landau polaritons with ultra-high
+cooperativity. Nature Photonics 12, 324 (2018).
+[13] A. Frisk, A. Miranowicz, S. De Liberato, S. Savasta, and F. Nori, Ultrastrong coupling
+between light and matter. Nature Review Physics 1, 19 (2019).
+[14] D. Walls, Higher order effects in the single atom field mode interaction. Physics Letters A
+42, 217 (1972).
+[15] S.-P. Wang, G.-Q. Zhang, Y. Wang, Z. Chen, T. Li, J. S. Tsai, S.-Y. Zhu, and J. Q.
+You, Photon-Dressed Bloch-Siegert Shift in an Ultrastrongly Coupled Circuit Quantum
+Electrodynamical System. Physical Review Applied 13 ,054063 (2020).
+[16] R. R. Puri, Mathematical Methods of Quantum Optics. Springer (2011).
+[17] The reason we focus on short times here is a purely technical requirement from the proof
+(see Appendix). Indeed, numerics shows that the errors are even larger for generic later
+times.
+[18] D. Gottesman, Opportunities and Challenges in Fault-Tolerant Quantum Computation.
+arXiv:2210.15844 (2022).
+[19] A. Grimsmo and S. Puri, Quantum Error Correction with the Gottesman-Kitaev-Preskill
+Code. PRX Quantum, 2, 020101 (2021).
+[20] B. Vlastakis, G. Kirchmair, Z. Leghtas, S. E. Nigg, L. Frunzio, S. M. Girvin, M. Mirrahimi,
+M. H. Devoret, and R. J. Schoelkopf, Deterministically Encoding Quantum Information
+Using 100-Photon Schrödinger Cat States. Science 342, 607 (2013).
+[21] M. Reed and B. Simon. Methods of mathematical physics 1. Functional analysis. Academic
+Press (1980).
+16
+
+[22] M. Reed and B. Simon. Methods of mathematical physics 2. Fourier analysis, self-
+adjointness. Academic Press (1975).
+17
+
diff --git a/WtE0T4oBgHgl3EQfVwCs/content/tmp_files/load_file.txt b/WtE0T4oBgHgl3EQfVwCs/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bfc8daae1c3be78526428d68ed10805a06323ce7
--- /dev/null
+++ b/WtE0T4oBgHgl3EQfVwCs/content/tmp_files/load_file.txt
@@ -0,0 +1,375 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf,len=374
+page_content='Taming the Rotating Wave Approximation Daniel Burgarth1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Paolo Facchi2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Robin Hillier3,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' and Marilena Ligabò4 1Center for Engineered Quantum Systems,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Macquarie University,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2109 NSW,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Australia 2Dipartimento di Fisica,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Università di Bari,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' I-70126 Bari,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Italy,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' and INFN,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Sezione di Bari,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' I-70126 Bari,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Italy 3Department of Mathematics and Statistics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Lancaster University,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Lancaster LA1 4YF,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' UK 4Dipartimento di Matematica,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Università di Bari,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' I-70125 Bari,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Italy January 9,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2023 Abstract The interaction between light and matter is one of the oldest research areas of quantum mechanics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' and a field that just keeps on delivering new insights and applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' With the arrival of cavity and circuit quantum electrodynamics we can now achieve strong light- matter couplings which form the basis of most implementations of quantum technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' But quantum information processing also has high demands requiring total error rates of fractions of percentage in order to be scalable (fault-tolerant) to useful applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Since errors can also arise from modelling, this has brought into center stage one of the key approximations of quantum theory, the Rotating Wave Approximation (RWA) of the quantum Rabi model, leading to the Jaynes-Cummings Hamiltonian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' While the RWA is often very good and incredibly useful to understand light-matter interactions, there is also growing experimental evidence of regimes where it is a bad approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Here, we ask and answer a harder question: for which experimental parameters is the RWA, although perhaps qualitatively adequate, already not good enough to match the demands of scalable quantum technology?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For example, when is the error at least, and when at most, 1%?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' To answer this, we develop rigorous non-perturbative bounds taming the RWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We find that these bounds not only depend, as expected, on the ratio of the coupling strength and the oscillator frequency, but also on the average number of photons in the initial state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This confirms recent experiments on photon-dressed Bloch-Siegert shifts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We argue that with experiments reporting controllable cavity states with hundreds of photons and with quantum error correcting codes exploring more and more of Fock space, this state- dependency of the RWA is increasingly relevant for the field of quantum computation, and our results pave the way towards a better understanding of those experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' The Rotating Wave Approximation (RWA) is one of the oldest and most important ap- proximations in Quantum Theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' The starting point is at the birthplace of Nuclear Magnetic Resonance (NMR) in 1938, when Rabi and co-authors realized that rather than using rotating 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='02269v1 [quant-ph] 5 Jan 2023 fields, “it is more convenient experimentally to use an oscillating field, in which case the transition probability is approximately the same for weak oscillating fields near the resonance frequency” [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This was significant: Rabi had shown earlier that the Schrödinger equation for rotating fields is easily solved analytically [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This approximation was a crucial step in understanding driven quantum dynamics, as the time-dependent Schrödinger equation is notoriously hard to solve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Perhaps this is the key reason for the popularity [3] of the RWA: it provides understanding and intuition of resonant driving.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In fact, the importance of these ideas and the resulting techniques of NMR led to Rabi being awarded the Nobel Laureate in Physics in 1944.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' But what justified the approximation, and how did Rabi get to it?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Primarily reporting an experimental finding, Rabi himself does not provide justification, but over the last 80 years many different theoretical methods were used to provide justification and deeper understanding of the RWA (the literature is extensive, but see for instance [4, 5, 6, 7, 8]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Rabi described the atom as a two-level system and the field classically.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In the full quantum description of light-matter interaction the situation is much more complicated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' By the 1960s Quantum Electrodynamics was well established, and the electromagnetic field is now itself a quantum system described by unbounded operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Jaynes and Cummings [9] developed the full quantum mechanical version of the Rabi model (now called Quantum Rabi Model) H = Ω 2 σz + ωa†a + λσx(a + a†), (1) and applied the RWA to obtain the Jaynes-Cummings model HRWA = Ω 2 σz + ωa†a + λ(σ+a + σ−a†).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (2) Here, Ω is the energy difference between the two states of the atom, ω the light frequency and λ the strength of the light-matter coupling;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' we always use ℏ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Due to its simplicity and wide range of applicability, the Jaynes-Cummings model is the main work horse of light-matter interactions and, by extension, quantum technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For an excellent overview of its scope see [10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' While at the time of the original paper the RWA was rather natural, given that the bare coupling between matter and light tends to be extremely weak, in cavity and circuit QED nowadays it is well understood that the effective coupling can be enhanced to a level where the RWA breaks down.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This is often referred to as the Ultrastrong Coupling regime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For examples of experiments, see [11] and [12], for a recent review see [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' While there is no rigorous derivation of the RWA for the Jaynes-Cummings model till date, the common lore is that the ratio g ≡ λ/ω between the light-matter coupling and the light frequency is the key parameter [10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This is motivated by perturbative arguments and of course backed up by extensive numerical studies and simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For a summary of the different regimes see Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1 in [10], where it is argued that for g ≈ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1 the RWA breaks down.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' On the other hand, this picture changes for high photon numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Indeed, Walls showed [14] that the Bloch-Siegert shift (taken as a sign of the breakdown of the RWA) scales with the number of photons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This was also observed experimentally [15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' See also [16] for a perturbative argument that λ � ⟨a†a⟩ ≪ ω is a more relevant condition in that regime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2 Figure 1: Light-matter interactions have been a major driver in quantum physics for half a decade.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Often, atoms are placed into cavities to amplify their effective coupling strength with photons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Here, we show that the rotating wave approximation is not only determined by such coupling strength and the frequency of the driving, but also by the number of photons (naively depicted as golden spheres) in the cavity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' What this means is that the quality of the RWA does not only depend on the parameters of the model, but also on the initial state of the system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' See Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2 for a numerical example.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Indeed, we prove that there are short times [17] t ≤ π/ω for which ∥e−itH − e−itHRWA∥ ≥ 1 6 (3) for any parameter value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This should be considered as a big error, because the biggest difference between two unitaries is 2 and because modern quantum technology demands errors well below 1% (see below).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Does this mean that the RWA is wrong?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' No, because we also show that for any state ϕ and any time t, e−itHϕ − e−itHRWAϕ → 0, as g → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (4) This is our main result, providing a rigorous justification to the RWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' It does not contradict Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (3), but is a typical phenomenon of unbounded Hamiltonians such as H and HRWA: there is no norm convergence, only state-dependent convergence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This is one the key technicalities that make it hard to apply standard perturbative arguments for the RWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Let us discuss the relevance of this photon-dependence in the context of quantum technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For fault-tolerant quantum computation, very high fidelity with error rates < 10−3 are required [18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Moreover, modern qubit designs such as GKP [19] and CAT qubits use cavity states and 3 Exact Numerics Upper Bound Lower Bound 10 20 30 40 50 Photons 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='12 Error Figure 2: Bounding the error of the RWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We consider a Fock state evolving under the quantum Rabi and the Jaynes-Cummings model, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We show our analytical upper and lower bounds and the exact numerical norm difference between the two models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We see that the error grows with the photon number, and that the bounds provide a good understanding of the scaling (other parameters here g = λ ω = 1 100,∆ = 0 , t ≈ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='04/ω).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' explore high numbers of photons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In particular, CAT states have been created with about 100 photons [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' It is therefore necessary to have a good handle of the error of the RWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Since quantum algorithms also invoke dynamics, it is not sufficient to simply match spectral properties, as it is usually done, but we need to bound the difference in evolution operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' The interesting evolution time regime here are short times up to π/ω: already there, the RWA dynamics can deviate substantially.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We show that the maximal error ϵn that the RWA has for an n−photon Fock state in a short time interval up to π/ω is bounded between 5g √ n + 3 ≥ ϵn ≥ 1 6 − 1 216g2n − 7 12n, (5) proving that the RWA becomes good for small g but bad for large n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Tighter and more general bounds and the full proofs of our results are provided in the Appendix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' See also Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2 for numerical examples of these refined bounds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' These bounds prove that g√n is the right parameter (as anticipated by the perturbative argument [16]) for the validity of the RWA for Fock states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For more general states, see the Appendix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' These bounds will be useful for experimentalists in quantum information to judge if they should apply the RWA or not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 4 We would now like to explain the idea which allows us to tame the RWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Although there are many different conceptual ideas trying to justify the RWA, almost all of them agree that ‘highly oscillatory terms’ in a Hamiltonian may sometimes be discarded to a good approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' But why?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Interestingly, some have argued that such terms are not observable, since measurements take finite time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This is plausible;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' however it turns out that even if measurements are instanta- neous, the RWA can be taken.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Others argue on the basis of first order perturbation theory, when the term involves an integral over the Hamiltonian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This gives a good qualitative picture but makes it impossible to compute a rigorous and precise picture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In a more recent work [8] a differ- ent route was taken: by an integration by part, the difference between two evolutions can indeed be written in terms of an integral over the difference of their generating Hamiltonians, where fast oscillations average out.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This allows one to prove and provide bounds for the RWA, but only in the finite dimensional case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Here, we develop an integration by part to unbounded operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In the general case, this is hard, so we are employing several structures of the specific problem of the quantum Rabi model to simplify the analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' First, both H and HRWA are time-independent, so we can use the rich theory of semigroups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Secondly, HRWA has many conserved quantities and can only increase and decrease the photon number by one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Finally, all involved quantities are well-defined on the subspace of rapidly decreasing functions and leave it invariant, which allows us to work on that subspace.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We refer to the Appendix for the mathematical details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' To summarize, after decades of work and conjectures around the RWA for the highly relevant quantum Rabi model, we have now got a rigorous proof and in addition a complete quantitative measure in terms of lower and upper bounds on the error of the approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In particular, this confirms the experimental and numerical findings that the error becomes large for large ratio g between light-matter coupling and light frequency or for large photon numbers and hence the dependence on the state of the system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In practice, for given fixed photon number and given maximally permissible error this tells us how small g has to be in order for the RWA to work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Since experiments are working with ever growing systems, our results will be of immediate relevance to the understanding and setup of those experiments and further developments in quantum technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We expect that the methods developed for our proof can be applied to tame the RWA for other interesting models, such as systems with multiple modes, nonlinearities and other descendants of the Jaynes-Cummings model [10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Acknowledgements DB acknowledges funding by the Australian Research Council (project numbers FT190100106, DP210101367, CE170100009).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' PF and ML were partially supported by the Italian National Group of Mathematical Physics (GNFM-INdAM), by Istituto Nazionale di Fisica Nucleare (INFN) through the project “QUANTUM”, and by Regione Puglia and QuantERA ERA-NET Cofund in Quantum Technologies (GA No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 731473), project PACE-IN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 5 Appendix 1 Time evolution of the Rabi and the Jaynes-Cummings models We consider the infinite dimensional Hilbert space L2(R), the creation operator a† = 1 √ 2 � x − d dx � and the annihilation operator a = 1 √ 2 � x + d dx � on Schwartz space S (R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' A fundamental feature of these two operators is that their commutator is the identity operator, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [a, a†] = I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (6) Now we consider the following two Hamiltonians H = Ω 2 σz ⊗ I + I ⊗ ωa†a + λσx ⊗ (a + a†) (7) and HRWA = Ω 2 σz ⊗ I + I ⊗ ωa†a + λ(σ+ ⊗ a + σ− ⊗ a†) (8) on C2 ⊗ S (R), and we also denote their closures by H and HRWA, with suitable dense domains.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Here λ, ω, Ω ∈ R, σx = � 0 1 1 0 � , σy = � 0 −i i 0 � , σz = � 1 0 0 −1 � , (9) are the Pauli matrices and σ+ = � 0 1 0 0 � , σ− = � 0 0 1 0 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (10) In what follows, we usually work on C2 ⊗ S (R) without saying so explicitly every time, as this subspace of C2 ⊗ L2(R) forms a common invariant core of the operators we are studying in this section;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' this can be shown following the standard methods in [22, Sec.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' To simplify the notation, we usually use the same notation for an operator on this core and its closure on the whole domain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1 Interaction picture: time dependent Hamiltonians H1(t) and H2(t) We consider the following Hamiltonian H0 = ω 2 σz ⊗ I + I ⊗ ωa†a, (11) and define for all t ∈ R U1(t) = eitH0e−itH, U2(t) = eitH0e−itHRWA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (12) 6 We have that for all j ∈ {1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2}: Uj(0) = I and idUj(t) dt = Hj(t)Uj(t),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (13) where for all t ∈ R: H1(t) = eitH0(H − H0)e−itH0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' H2(t) = eitH0(HRWA − H0)e−itH0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (14) We define the detuning between field and atom as ∆ = Ω − ω and compute H1(t): H1(t) = eitH0(H − H0)e−itH0 = eitH0 �∆ 2 σz ⊗ I + λσx ⊗ (a + a†) � e−itH0 = ∆ 2 σz ⊗ I + λ(eitωσz/2σxe−itωσz/2) ⊗ (eitωa†a(a + a†)e−itωa†a) (15) We get eitωσz/2σxe−itωσz/2 = cos (tω) σx − sin (tω) σy,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (16) and eitωa†aae−itωa†a = e−itωa,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' eitωa†aa†e−itωa†a = eitωa†.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (17) Therefore, H1(t) = ∆ 2 σz ⊗ I + λ (cos (tω) σx − sin (tω) σy) ⊗ � e−itωa + eitωa†� = ∆ 2 σz ⊗ I + λ � σ+ ⊗ a + σ− ⊗ a† + e2itωσ+ ⊗ a† + e−2itωσ− ⊗ a � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (18) In a similar way we compute H2(t): H2(t) = eitH0(HRWA − H0)e−itH0 = eitH0 �∆ 2 σz ⊗ I + λ(σ+ ⊗ a + σ− ⊗ a†) � e−itH0 = ∆ 2 σz ⊗ I + λ(σ+ ⊗ a + σ− ⊗ a†).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (19) We notice that H2 is time-independent, and again we use the same symbol for its closure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='2 Computation of U2(t) Notice that, since H2 is time-independent, {U2(t)}t∈R is a unitary group: for all t ∈ R, U2(t) = eitH0e−itHRWA = e−it( ∆ 2 σz⊗I+λ(σ+⊗a+σ−⊗a†)) = e−itH2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (20) 7 We observe that C2 ⊗ S (R) ⊂ D(H2) is a set of analytic vectors for K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Let e1 = � 1 0 � , e2 = � 0 1 � (21) be the canonical orthonormal basis of C2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Using the following identities σ+e1 = � 0 0 � , σ+e2 = e1, σ−e1 = e2, σ−e2 = � 0 0 � , (22) we compute the even and the odd powers of H2 on vectors e1 ⊗ ψ and e2 ⊗ ψ, with ψ ∈ S (R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For all j ∈ N, H2j 2 (e1 ⊗ ψ) = j � ℓ=0 � j ℓ � λ2ℓ �∆ 2 �2(j−ℓ) e1 ⊗ (aa†)ℓψ, (23) H2j 2 (e2 ⊗ ψ) = j � ℓ=0 � j ℓ � λ2ℓ �∆ 2 �2(j−ℓ) e2 ⊗ (a†a)ℓψ, (24) H2j+1 2 (e1 ⊗ψ) = j � ℓ=0 � j ℓ � � λ2ℓ �∆ 2 �2(j−ℓ)+1 e1 ⊗ (aa†)ℓψ + λ2ℓ+1 �∆ 2 �2(j−ℓ) e2 ⊗ a†(aa†)ℓψ � , (25) H2j+1 2 (e2⊗ψ) = j � ℓ=0 � j ℓ � � −λ2ℓ �∆ 2 �2(j−ℓ)+1 e2 ⊗ (a†a)ℓψ + λ2ℓ+1 �∆ 2 �2(j−ℓ) e1 ⊗ a(a†a)ℓψ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (26) Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' U2(t)(C2 ⊗ S (R)) ⊂ C2 ⊗ S (R) for all t ∈ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' By the N-representation theorem for S (R) [21, Thm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='13], we have that ψ ∈ S (R) if and only if for all m ∈ N: sup n∈N |⟨ϕn|ψ⟩|nm < +∞, (27) where {ϕn}n∈N is the orthonormal eigenbasis of the number operator a†a, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' a†aϕn = nϕn for all n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We have that for all n ∈ N and t ∈ R: U2(t)(e1 ⊗ ϕn) = e1 ⊗ an(t)ϕn + e2 ⊗ bn(t)ϕn+1 (28) and U2(t)(e2 ⊗ ϕn) = e1 ⊗ cn(t)ϕn−1 + e2 ⊗ dn(t)ϕn (29) where an(t) = cos � �t � λ2(n + 1) + �∆ 2 �2 � � − i∆ 2 sin � t � λ2(n + 1) + � ∆ 2 �2 � � λ2(n + 1) + � ∆ 2 �2 ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (30) 8 bn(t) = −iλ √ n + 1 sin � t � λ2(n + 2) + � ∆ 2 �2 � � λ2(n + 2) + � ∆ 2 �2 ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (31) cn(t) = −iλ√n sin � t � λ2n + � ∆ 2 �2 � � λ2n + � ∆ 2 �2 (32) and dn(t) = cos � �t � λ2n + �∆ 2 �2 � � + i∆ 2 sin � t � λ2n + � ∆ 2 �2 � � λ2n + � ∆ 2 �2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (33) Let ψ ∈ S (R) and t ∈ R, then U2(t)(e1 ⊗ ψ) = e1 ⊗ ψ1 + e2 ⊗ ψ2, U2(t)(e2 ⊗ ψ) = e1 ⊗ ψ3 + e2 ⊗ ψ4, (34) where ψ1 = +∞ � n=0 ⟨ϕn|ψ⟩an(t)ϕn, ψ2 = +∞ � n=0 ⟨ϕn|ψ⟩bn(t)ϕn+1, (35) and ψ3 = +∞ � n=1 ⟨ϕn|ψ⟩cn(t)ϕn−1, ψ4 = +∞ � n=0 ⟨ϕn|ψ⟩dn(t)ϕn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (36) Notice that for all m ∈ N and for all j ∈ {1, 2, 3, 4}: sup n∈N |⟨ϕn|ψj⟩|nm < +∞, (37) hence ψj ∈ S (R) and therefore U2(t)(C ⊗ S (R)) ⊂ C ⊗ S (R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For all t ∈ R and Ψ ∈ C2 ⊗ S (R), we have � t 0 (H2 − H1(s))Ψ ds = −λ sin (tω) ω � eitωσ+ ⊗ a† + e−itωσ− ⊗ a � Ψ, (38) and we denote the closure of this operator by S21(t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Moreover: S21(t)(C2 ⊗ S (R)) ⊂ C2 ⊗ S (R) for all t ∈ R;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' for all Ψ ∈ C2 ⊗ S (R) and for all t ∈ R: d dtS21(t)Ψ = (H2(t) − H1(t))Ψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 9 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' On C2 ⊗ S (R), we have S21(t) := � t 0 (H2 − H1(s)) ds = −λ � t 0 � e2isωσ+ ⊗ a† + e−2isωσ− ⊗ a � ds = − λ 2iω � e2isω�s=t s=0 σ+ ⊗ a† + λ 2iω � e−2isω�s=t s=0 σ− ⊗ a = − λ 2iω � e2itω − 1 � σ+ ⊗ a† + λ 2iω � e−2itω − 1 � σ− ⊗ a = −λ sin (tω) ω � eitωσ+ ⊗ a† + e−itωσ− ⊗ a � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (39) Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For all Ψ ∈ C2 ⊗ S (R): i(U2(t) − U1(t))Ψ = S21(t)U2(t)Ψ + (40) +i � t 0 U1(t)U1(s)†(S21(s)H2 − H1(s)S21(s))U2(s)Ψ ds, Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Let Ψ ∈ C2 ⊗ S (R), by Lemmas 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1 and 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='2 we have that for all s ∈ R: U2(s)Ψ, S21(s)H2U2(s)Ψ, H1(s)S21(s)U2(s)Ψ ∈ C2 ⊗ S (R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (41) By equation (13) we have that i(U2(t) − U1(t))Ψ = iU1(t)(U1(t)†U2(t) − I)Ψ = iU1(t) � U1(s)†U2(s) �t 0 Ψ = iU1(t) � t 0 d dsU1(s)†U2(s)Ψ ds = U1(t) � t 0 U1(s)†(H2 − H1(s))U2(s)Ψ ds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (42) We observe that for all s ∈ R: d ds � U1(s)†S21(s)U2(s)Ψ � = iU1(s)†(H1(s)S21(s) − S21(s)H2)U2(s)Ψ + +U1(s)†(H2 − H1(s))U2(s)Ψ, (43) therefore i(U2(t) − U1(t))Ψ = U1(t) � t 0 U1(s)†(H2 − H1(s))U2(s)Ψ ds = S21(t)U2(t)Ψ +i � t 0 U1(t)U1(s)†(S21(s)H2 − H1(s)S21(s))U2(s)Ψ ds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (44) 10 Notice that a similar Lemma might hold with U1(t) and U2(t) interchanged.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This would however be much harder to prove, as our current prove relies on the simple structure of the Jaynes-Cummings interaction through Lemma 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2 Computation of bounds and the rotating wave approximation Without loss of generality, we can assume λ, Ω, ω > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1 Upper bound for generic vectors Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For all Ψ ∈ C2 ⊗ S (R) and t ∈ R: ∥(U2(t) − U1(t))Ψ∥ ≤ λ ω � ∥(N + 2)1/2Ψ| + |t| � |∆|∥(N + 2)1/2Ψ∥ + 3λ∥ � (N + 2)(N + 3) �1/2Ψ∥ �� , (45) where N = I ⊗ a†a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Moreover for all Ψ ∈ C2 ⊗ L2(R): lim ω→+∞ ��(e−itHRWA − e−itH)Ψ �� = 0, (46) uniformly for t in compact sets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This theorem proves (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' In particular, (46) shows that mathematically the rotating wave approximation is correct in the limit ω → ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' How appropriate the approximation is in practice with finite parameters can be computed in (45), which provides a concrete upper bound on the norm difference of the time evolution of an initial state under the actual time evolution and under the rotating wave approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' First we prove (45).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Let Ψ ∈ C2 ⊗ S (R) and t ∈ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' First of all we observe that ∥(I ⊗ a)Ψ∥2 = ⟨(I ⊗ a)Ψ|(I ⊗ a)Ψ⟩ = ⟨Ψ|(I ⊗ a†a)Ψ⟩ = ⟨Ψ|NΨ⟩ = ∥N1/2Ψ∥2, (47) ∥(I ⊗ a†)Ψ∥2 = ⟨Ψ|(N + 1)Ψ⟩ = ∥(N + 1)1/2Ψ∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (48) Moreover, the conservation law [H2, N] = 0, N = P+ + N, P+ = σ+σ− ⊗ I, (49) implies that U2(t)†NU2(t) = U2(t)†NU2(t) − U2(t)†P+U2(t) = N − U2(t)†P+U2(t) = N + � P+ − U2(t)†P+U2(t) � (50) ≤ N + 1 on C2 ⊗ S (R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 11 Start from the equality � U2(t) − U1(t) � Ψ = −iS21(t)U2(t)Ψ + � t 0 U1(t)U1(s)†� S21(s)H2 − H1(s)S21(s) � U2(s)Ψ ds .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (51) Let u(t) = U2(t)Ψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' One gets ∥S21(t)U2(t)Ψ∥2 = ⟨S21(t)u(t)|S21(t)u(t)⟩ = �λ sin (tω) ω �2 ⟨u(t)| � σ+σ− ⊗ a†a + σ−σ+ ⊗ aa†� u(t)⟩ = �λ sin (tω) ω �2 ∥ � σ+σ− ⊗ n1/2 + σ−σ+ ⊗ (n + 1)1/2� u(t)∥2 ≤ λ2 ω2 ∥(N + 1)1/2u(t)∥2 ≤ λ2 ω2 ∥(N + 2)1/2Ψ∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (52) Moreover, let V (t) = H2 − H1(t) = −λ � e2itωσ+ ⊗ a† + e−2itωσ− ⊗ a � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (53) Then X(t) = S21(t)H2 − H1(t)S21(t) = [S21(t), H2] + V (t)S21(t) = −λ sin (tω) ω � ∆ � −eitωσ+ ⊗ a† + e−itωσ− ⊗ a � +λσ+σ− ⊗ � eitωa†2 − e−itωa2 − eitωa†a � +λσ−σ+ ⊗ � −eitωa†2 + e−itωa2 − e−itωaa†�� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (54) We want to estimate ∥X(t)u(t)∥.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' First we observe that ∥ � −eitωσ+ ⊗ a† + e−itωσ− ⊗ a � u(t)∥ ≤ ∥(N + 1)1/2u(t)∥.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (55) Moreover for all ψ ∈ S (R): ∥ � eitωa†2 − e−itωa2 − eitωa†a � ψ∥ ≤ ∥a†2ψ∥ + ∥a2ψ∥ + ∥a†aψ∥ ≤ 3∥((a†a + 1)(a†a + 2))1/2ψ∥ (56) and ∥ � −eitωa†2 + e−itωa2 − e−itωaa†� ψ∥ ≤ ∥a†2ψ∥ + ∥a2ψ∥ + ∥aa†aψ∥ ≤ 3∥((a†a + 1)(a†a + 2))1/2ψ∥.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (57) 12 Therefore, ∥X(t)u(t)∥ ≤ λ ω � |∆|∥(N + 1)1/2u(t)∥ + 3λ∥((N + 1)(N + 2))1/2u(t)∥ � ≤ λ ω � |∆|∥(N + 2)1/2Ψ∥ + 3λ∥((N + 2)(N + 3))1/2Ψ∥ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (58) Taking things together we get ∥(U2(t) − U1(t))Ψ∥ ≤ λ ω � ∥(N +2)1/2Ψ∥+|t| � |∆|∥(N +2)1/2Ψ∥+3λ∥ � (N +2)(N +3) �1/2Ψ∥ �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (59) Therefore lim ω→+∞ ��(e−itHRWA − e−itH)Ψ �� = lim ω→+∞ ∥(U2(t) − U1(t))Ψ∥ = 0, (60) for all Ψ ∈ C2 ⊗ S (R), and since the latter is dense in C2 ⊗ L2(R), (46) follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='2 Lower bound Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For all Ψ ∈ C2 ⊗ S (R) and for all 0 ≤ t ≤ π/ω: ∥(U2(t) − U1(t))Ψ∥ ≥ λ ω sin(tω)∥(N − 1)1/2 + Ψ∥ − λ ω2 (1 − cos(tω)) � |∆|∥(N + 2)1/2Ψ∥ + 3λ∥ � (N + 2)(N + 3) �1/2Ψ∥ � , where (N − 1)+ denotes the positive part of the operator N − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This theorem reveals the limitation of the rotating wave approximation in practice: the error grows with the photon number, so for larger systems the rotating wave approximation may no longer be justified.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Start from the equality � U2(t) − U1(t) � Ψ = −iS21(t)U2(t)Ψ + � t 0 U1(t)U1(s)†� S21(s)H2 − H1(s)S21(s) � U2(s)Ψ ds .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (61) We have that ∥(U2(t) − U1(t))Ψ∥ ≥ ∥S21(t)Ψ∥ − � t 0 ∥(S21(s)H2 − H1(s)S21(s))U2(s)Ψ∥ ds (62) We define u(t) = U2(t)Ψ and we have ∥S21(t)U2(t)Ψ∥2 = �λ sin (tω) ω �2 ⟨u(t)| � σ+σ− ⊗ a†a + σ−σ+ ⊗ aa†� u(t)⟩ ≥ �λ sin (tω) ω �2 ∥N1/2u(t)∥2, (63) 13 moreover, by (50), we have U2(t)†NU2(t) ≥ N − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (64) Hence, one can show that ∥S21(t)U2(t)Ψ∥2 ≥ �λ sin (tω) ω �2 ∥(N − 1)1/2 + Ψ∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (65) Moreover, for 0 ≤ t ≤ π/ω we have � t 0 ∥(S21(s)H2 − H1(s)S21(s))U2(s)Ψ∥ ds ≤ λ ω � |∆|∥(N + 2)1/2Ψ∥ + 3λ∥((N + 2)(N + 3))1/2Ψ∥ � � t 0 sin(ωs) ds = λ(1 − cos(ωt)) ω2 � |∆|∥(N + 2)1/2Ψ∥ + 3λ∥((N + 2)(N + 3))1/2Ψ∥ � (66) and hence, for 0 ≤ t ≤ π/ω, we get ∥(U2(t) − U1(t))Ψ∥ ≥ ∥S21(t)Ψ∥ − � t 0 ∥(S21(s)H2 − H1(s)S21(s))U2(s)Ψ∥ ds ≥ λ ω sin(tω)∥(N − 1)1/2 + Ψ∥ (67) − λ ω2 (1 − cos(tω)) � |∆|∥(N + 2)1/2Ψ∥ + 3λ∥ � (N + 2)(N + 3) �1/2Ψ∥ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='3 Applying the bounds to Fock states To understand the scaling better, let us apply the above bounds to (normalised) Fock states Φj,n = ej ⊗ ϕn ∈ C2 ⊗ S (R), with j ∈ {1, 2} and n ∈ N, n > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For simplicity, we consider the case ∆ = 0, but the argument is easily generalised.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' From the upper bound (45) we obtain ∥(U2(t) − U1(t))Φj,n∥ ≤ λ(n + 2)1/2 ω � 1 + 3|t|λ(n + 3)1/2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (68) For the lower bound we have, for 0 ≤ t ≤ π/ω, ∥(U2(t)−U1(t))Φj,n∥ ≥ λ ω sin(tω)(n − 1)1/2 − 3λ2 ω2 (1 − cos(tω)) � (n + 2)(n + 3) �1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (69) We compute that this as a function of t has a maximum at t∗ = cos−1 � � 3λ � 9λ2+ (n−1)ω2 (n+2)(n+3) � � ω (70) 14 At this time, the right hand side of Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (69) evaluates as g � 9g2(n + 2)(n + 3) − 3g � (n + 2)(n + 3) (9g2(n + 2)(n + 3) + n − 1) + n − 1 � � 9g2(n + 2)(n + 3) + n − 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (71) Here, we have set g ≡ λ ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' We can expand this in order of n−1 to obtain a slightly simpler exact lower bound (valid for n > 0) sup t∈[0, π ω ] ∥(U2(t)−U1(t))Φj,n∥ ≥ 1 6 − 1 216g2n − 7 12n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' (72) Focussing on the same time interval also for the upper bound and linearising it, we can conclude that 5g √ n + 3 ≥ sup t∈[0, π ω ] ∥(U2(t)−U1(t))Φj,n∥ ≥ 1 6 − 1 216g2n − 7 12n, (73) proving (5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' This bound is not necessarily a sharp bound but it shows nicely that the short-time error becomes small for small g and large for high photon number n, hence it provides us with a quantitative condition on g in order to reduce the error below a certain bound for given photon number.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' For high photon numbers n → ∞, there is a time such that the difference becomes greater than 1 6, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=', ∥e−itH − e−itHRWA∥ ≥ 1 6 (74) by taking the supremum over all states in (72), which means that the rotating wave approximation does not work for arbitrarily high photon numbers;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' this proves (3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' References [1] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Rabi, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Zacharias, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Millman, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Kusch, A New Method of Measuring Nuclear Magnetic Moment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review 53, 318 (1938).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [2] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Rabi, Space Quantization in a Gyrating Magnetic Field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review 51, 652 (1937).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [3] Google Scholar reports close to a million hits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [4] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Bloch and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Siegert, Magnetic Resonance for Nonrotating Fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review 57, 522 (1940).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [5] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Shirley, Solution of the schrödinger equation with a hamiltonian periodic in time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review 138, B979 (1965).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [6] U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Haeberlen and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Waugh, Coherent Averaging Effects in Magnetic Resonance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review 175, 453 (1968).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 15 [7] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Agarwal, Rotating-Wave Approximation and Spontaneous Emission.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review A 7, 1195 (1973).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [8] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Burgarth, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Facchi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Gramegna, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Yuasa, One bound to rule them all: from Adiabatic to Zeno.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Quantum 6, 737 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [9] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Jaynes and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Cummings, Comparison of quantum and semiclassical radiation theories with application to the beam maser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Proceedings of the IEEE 51, 89 (1963).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [10] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Larson and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Mavrogordatos, The Jaynes-Cummings model and its descendants modern research directions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' IoP Publishing (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [11] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Forn-Díaz, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Lisenfeld, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Marcos, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' García-Ripoll, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Solano, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Harmans, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Mooij, Observation of the Bloch-Siegert Shift in a Qubit-Oscillator System in the Ultrastrong Coupling Regime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review Letters 105, 237001 (2010).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [12] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Li, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Bamba, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Zhang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Fallahi, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Gardner, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Gao, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Lou, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Yoshioka, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Manfra, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Kono, Vacuum Bloch-Siegert shift in Landau polaritons with ultra-high cooperativity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Nature Photonics 12, 324 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [13] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Frisk, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Miranowicz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' De Liberato, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Savasta, and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Nori, Ultrastrong coupling between light and matter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Nature Review Physics 1, 19 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [14] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Walls, Higher order effects in the single atom field mode interaction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physics Letters A 42, 217 (1972).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [15] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='-P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Wang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Chen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Li, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Tsai, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Zhu, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' You, Photon-Dressed Bloch-Siegert Shift in an Ultrastrongly Coupled Circuit Quantum Electrodynamical System.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Physical Review Applied 13 ,054063 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [16] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Puri, Mathematical Methods of Quantum Optics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Springer (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [17] The reason we focus on short times here is a purely technical requirement from the proof (see Appendix).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Indeed, numerics shows that the errors are even larger for generic later times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [18] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Gottesman, Opportunities and Challenges in Fault-Tolerant Quantum Computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' arXiv:2210.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content='15844 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [19] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Grimsmo and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Puri, Quantum Error Correction with the Gottesman-Kitaev-Preskill Code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' PRX Quantum, 2, 020101 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [20] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Vlastakis, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Kirchmair, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Leghtas, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Nigg, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Frunzio, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Girvin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Mirrahimi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Devoret, and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Schoelkopf, Deterministically Encoding Quantum Information Using 100-Photon Schrödinger Cat States.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Science 342, 607 (2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' [21] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Reed and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Simon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Methods of mathematical physics 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Functional analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Academic Press (1980).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 16 [22] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Reed and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Simon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Methods of mathematical physics 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Fourier analysis, self- adjointness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' Academic Press (1975).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
+page_content=' 17' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/WtE0T4oBgHgl3EQfVwCs/content/2301.02269v1.pdf'}
diff --git a/X9AyT4oBgHgl3EQfvflc/vector_store/index.pkl b/X9AyT4oBgHgl3EQfvflc/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..b6427ee463b165550c9b08326f8ee570c57ab9d7
--- /dev/null
+++ b/X9AyT4oBgHgl3EQfvflc/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:970e331e6de64d1fb3cab513fffa7f5d797f862bee32e4b076e49ca39f2283eb
+size 253400
diff --git a/Y9E1T4oBgHgl3EQfcQTz/content/tmp_files/2301.03183v1.pdf.txt b/Y9E1T4oBgHgl3EQfcQTz/content/tmp_files/2301.03183v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..2b0262d5dd0867e88203d57caa1315e44aa20f1b
--- /dev/null
+++ b/Y9E1T4oBgHgl3EQfcQTz/content/tmp_files/2301.03183v1.pdf.txt
@@ -0,0 +1,2792 @@
+Minimax Weight Learning for Absorbing MDPs
+Fengying Li
+Yuqiang Li
+Xianyi Wu
+School of Statistics, KLATASDS-MOE, East China Normal University,
+Shanghai 200062, PR China
+Abstract
+Reinforcement learning policy evaluation problems are often modeled as finite or discounted/averaged infinite-
+horizon MDPs. In this paper, we study undiscounted off-policy policy evaluation for absorbing MDPs. Given the
+dataset consisting of the i.i.d episodes with a given truncation level, we propose a so-called MWLA algorithm to
+directly estimate the expected return via the importance ratio of the state-action occupancy measure. The Mean
+Square Error (MSE) bound for the MWLA method is investigated and the dependence of statistical errors on the
+data size and the truncation level are analyzed. With an episodic taxi environment, computational experiments
+illustrate the performance of the MWLA algorithm.
+Keywords:
+Absorbing MDP, Off-policy, Minimax weight learning, Policy evaluation, Occupancy measure
+1. Introduction
+Off-policy evaluation (OPE) in reinforcement learning means to estimate expected returns
+of target policies with data collected by behavior policies. It is very crucial in circumstances
+where deploying new strategies is expensive, risky or even dangerous, such as medicine (Murphy
+et al., 2001), education (Mandel et al., 2014), economics (Hirano et al., 2003), and recommender
+systems (Li et al., 2011). The existing off-policy evaluation is mostly based on importance sam-
+pling techniques, and thus suffers from high variance exponentially increasing in time horizon,
+known as “the curse of the horizon” (Jiang and Li, 2016; Li et al., 2015). Recently, a promis-
+ing idea of using Marginalized Importance Sampling (MIS) is proposed. For example, for an
+infinite-horizon discounted MDP, Liu et al. (2018) computes the importance weights regarding
+the state distribution by solving a minimax optimization problem, and proposes a method to
+estimate the expected return estimation and Uehara et al. (2020) proposes a Minimax Weight
+Learning (MWL) algorithm that directly estimates the ratio of the state-action distribution
+without reference to the information of behavior policy.
+In many reinforcement learning applications, the objectives of learning are to achieve some
+prescribed goals so that the processes are terminated at certain finite but random times. Exam-
+ples can be found in many robotic applications. In situations as such, it is then not appropriate
+to use finite-time and infinite-time to model the environments. MDPs with absorbing states,
+with the absorption to reflect the termination of processes, are thus suitable. Moreover, from
+Preprint submitted to ******
+January 10, 2023
+arXiv:2301.03183v1 [cs.LG] 9 Jan 2023
+
+a theoretical perspective, absorbing MDPs extend infinite-horizon discounted MDP processes
+(Altman, 1999).
+The theory of controllable absorbing MDPs has been widely studied and understood. A stop-
+ping time to reach the absorbing state is discussed in Chatterjee et al. (2008) and Iida and Mori
+(1996)), which depends on the state and action. The minimization of the expected undiscounted
+cost until the state enters the absorbing set is studied in, for example, pursuit problems, tran-
+sient programming and first pass problems (Eaton and Zadeh, 1962; Derman, 1970; Kushner,
+1971). Other research efforts include stochastic shortest path problem (Bertsekas and Tsitsiklis,
+1991), the control-to-exit time problem (Kesten and Spitzer, 1975; Borkar, 1988), and so on.
+There are also quite a few of problems that can be included in the category of absorbing MDPs,
+including board games (a game terminates once the winner is determined), trips through a maze,
+and dialog systems (a session terminates when the conversation is concluded) (Jiang, 2017), to
+name just a few.
+We aim to extend MWL method aforestated to the problems of OPE for absorbing MDPs.
+Since the absorbing MDPs have indefinite-horizon, and some of them are short, it is natural to
+assume that data consists of trajectories, i.e. each trajectory is treated as a single data point.
+However, in practice many trajectories are not recorded completely because they are too long,
+expensive or other reasons. In other words, the collected trajectory may be truncated. Naturally,
+an interesting and important problem in theory and practice is to quantify the possible errors
+resulting from the truncated data. Though researchers frequently need to handle the OPE in
+MIS experiments where many benchmark environments are indeed episodic and have random
+trajectory lengths, to our knowledge, there is few literature to discuss the absorbing MDPs and
+study their errors of OPE from the data truncation.
+In this paper, we propose a new but MWL-like algorithm (MWLA for short) with data set
+consisting of the truncated trajectories of an absorbing MDP, or in other words, the truncated
+episode data.
+We derive the estimation of the expected total return.
+An upper bound of
+the MSE of the MWLA algorithm is established in Theorems 4.1 and 4.2, showing that the
+MSE mainly consists of three parts: statistical error, approximation error, and optimization
+error. We explicitly evaluate the statistical errors by the truncation level and data size, and
+more importantly, we get the uniformly bound of MSE by optimizing the truncations when the
+truncation level is relatively large. Some numerical experiments in the episodic taxi environment
+to show the effectivity of our algorithm.
+The left is organized as follows: Section 2 introduces the model and specifies some basic
+settings. The MWLA algorithm and its theoretical guarantees with unknown behavior policy is
+presented in Section 3. A parallel version of MWLA, referred to MSWLA there, is also discussed
+for absorbing MDP with known behavior policies. In Section 4, under the assumption that the
+data consists of i.i.d. episodes, MSE bounds for the MWLA method are given in Theorems 4.1
+and 4.2. When the function classes are VC classes, compared with Theorem 9 in Uehara et al.
+(2020), it is found that our statistical error is related to the truncation length H. The relevant
+work is discussed in Section 5 in more detail so as to clarify their connection to and differences
+2
+
+between the current work. In Section 6, a computer experiment is reported under the episodic
+taxi environment, compared with on-policy, naive-average, and MSWLA methods, estimations
+of return and MSE are given for different episode numbers and different truncation lengths. All
+theoretical proofs and the pseudo-code of the algorithm are deferred to Appendix.
+2. Basic setting
+An MDP is a controllable rewarded Markov process that is represented by a standard tuple
+M = (S, A, R, P, µ) of a state space S, an action space A, a reward distribution R which maps a
+state-action pair (s, a) to a probability distribution R(s, a) with expectation R(s, a), a transfer
+probability function P : (s, a, s′) ∈ S×A×S → P(s′|s, a) ∈ [0, 1] and an initial state distribution
+µ. The space S × A is assumed enumerable and R bounded.
+A policy π := π(a|s) is a time-homogeneous mapping from S to the family of all distributions
+on A, under which the agent interacts with the environment consecutively: starting with an
+initial state s0 ∼ µ, at any integer time t ≥ 0, an action at ∼ π(·|st) is sampled, a scalar reward
+rt from the distribution R(st, at) is collected, and a next state st+1 ∼ P(·|st, at) is then assigned
+by the environment. The probability distribution generated by M under a policy π and an
+initial distribution µ is denoted by Pµ,π and uses Eµ,π for its expectation operation. When the
+initial state s0 = i, the probability distribution and expectation are indicated by Pi,π and Ei,π
+respectively, so that Pµ,π = �
+i∈S µ(i)Pi,π and Eµ,π = �
+i∈S µ(i)Ei,π. We also use P(s,a),π and
+E(s,a),π to indicate the probability and expectation generated by M when it is initialized by the
+state-action pair (s, a) and then follows policy π.
+An absorbing state, which is denoted by ξ, in S is such that r(ξ, a) = 0 and P(ξ|ξ, a) = 1
+for all a ∈ A. Without loss of generality, the absorbing state is assumed to be unique. For a
+trajectory, denote the terminal time by T .= min{n ≥ 1, sn = ξ}, where and in the paper .=
+reads “defined as”. An MDP is absorbing if Pi,π(T < ∞) = 1 for all states i ̸= ξ and all policies
+π. Denote S0 = S \ {ξ}. We need the following assumption on T.
+Assumption 2.1. sup(s,a)∈S0×A,π E(s,a),π(T) < ∞.
+The expected return of a rewarded Markov process depends only on the transition probability
+and the mean rewards rather than the distributions of the rewards at any state-action pair. For
+an MDP, the expected return under a policy π is
+Rπ .= Eµ,π
+�T−1
+�
+t=0
+rt
+�
+=
+∞
+�
+t=0
+Eµ,π [R(st, at)] .
+(1)
+Remark 2.1. With the expected return defined in (1) as the objective, the MDP models with
+absorbing states are more general than the infinite-horizon MDPs M = (S, A, r, P, µ, γ) under
+discounted expected return Rπ = Eµ,π[
+∞
+�
+t=0
+γtrt], where γ is a discount factor. This can be done by
+introducing a dummy absorbing state κ such that the probability of transferring to κ is 1−γ from
+any state action pair (s, a) because they have the same Bellman optimality equations (Altman,
+3
+
+1999). However, it worth noting that, unlike in the standard discounted infinite-horizon MDPs,
+in this artificially constructed MDP, as the survival probability of the system, the parameter γ
+is unknown and needs to be estimated.
+Especially, for an arbitrary but fixed state-action pair (s, a) ∈ S0×A, regarding the indication
+function 1(s,a) as a special mean reward function, i.e., collecting a unit reward at the state-action
+pair (s, a) and zero otherwise, we can introduce the so-called occupancy measure as
+dπ(s, a) .= Eµ,π
+� T−1
+�
+t=0
+1(s,a)(st, at)
+�
+=
+∞
+�
+t=0
+Pµ,π(st = s, at = a).
+(2)
+Note that dπ(s, a) < ∞ by Assumptions 2.1. Here, note that dπ(s, a) implicitly depends on the
+initial distribution µ. Conceptually, dπ can be retrieved from the commonly known Chapmann-
+Kolmogorov equation
+dπ(s, a) = µ(s)π(a|s) +
+∞
+�
+t=1
+Pµ,π(st = s, at = a)
+= µ(s)π(a|s) +
+∞
+�
+t=1
+�
+(s′,a′)∈S0×A
+Pµ,π(st−1 = s′, at−1 = a′)P(s|s′, a′)π(s|a)
+= µ(s)π(a|s) +
+�
+(s′,a′)∈S0×A
+P(s|s′, a′)π(s|a)dπ(s′, a′), for all (s, a) ∈ S0 × A.
+(3)
+Conversely, with the measure dπ(s, a), it follows that
+Rπ =
+�
+(s,a)∈S0×A
+R(s, a)dπ(s, a)
+is an integration of the mean reward function with respect to the occupancy measure.
+Remark 2.2. If we define a function Φπ(q) for any q ∈ B(S0 × A) by
+Φπ(q) .= Eµ,π[
+T−1
+�
+t=0
+q(st, at)],
+where B(S0×A) denotes the class of bounded functions on S0×A, with q(ξ, a) = 0 for all a ∈ A.
+We can readily see that
+Rπ = Φπ(R) and dπ(s, a) = Φπ(1(s,a)).
+Furthermore, a simple recursive argument shows that
+Φπ(q) =
+�
+(s,a)∈S0×A
+q(s, a)dπ(s, a)
+=
+�
+s∈S0
+µ(s)q(s, π) +
+�
+(s,a,s′)∈S0×A×S0
+P(s′|s, a)q(s′, π)dπ(s, a),
+4
+
+where for any function q, q(s, π) is the shorthand for �
+a∈A π(a|s)q(s, a). A direct result of the
+equality is
+�
+(s,a)∈S0×A
+�
+�q(s, a) −
+�
+s′∈S0
+P(s′|s, a)q(s′, π)
+�
+� dπ(s, a) =
+�
+s′∈S0
+µ(s′)q(s′, π).
+(4)
+Denote by dπ(s, a, s′) = dπ(s, a)P(s′|s, a). Then equation (4) can be rewritten as
+�
+(s,a,s′)∈S0×A×S0
+q(s′, π)dπ(s, a, s′) −
+�
+(s,a)∈S0×A
+q(s, a)dπ(s, a) + Es∼µ
+�
+q(s, π)
+�
+= 0.
+(5)
+3. Minimax Weight Learning for absorbing MDP (MWLA)
+In the rest of this paper, we aim to estimate the expected return of a target policy πe under a
+given initial distribution µ, by using a set of offline trajectories τ i, i = 1, . . . , m collected from a
+possibly different and unknown behavior policy πb and truncated at a level H a priori specified.
+More clearly, let
+Z = (s0, a0, s1, a1, · · · , sT−1, aT−1) and Zi = (si
+0, ai
+0, si
+1, ai
+1, · · · , si
+Ti−1, ai
+Ti−1), i = 1, 2 · · · , m
+(6)
+be a representative episodes of an absorbing MDP with probability distribution Pµ,πb and its
+i.i.d. copies, respectively. The dataset D contains m i.i.d. trajectories {τ i, i = 1, · · · , m}. Each
+τ i is a realization of Zi but only has the first li = Ti ∧ H consecutive sample transitions, i.e.
+τ i =
+�
+si
+0, ai
+0, ri
+0, si
+1, ai
+1, ri
+1, . . . , si
+li−1, ai
+li−1, ri
+li−1, si
+Ti∧H
+�
+, i = 1, 2, . . . , m,
+where Ti denotes the absorbing time of episode i. Note that Ti may not be observed. But
+Ti, i = 1, 2, · · · , m, are i.i.d with the distribution Pµ,πb(T = k) for each positive integer k.
+The goal of this paper is to estimate the expected return Rπe by using the data aforedescribed
+and analyze or control the errors caused by the truncation level H specified beforehand.
+For the infinite-horizon discounted MDPs, Uehara et al. (2020) proposes a Minimax Weight
+Learning (MWL) algorithm to handle the estimation of the expected discounted return, which
+is agnostic to the knowledge of πb. Their method uses a discriminator function class Q to learn
+the importance weight w (see equation (7) below) on state-action pairs. One of their important
+tools is the (normalized) discounted occupancy which can be approximated well considering the
+given discount factor γ and the suitable dataset (for example, the dataset consisting of i.i.d.
+tuples (s, a, r, s′)). However in our setting, the normalized occupancy is invalid since the reward
+is not discounted.
+In this section, we extend the MWL method to the absorbing MDPs. Our method is es-
+sentially based on another occupancy measure defined by the first equation in (2). The corre-
+spondingly developed algorithm, as what is indicated in the title of this section, is referred to
+as MWLA.
+5
+
+For any (s, a) ∈ S0 × A, let
+w πe
+πb (s, a) := dπe(s, a)
+dπb(s, a),
+almost everywhere
+dπb.
+(7)
+Observe that
+Rπe =
+�
+(s,a)∈S0×A
+R(s, a)dπe(s, a) =
+�
+(s,a)∈S0×A
+w πe
+πb (s, a)R(s, a)dπb(s, a) = Φπb(w πe
+πb R),
+if dπe(s, a) > 0 implies dπb(s, a) > 0. With dπb, R and w πe
+πb being estimated by ˆdπb, ˆR and ˆw πe
+πb ,
+respectively, a plug-in idea suggests that Rπe can be simply estimated by
+ˆΦπb( ˆw πe
+πb
+ˆR) .=
+�
+(s,a)∈S0×A
+ˆw πe
+πb (s, a) ˆR(s, a) ˆdπb(s, a),
+in which the key is to estimate w πe
+πb .
+We need the following technical assumption.
+Assumption 3.1. There exists a constant Cw > 0, such that sup(s,a)∈S0×A w πe
+πb (s, a) ≤ Cw.
+With equality (5), we formally introduce a loss function
+L(w, q) .=
+�
+(s,a,s′)∈S0×A×S0
+w(s, a)q(s′, πe)dπb(s, a, s′)
+−
+�
+(s,a)∈S0×A
+w(s, a)q(s, a)dπb(s, a) + Es∼µ
+�
+q(s, πe)
+�
+.
+(8)
+Obviously,
+B(S0 × A) × B(S0 × A) ⊂ D(L) .=
+�
+(w, q) : L(w, q) < ∞
+�
+.
+Recall that identity (5) simply states that
+L(w πe
+πb , q) = 0 for all q ∈ B(S0 × A).
+Conversely, by taking a family of particular functions {q(s, a) = 1(¯s,¯a)(s, a) : (¯s, ¯a) ∈ S0 × A},
+as what has been done in (3), we have the following result on the uniqueness of the solution to
+this system of equations.
+Theorem 3.1. Assume dπe is the unique solution to the systems of equations on q,
+q(s, a) = µ(s)πe(a|s) +
+�
+(s′,a′)∈S0×A
+P(s|s′, a′)πe(s|a)q(s′, a′), (s, a) ∈ S0 × A.
+(9)
+If Assumption 3.1 holds and dπe(s, a) > 0 implies dπb(s, a) > 0 for all (s, a) ∈ S0 × A, then
+6
+
+w = w πe
+πb is the unique bounded solution to the system of equations L(w, q) = 0 for each
+q ∈ l2(S0 × A) .=
+�
+�
+�g :
+�
+(s,a)∈S0×A
+g2(s, a) < ∞
+�
+�
+� .
+Theorem 3.1 simply states that
+w πe
+πb = argmin
+w
+max
+q∈l2(S0×A) L(w, q)2.
+(10)
+In order to retrieve the solution to (10), we introduce two function classes: W : S0 × A → R as
+working class of w πe
+πb , and Q : S0 × A → R to be treated as discriminators, then use
+w∗(s, a) .= argmin
+w∈W
+max
+q∈Q L(w, q)2
+to approximate w πe
+πb .
+The theorem below will be helpful in bounding the estimation error of occupancy measure
+ratio by means of the mini-max loss via the identification function class Q chosen properly.
+Theorem 3.2. The following bounds hold:
+(1)
+max
+q∈Q |L(w, q)| ≥∥ dπe − wdπb ∥∞,
+if
+{±Vs′,a′,πe : (s′, a′) ∈ S0 × A} ⊆ Q;
+(2)
+min
+w∈W max
+q∈Q |L(w, q)| ≥∥ w πe
+πb − w∗ ∥∞,
+if
+{±Vs′,a′,πe/dπb(s′, a′) : (s′, a′) ∈ S0 × A} ⊆ Q.
+where ∥ · ∥∞ denotes the supremum norm and
+Vs,a,πe(s′, a′) .=
+∞
+�
+t=0
+P(s,a),πe(st = s′, at = a′),
+∀ (s, a) ∈ S0 × A.
+(11)
+In order to construct estimators for dπb, R and w πe
+πb with the dataset of m i.i.d. trajectories
+τ i, for all (s, a, s′) ∈ S0 × A × S0, define
+ˆdi
+πb(s, a) .=
+li−1
+�
+t=0
+1(s,a)(si
+t, ai
+t),
+ˆdi
+πb(s, a, s′) .=
+li−1
+�
+t=0
+1(s,a,s′)(si
+t, ai
+t, si
+t+1)
+(12)
+and
+ˆri(s, a) .=
+�li−1
+t=0 ri
+t1(s,a)(si
+t, ai
+t)
+ˆdiπb(s, a)
+if ˆdi
+πb(s, a) > 0 and 0 otherwise
+to be the empirical occupancy measures and rewards from the single i-th episode, respectively.
+7
+
+Let
+ˆRm(s, a) = 1
+m
+m
+�
+i=1
+ˆri(s, a),
+ˆdm(s, a) = 1
+m
+m
+�
+i=1
+ˆdi
+πb(s, a).
+In addition, for any w, q ∈ B(S0 × A), introduce the empirical loss function
+ˆLm(w, q) = 1
+m
+m
+�
+i=1
+�
+(s,a,s′)∈S0×A×S0
+w(s, a)q(s′, πe) ˆdi
+πb(s, a, s′)
+− 1
+m
+m
+�
+i=1
+�
+(s,a)∈S0×A
+w(s, a)q(s, a) ˆdi
+πb(s, a) + Es∼µ
+�
+q(s, πe)
+�
+.
+(13)
+An estimator of w πe
+πb can then be defined as
+ˆwm(s, a) .= argmin
+w∈W
+max
+q∈Q
+ˆLm(w, q)2.
+Naturally, we estimate Rπe by
+ˆRπe
+.=
+�
+(s,a)∈S0×A
+ˆwm(s, a) ˆRm(s, a) ˆdm(s, a).
+This estimation procedure is referred to as an MWLA algorithm.
+Remark 3.1. Unlike the MWL algorithm in Uehara et al. (2020), the estimators defined here
+are not based on the (s, a, r, s′) tuple data but on truncated episodes. Clearly, the performance of
+the estimation depends on the sample size m and the truncation level H. We need to understand
+how the errors vary as m and H, which can help us to understand better the effects of truncating
+episodes and/or find a suitable level H to balance the errors caused by the truncation. We discuss
+this problem in next section, which also is one of the main contributions of this paper.
+Below is an example of the MWLA algorithm applied to the tabular MDP with an absorbing
+state.
+Example 3.1. Consider an absorbing tabular MDP with S0 = {0, 1, . . . , n−1}, A = {0, 1, . . . , h−
+1}. The function classes are
+W = Q =
+�
+ga(k, l) .= akh+l, k ∈ S0, l ∈ A : a = (a0, · · · , anh−1)⊤ ∈ Rnh�
+.
+For every 0 ≤ k ≤ n− 1, 0 ≤ l ≤ h− 1, denote by 1(k,l) the nh-dimensional column vector whose
+(kh + l)-th component is 1 and the others are 0, let 1(k,πe) = �h−1
+l=0 πe(l|k)1(k,l). Then, taking
+8
+
+w = gu and q = gv the empirical loss function is
+ˆLm(w, q) = 1
+m
+m
+�
+i=1
+�
+(k,l,v)∈S0×A×S0
+gu(k, l)gv(v, πe) ˆdi
+πb(k, l, v)
+− 1
+m
+m
+�
+i=1
+�
+(k,l)∈S0×A
+gv(k, l) ˆdi
+πb(k, l) +
+�
+k∈S0
+µ(k)gv(k, πe)
+= (u⊤ ˆA + b⊤)v,
+where
+ˆA = 1
+m
+m
+�
+i=1
+n−1
+�
+k=0
+h−1
+�
+l=0
+n−1
+�
+v=0
+1(k,l)1⊤
+(v,πe) ˆdi
+πb(k, l, v) − 1
+m
+m
+�
+i=1
+n−1
+�
+k=0
+h−1
+�
+l=0
+1(k,l)1⊤
+(k,l) ˆdi
+πb(k, l)
+= 1
+m
+m
+�
+i=1
+Ti∧H−1
+�
+t=0
+1(si
+t,ai
+t)
+� �
+a∈A
+πe(a|si
+t+1)1⊤
+(si
+t+1,a) − 1⊤
+(si
+t,ai
+t)
+�
+is an nh × nh matrix and b = �
+(s,a)∈S0×A µ(s)πe(a|s)1(s,a) is an nh-dimensional vector. There-
+fore,
+ˆL2
+m(w, q) = v⊤(ˆA⊤u + b)(ˆA⊤u + b)⊤v = ∥v∥
+2∥ˆA⊤u + b∥
+2
+2,
+where ∥ · ∥2 denotes the Euclidean norm on Rnh.
+The estimator ˆwm is ereadily seen as the follows.
+(1) When the matrix ˆA is invertible, ˆwm = gˆu, where
+ˆu = −(ˆA⊤)−1b,
+since ˆLm(gˆu, q) ≡ 0 for any q ∈ Q.
+(2) When the matrix ˆA is not invertible, but the equation
+b = −ˆA⊤u
+(14)
+is consistent, ˆwm = gˆu where ˆu = −(ˆA+)⊤b is the minimum norm least square solution
+of (14), because we have ˆLm(gˆu, q) ≡ 0 for any q ∈ Q (ˆA+ is the Moore-Penrose inverse
+of ˆA).
+(3) When equation (14) is inconsistent, ˆwm = gˆu where u = −(ˆA+)⊤b is the minimum norm
+least square solution of (14), because we also have ˆL2
+m(w, q) ≥ ˆL2(gˆu, q) for any q ∈ Q.
+Remark 3.2. If we define dπ(s) = Φπ(1{s}), then dπ(s, a) = dπ(s)π(a|s) and from (5), we have
+that
+�
+(s,a,s′)∈S0×A×S0
+q(s′, π)dπ(s)π(a|s)P(s′|s, a) −
+�
+s∈S0
+q(s, π)dπ(s) + Es∼µ
+�
+q(s, π)
+�
+= 0.
+9
+
+For a given target policy πe, simply denote q(s, πe) by q(s), so that the equation above can be
+rewritten as
+�
+(s,a,s′)∈S0×A×S0
+w πe
+πb (s)q(s′)πe(a|s)
+πb(a|s)dπb(s, a, s′) −
+�
+s∈S0
+w πe
+πb (s)q(s)dπb(s) + Es∼µ
+�
+q(s)
+�
+= 0,
+where w πe
+πb (s) = dπe(s)
+dπb(s).
+With this equation, when the behavior policy πb is known, we can construct a corresponding
+estimate of the value function based on the minimax optimization problem:
+min
+w∈Ws max
+q∈Qs
+�
+�
+(s,a,s′)∈S0×A×S0
+w πe
+πb (s)q(s′)πe(a|s)
+πb(a|s)dπb(s, a, s′)−
+�
+s∈S0
+w πe
+πb (s)q(s)dπb(s)+Es∼µ
+�
+q(s)
+��2.
+For convenience, we refer to the method as the MSWLA algorithm which is essentially an
+extension of the method discussed in Liu et al. (2018). By similar arguments in Example 3.1,
+in the tabular case where S0 = {0, 1, . . . , n − 1} and the function classes Ws and Qs are Rn,
+the empirical loss function for the MSWLA algorithm is ˆLm(w, q) = (u⊤ ˆA + b⊤)v for any
+w ∈ Ws, q ∈ Qs, where
+ˆA = 1
+m
+m
+�
+i=1
+Ti∧H−1
+�
+t=0
+1{si
+t}
+�πe(ai
+t|si
+t)
+πb(ai
+t|si
+t)1⊤
+{si
+t+1} − 1⊤
+{si
+t}
+�
+,
+b = �
+s∈S0 µ(s)1{s}, and for any s ∈ S0, 1{s} is the n-dimensional column vector whose s-th
+entry is 1 and other elements are 0.
+4.
+MSE of estimated return
+In this section, we discuss how the error of ˆRπe varies with the episodic size m and the
+truncation level H in terms of the mean squared error (MSE).
+To state our main results, we need the so-called state-action function Qπe(s, a) which is
+defined by
+Qπe(s, a) .= E(s,a),πe(
+T−1
+�
+t=0
+r(st, at)).
+Denote by Hm the unique solution to the equation
+2mx2 + 2 ln m ln x − ln m = 0,
+x > 0.
+The following theorems state the results.
+Theorem 4.1. Suppose that
+10
+
+1) there exists constants K0, K1 such that for any w ∈ W, q ∈ Q,
+∥w∥2 :=
+�
+�
+(s,a)∈S0×A
+w2(s, a)
+�1/2
+≤ K0, ∥q∥∞ :=
+sup
+(s,a)∈S0×A
+|q(s, a)| ≤ K1;
+2) W and Q have finite pseudo-dimensions DW and DQ, respectively;
+3) Assumptions 2.1 and 3.1 hold;
+4) Qπe ∈ Q;
+5) There exists λ0 > 0 such that Eµ,πb(eλ0T ) = M0 < ∞.
+Then we have the following:
+1) When M0e−λ0H > Hm, there exists a constant C independent of H, m, such that
+E
+�
+( ˆRm − Rπe)2�
+≤ C
+�
+e−2λ0H + H2 ln m
+m
+�
++ 8 min
+w∈W max
+q∈Q L(w, q)2.
+2) When M0e−λ0H ≤ Hm, there exists a constant C independent of H, m, such that
+E
+�
+( ˆRm − Rπe)2�
+≤ C ln3 m
+m
++ 8 min
+w∈W max
+q∈Q L(w, q)2.
+Especially, if w πe
+πb ∈ W and M0e−λ0H ≤ Hm, then
+E
+�
+( ˆRm − Rπe)2�
+≤ C ln3 m
+m
+.
+Theorem 4.2. Suppose the assumptions in Theorem 4.1 hold and m ≥ e but Qπe ̸∈ Q.
+(1) When M0e−λ0H > Hm, there exists a constant C independent of H, m, such that
+E
+�
+( ˆRm − Rπe)2�
+≤ C
+�
+e−2λ0H + H2 ln m
+m
+�
++ 16 min
+w∈W max
+q∈Q L(w, q)2 + 4 max
+w∈W min
+q∈Q L2(w, Qπe − q).
+(2) When M0e−λ0H ≤ Hm, there exists a constant C independent of H, m, such that
+E
+�
+( ˆRm − Rπe)2�
+≤ C ln3 m
+m
++ 16 min
+w∈W max
+q∈Q L(w, q)2 + 4 max
+w∈W min
+q∈Q L2(w, Qπe − q).
+Obviously, the additional term max
+w∈W min
+q∈Q L(w, Qπe − q)2 becomes 0 when Qπe in the closure of Q
+under the metric || · ||∞.
+Theorem 4.1 and Theorem 4.2 provide upper bounds of MSE, which are related to the
+truncation level H and the number m of the episodes. On one hand, for a small truncation level H
+(i.e., M0e−λ0H > Hm), the estimate errors include four terms: the pure truncation term e−2λ0H,
+the mixing term H2 ln m/m generated by the randomness of sampling, the approximation error
+11
+
+min
+w∈W max
+q∈Q L2(w, q), and the optimization error max
+w∈W min
+q∈Q L2(w, Qπe −q), in which the first two are
+caused by the randomness of statistics and the other two are from the approximation of the two
+function classes W and Q to wπ/πb and Qπ. On the other hand, when the truncation level H
+is relatively large (i.e., M0e−λ0H ≤ Hm), the pure truncation term e−2λ0H and the mixing term
+H2 ln m/m can be dominated by C ln3 m/m which is dependent of the truncation level H. This
+observation also approves in some sense that our MWLA algorithm can eliminate the curse of
+the horizon.
+In the following are more remarks on the results.
+Remark 4.1. Consider the case Qπe ∈ Q. For the infinite horizon MDP with m i.i.d. tuples
+(si, ai, s′
+i, ri), the error bound of the MWL method consists of a statistical error ln m
+m +R2
+m(W, Q)
+and an approximation error min
+w∈W max
+q∈Q L(w, q)2, where Rm(W, Q) is the Rademacher complexity
+of the function class
+��
+s, a, s′�
+�→ |w(s, a)(q(s
+′, π) − q(s, a))| : w ∈ W, q ∈ Q
+�
+,
+as given in Theorem 9 of Uehara et al. (2020). Let DW,DQ be the VC-subgraph dimension (i.e.
+pseudo-dimension) of W, Q, respectively. Because Rm(W, Q) = O(
+�
+max (DW,DQ)
+m
+) (Corollary
+1 of Uehara et al., 2021), the statistical error is dominated by ln m
+m . In our method with m
+i.i.d. episodes, the MSE bound also includes an approximation error min
+w∈W max
+q∈Q L(w, q)2 and a
+statistical error. When M0e−λ0H ≤ Hm, the statistical error is bounded by ln3 m
+m
+which includes
+an extra factor ln2 m in form.
+Remark 4.2. For m > e, one has
+�
+ln m/(2m) < Hm < ln m
+�
+e/m (Lemma B.8 in Appendix).
+Therefore, when M0e−λ0H > Hm, it follows that H ≤ ln M0+ln 2/2+ln m/2 and H2 ln m
+m
+≤ C ln3 m
+m
+for some constant C. Whatever H is, the bounds in Theorem 4.1 are both less than
+C(e−2λ0H + ln3 m/m) + 8 min
+w∈W max
+q∈Q L(w, q)2.
+Remark 4.3. In the tabular setting, if we take W = {w : ∥w∥2 ≤ K0} and Q = {q : ∥q∥∞ <
+K1}, where K0 is a constant larger than Cw in Assumption 3.1, then all assumptions in Theorem
+4.1 hold. Hence,
+E
+�
+( ˆRm − Rπe)2�
+≤ C
+�
+e−2λ0H + ln3 m
+m
+�
+.
+5.
+Connections to related work
+The research of the off-policy evaluation, which has been mainly modeled as an infinite and
+fixed finite horizon MDP, can be divided into two categories according to whether the behavior
+policy is known.
+When the behavior policy is known, Importance Sampling (IS) is a method for reweighting
+rewards according to its likelihood ratio of πe over πb. Since the ratio is computed by a cumulative
+12
+
+product of the importance weight over action πe(a|s)
+πb(a|s) at each time step (Precup, 2000), the IS
+method suffers from a variance that is exponentially increasing in time horizon.
+To reduce
+that extremely high variance, a series of off-policy estimation methods are proposed based on
+IS. For example, the Weighted Importance Sampling method, Stepwise Importance Sampling
+method, and Doubly Robust(DR) method can reduce the variance to certain degree (Cassel et al.,
+1976; Robins et al., 1994; Robins and Rotnitzky, 1995; Bang and Robins, 2005). However, the
+exponential variance of IS-based methods cannot be significantly improved when the MDP has
+a high stochasticity (Jiang and Li, 2016).
+The marginalized importance sampling method proves a promising improvement over IS
+by successfully avoiding the trouble of exponential variance. For example, for a finite-horizon
+inhomogeneous MDP, compared to weighting the whole trajectory, using a ratio wt(s) πe(a|s)
+πb(a|s) with
+wt(s) = dπe,t(s)
+dπb,t(s) to reweight the rewards r (Xie et al., 2019) gives rise to a lower variance. In an
+infinite horizon setting, based on a discounted stationary distribution, Liu et al. (2018) proposes
+to use the ratio w πe
+πb (s) · πe(a|s)
+πb(a|s) with w πe
+πb (s) = dπe,γ(s)
+dπb,γ(s). The ratio w πe
+πb (s) is then estimated by a
+minimax procedure with two function approximators: one to model a weight function w πe
+πb (s),
+and the other to model V πb, as a discriminator class for distribution learning.
+A further effort is made on the case of unknown behavior policies by Hanna et al. (2019),
+showing that the importance sampling method with history-based behavior policy estimation
+has lower asymptotic variance than the one with a known behavior strategy. The fitted Q-
+iteration,which uses dynamic programming techniques to fit Qπe directly from the data, can
+overcome the curse of dimensionality, with a const of assuming that the function class contains
+Qπe and is closed under the bellman update Bπe, so as to avoid a high bias, see Ernst et al. (2005)
+and Le et al. (2019). Uehara et al. (2020) proposes the MWL algorithm by means of estimating
+marginalized importance weight w πe
+πb (s, a) = dπe,γ(s,a)
+dπb(s,a) . A Dualdice algorithm is further proposed
+to estimate the discounted stationary distribution ratios (Nachum et al., 2019a; Nachum et al.,
+2019b; Nachum and Dai, 2020) where the loss function can be considered as a derivative of the
+loss function in Uehara et al. (2020). Jiang and Huang (Jiang and Huang, 2020) briefly discuss
+about how to derive the equivalent of MWL/MQL as well as their value interval for finite-horizon
+MDPs and average-reward MDPs.
+In reinforcement learning, there are only a few efforts made on absorbing MDP, though many
+benchmark environments are indeed episodic and have random horizons, such as board games
+(a game terminates once the winner is determined), trips through a maze, and dialog systems (a
+session terminates when the conversation is concluded) (Jiang, 2017). Researchers often handle
+absorbing MDPs as a special case of finite-horizon MDPs by padding all trajectories (with
+random lengths) to the same length with absorbing states. Another way to handle it practically
+is to use the infinite-horizon setup (with a sufficiently large discount factor γ), and whenever a
+trajectory terminates, we imagine it continuous to infinity with absorbing states. However, when
+the random horizons are not bounded and the random episodes are not observed completely,
+especially, accompanied by the non-discounted rewards, new issues will arise. For example, how
+do the unobserved trajectories affect the results? As our results show, this problem, which is by
+13
+
+no means trivial, is essentially neglected when we simply apply the two ways mentioned above.
+In this paper we employ a direct way to handle the OPE for absorbing MDPs. Our contri-
+butions are two folds.
+1. We introduce the MWLA algorithm. The MWLA algorithm proposed in this paper is a
+variant of the MWL to fit the random horizon and truncated episodic data modeled by absorbing
+MDPs. The difference is that the MWL algorithm uses basically (s, a, r, s′)-tuple data and the
+MWLA algorithm uses episodic data.
+2. We explicitly evaluate the errors caused by the truncation level and data size, and more
+importantly, we get the uniform bound of MSE by optimizing the truncations when the trunca-
+tion level is relatively large.
+6. Experiments
+6.1. Setting
+The environment Taxi (Dietterich, 2000) is a two-dimensional grid world that simulates a
+taxi moving along a grid. The experiment is conducted with a 5 × 5 grid as shown in Figure
+1. The four corners are marked as R(ed), B(lue), G(reen), and Y(ellow). Initially, the taxi
+randomly chooses a corner to wait for a passenger, who appears or disappears with probability
+at each of the four corners, and that passenger wishes to be transported to one of the four
+corners (also chosen randomly). The taxi must pick up the passenger and drops him off at a
+destination. An episode ends once a passenger is successfully dropped off at his destination.
+Figure 1: Taxi Grid
+The elements of the model are as follows.
+There are a total of 2000 states (25 × 24 × 5), including 25 taxi locations, 24 passenger
+appearance status, and 5 taxi status (empty or one of 4 destinations). There are four navigation
+actions that move the taxi one square North, South, East, or West, respectively. Each action
+is deterministic. Besides, we can only take 3 and 2 actions when the taxi is at the boundary
+and the corner, respectively. When a passenger is picked up in Taxi, it receives a reward of -1
+at every time step, but a reward of 0 is collected if she is successfully dropped off at the right
+place. A reward of -2 at each time step is collected if the taxi is empty.
+As in Liu et al. (2018), we first run a Q-learning with 400,000 iterations to produce a
+policy and then run 60,000 iterations to produce another policy.
+The two policies are then
+regularized such that the probability of crossing any boundary of the grid is 0 and the sums of
+the probabilities of the actions are 1. The first policy is treated as the target policy πe and the
+second as the policy π+. The behaviors are πb = απe + (1 − α)π+ with α ∈ {0.2, 0.4}.
+14
+
+4
+R
+D
+3
+D
+2
+1
+0
+Y
+B
+0
+2
+3
+4Taxi environment above differs from the ones analyzed by others in the following aspects:
+(1) Initial Location. While the taxi starts in a randomly-chosen square in Liu et al. (2018)
+and Uehara et al. (2020)), it starts in a randomly-chosen corner here.
+(2) Actions. In previous research (Dietterich, 2000; Liu et al., 2018; Uehara et al., 2020),
+there are other two actions: a Pickup action and a Putdown action.
+In the experiment, the true expected return of the target policy is approximated by a set
+of 2 × 106 on-policy Monte-Carlo trajectories, truncated at H = 500 to assure that most trials
+stop at the absorbing state.
+6.2. Results
+Provided here are the experiment results for MWLA and MSWLA, together with an on-
+policy algorithm and a naive averaging baseline algorithm. The on-policy algorithm estimates
+the expected return by the direct average over a set of trajectories generated by the target policy
+itself, and the naive average baseline algorithm does it by a direct average over another set of
+trajectories generated by the behavior policy, all truncated at H. We also show the results of
+MWL applied to the data.
+The first experiment is on the MSEs of the four methods applied to m = 15000, 20000, 30000,
+40000 and 50000 trajectories with policy mixing ratio α = 0.2 and varying truncation level
+H = {20, 50, 100, 150, 200}. A total of 100 duplicates of every parameter setting are generated
+with different random seeds. Two strategies of visualization are employed. In the upper panel of
+Figure 2, every graph corresponds to a fixed episode size. It appears that the MSE of MWLA and
+MSWLA decreases at the beginning and then varies slowly when the truncation level increases,
+MWLA is better than MSWLA to a moderate degree, and both are significantly lower than
+the on-policy and naive averaging baseline algorithms. The lower panel of Figure 2 provides a
+different perspective for the results, in which every graph corresponds to a fixed H. Figure 3
+visualizes the results in the same setting as in Figure 2 but now we use the policy ratio α = 0.4
+and it does show similar results.
+The estimated returns are depicted in Figures 4. The estimates by both MSWLA and MWLA
+approach the true value as the number of trajectories increases. MSWLA is slightly better than
+MWLA in mean but worse in fluctuation, so that gives rise to a slightly higher MSE as shown
+in Figure 2 and 3.
+7. Conclusions
+In this paper, we study the off-policy policy evaluation problem in reinforcement learning for
+absorbing MDPs. We propose the MWLA algorithm, which extends the MWL (Uehara et al.,
+2020) algorithm to the episode data with truncation. Assuming the collected data are i.i.d.
+episodes with a given truncation level, we obtain the estimator of the expected total return Rπe
+and theoretically study their MSE under certain conditions. We evaluate the statistical errors by
+the data size and the truncation level. If the behavior policy is known, we also briefly introduce
+the MSWLA algorithm. At last, in the episodic taxi environment, our method is compared to
+15
+
+Agend: The horizontal axis indicates the truncation level H and the vertical the logarithm of the MSE.
+Agend: The horizontal axis indicates the number of trajectories and the vertical the MSE, both are scaled in
+logarithm.
+Figure 2: MSE of the four algorithms: α = 0.2
+Agend: The horizontal axis indicates the truncation level H and the vertical the logarithm of the MSE.
+Agend: The horizontal axis indicates the number of trajectories and the vertical the MSE, both are scaled in
+logarithm.
+Figure 3: MSE of the four algorithms: α = 0.4
+16
+
+On-Policy
+Naive Average
+MSWLA
+MWLA
+3
+3
+log10MSE
+2 -
+2
+0
+1
+0
+50
+100150
+200
+50
+100150
+200
+50
+100150
+200
+50
+100150
+200
+50
+100150
+200
+(a)m=15000
+(b) m=20000
+(c) m=30000
+(d) m=40000
+(e) m=50000On-Policy
+Naive Average
+MSWLA
+MWLA
+2
+0
+0.
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+(a) logTrajecories, H=20
+(b) logTrajecories, H=50
+(c) logTrajecories, H=100
+(d) logTrajecories, H=150
+(e) logTrajecories, H=200On-Policy
+¥一
+Naive Average
+MSWLA
+MWLA
+3
+3
+log10MSE
+2 -
+2
+1
+0
+1
+-1
+1
+50
+100
+150
+50
+100150
+200
+50
+100150
+200
+50
+100
+150
+200
+50
+100150
+200
+(a)m=15000
+(b) m=20000
+(c) m=30000
+(d) m=40000
+(e) m=50000On-Policy
+Naive Average
+MSWLA
+MWLA
+2
+0
+0.
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+(a) logTrajecories, H=20
+(b) logTrajecories, H=50
+(c) logTrajecories, H=100
+(d) logTrajecories, H=150
+(e) logTrajecories, H=200α = 0.2
+α = 0.4
+Figure 4: Estimated expected return.
+the on-policy, naive-average, and MSWLA algorithm, the estimation of return and mean squared
+errors are given for different number of trajectories and different truncation lengths. It is shown
+that the MWLA method has the lower MSE as the number of episodes and truncation length
+increase, significantly improving the accuracy of policy evaluation.
+References
+Eitan Altman. Constrained Markov decision processes: stochastic modeling. Routledge, 1999.
+Martin Anthony, Peter L Bartlett, Peter L Bartlett, et al. Neural network learning: Theoretical foundations,
+volume 9. cambridge university press Cambridge, 1999.
+Heejung Bang and James M Robins. Doubly robust estimation in missing data and causal inference models.
+Biometrics, 61(4):962–973, 2005.
+Dimitri P Bertsekas and John N Tsitsiklis. An analysis of stochastic shortest path problems. Mathematics
+of Operations Research, 16(3):580–595, 1991.
+Vivek S Borkar. A convex analytic approach to markov decision processes. Probability Theory and Related
+Fields, 78(4):583–602, 1988.
+Claes M Cassel, Carl E S¨arndal, and Jan H Wretman. Some results on generalized difference estimation and
+generalized regression estimation for finite populations. Biometrika, 63(3):615–620, 1976.
+Debasish Chatterjee, Eugenio Cinquemani, Giorgos Chaloulos, and John Lygeros. Stochastic control up to
+a hitting time: optimality and rolling-horizon implementation. arXiv preprint arXiv:0806.3008, 2008.
+Cyrus Derman. Finite state markovian decision processes. Technical report, 1970.
+Thomas G Dietterich.
+Hierarchical reinforcement learning with the maxq value function decomposition.
+Journal of artificial intelligence research, 13:227–303, 2000.
+Jo H Eaton and LA Zadeh. Optimal pursuit strategies in discrete-state probabilistic systems. 1962.
+17
+
+MSWLA
+MWLA
+True
+77.25
+77.20
+77.0
+I Return
+77.30
+77.25
+77.1
+-77.30
+xpected
+77.30
+77.2
+77.35
+77.35
+77.35
+77.3
+77.40
+77.40
+77.40
+77.4
+77.45
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+(a) logTrajectores, H=50
+(b) logTrajectores, H=100
+(c) logTrajectores, H=150
+(d) logTrajectores, H=200MSWLA
+MWLA
+True
+76.9
+-77.25
+77.275
+77.0
+77.30
+IReturn
+-77.300
+77.1
+77.30
+-77.325
+Expected
+77.35
+77.2
+-77.350
+77.35
+77.3
+77.40
+-77.375
+77.40
+77.400
+77.4
+77.45
+77.425
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+4.2
+4.4
+4.6
+(a) logTrajectores, H=50
+(b) logTrajectores, H=100
+(c) logTrajectores, H=150
+(d) logTrajectores, H=200Damien Ernst, Pierre Geurts, and Louis Wehenkel. Tree-based batch mode reinforcement learning. Journal
+of Machine Learning Research, 6:503–556, 2005.
+Josiah Hanna, Scott Niekum, and Peter Stone. Importance sampling policy evaluation with an estimated
+behavior policy. In International Conference on Machine Learning, pages 2605–2613. PMLR, 2019.
+David Haussler. Sphere packing numbers for subsets of the boolean n-cube with bounded vapnik-chervonenkis
+dimension. Journal of Combinatorial Theory, Series A, 69(2):217–232, 1995.
+Keisuke Hirano, Guido W Imbens, and Geert Ridder. Efficient estimation of average treatment effects using
+the estimated propensity score. Econometrica, 71(4):1161–1189, 2003.
+Tetsuo Iida and Masao Mori. Markov decision processes with random horizon. Journal of the Operations
+Research Society of Japan, 39(4):592–603, 1996.
+Nan Jiang. A theory of model selection in reinforcement learning. PhD thesis, 2017.
+Nan Jiang and Jiawei Huang. Minimax value interval for off-policy evaluation and policy optimization. In
+34th Conference on Neural Information Processing Systems, 2020.
+Nan Jiang and Lihong Li. Doubly robust off-policy value evaluation for reinforcement learning. In Interna-
+tional Conference on Machine Learning, pages 652–661. PMLR, 2016.
+Harry Kesten and Frank Spitzer. Controlled markov chains. The Annals of Probability, pages 32–40, 1975.
+HJ Kushner. Introduction to stochastic control. holt, rinehart and wilson. New York, 1971.
+Hoang Le, Cameron Voloshin, and Yisong Yue. Batch policy learning under constraints. In International
+Conference on Machine Learning, pages 3703–3712. PMLR, 2019.
+Lihong Li, Wei Chu, John Langford, and Xuanhui Wang. Unbiased offline evaluation of contextual-bandit-
+based news article recommendation algorithms. In Proceedings of the fourth ACM international conference
+on Web search and data mining, pages 297–306, 2011.
+Lihong Li, R´emi Munos, and Csaba Szepesv´ari. Toward minimax off-policy value estimation. In Artificial
+Intelligence and Statistics, pages 608–616. PMLR, 2015.
+Qiang Liu, Lihong Li, Ziyang Tang, and Dengyong Zhou. Breaking the curse of horizon: Infinite-horizon
+off-policy estimation. Advances in Neural Information Processing Systems, 31, 2018.
+Travis Mandel, Yun-En Liu, Sergey Levine, Emma Brunskill, and Zoran Popovic. Offline policy evaluation
+across representations with applications to educational games. In AAMAS, volume 1077, 2014.
+Susan A Murphy, Mark J van der Laan, James M Robins, and Conduct Problems Prevention Research
+Group. Marginal mean models for dynamic regimes. Journal of the American Statistical Association, 96
+(456):1410–1423, 2001.
+Ofir Nachum and Bo Dai.
+Reinforcement learning via fenchel-rockafellar duality.
+arXiv preprint
+arXiv:2001.01866, 2020.
+Ofir Nachum, Yinlam Chow, Bo Dai, and Lihong Li. Dualdice: Behavior-agnostic estimation of discounted
+stationary distribution corrections. Advances in Neural Information Processing Systems, 32, 2019a.
+Ofir Nachum, Bo Dai, Ilya Kostrikov, Yinlam Chow, Lihong Li, and Dale Schuurmans. Algaedice: Policy
+gradient from arbitrary experience. arXiv preprint arXiv:1912.02074, 2019b.
+David Pollard. Convergence of stochastic processes. Springer Science & Business Media, 2012.
+Doina Precup.
+Eligibility traces for off-policy policy evaluation.
+Computer Science Department Faculty
+Publication Series, page 80, 2000.
+James M Robins and Andrea Rotnitzky. Semiparametric efficiency in multivariate regression models with
+missing data. Journal of the American Statistical Association, 90(429):122–129, 1995.
+James M Robins, Andrea Rotnitzky, and Lue Ping Zhao. Estimation of regression coefficients when some
+regressors are not always observed. Journal of the American statistical Association, 89(427):846–866, 1994.
+18
+
+Masatoshi Uehara, Jiawei Huang, and Nan Jiang. Minimax weight and q-function learning for off-policy
+evaluation. In International Conference on Machine Learning, pages 9659–9668. PMLR, 2020.
+Masatoshi Uehara, Masaaki Imaizumi, Nan Jiang, Nathan Kallus, Wen Sun, and Tengyang Xie.
+Finite
+sample analysis of minimax offline reinforcement learning: Completeness, fast rates and first-order efficiency.
+arXiv preprint arXiv:2102.02981, 2021.
+Tengyang Xie, Yifei Ma, and Yu-Xiang Wang.
+Towards optimal off-policy evaluation for reinforcement
+learning with marginalized importance sampling. Advances in Neural Information Processing Systems, 32,
+2019.
+19
+
+Appendix
+A. Proof of Theorems in Section 3
+The proof of Theorem 3.1 is read as follows.
+Proof of Theorem 3.1.
+Proof. It suffices to prove the uniqueness. For any w ∈ B(S0 × A), let
+Fw(s, a) = µ(s)πe(a|s) +
+�
+(¯s,¯a)∈S0×A
+w(¯s, ¯a)P(s|¯s, ¯a)πe(a|s)dπb(¯s, ¯a) − w(s, a)dπb(s, a),
+for any (s, a) ∈ S0 × A. Assumption 2.1 implies that Fw ∈ l2(S0 × A). Moreover, it is easy to
+check that L(w, q) = ⟨Fw, q⟩, where ⟨·, ·⟩ is the inner product in the Hilbert space l2(S0 × A).
+From the assumption L(w, q) = 0, ∀q ∈ l2(X × A), we know Fw ≡ 0. The assumption that dπe
+is the unique solution of (9) implies wdπb = dπe, i.e. w = w πe
+πb .
+□
+In order to prove Theorem 3.2, for any function q on S0 × A, we define a one-step forward
+operator under policy π, writing Lπ, by
+(Lπq)(s, a) .=
+�
+(s′,a′)∈S0×A
+q(s′, a′)π(a′|s′)P(s′|s, a) = Eπ [q(st+1, at+1)|st = s, at = a] , (A.1)
+Then, the so-called state-action function
+Qπ(s, a) .= E(s,a),π(
+T−1
+�
+t=0
+r(st, at))
+= R(s, a) +
+�
+s′∈S0
+P(s′|s, a)Es′,π
+�T−1
+�
+t=0
+rt
+�
+= R(s, a) +
+�
+(s′,a′)∈S0×A
+P(s′|s, a)π(a′|s′)Qπ(s′, a′)
+= R(s, a) + (LπQπ)(s, a).
+(A.2)
+and the loss function
+L(w, q) =
+�
+s,a∈S0×A
+w(s, a) [Lπeq(s, a) − q(s, a)] dπb(s, a) + Es∼µ
+�
+q(s, πe)
+�
+.
+(A.3)
+By Theorem 3.1, we can further change the form of L(w, q) as
+L(w, q) =
+�
+(s,a)∈S0×A
+�
+wπ/πb(s, a) − w(s, a)
+�
+[q(s, a) − (Lπeq)(s, a)] dπb(s, a).
+(A.4)
+Now we give the proof of Theorem 3.2.
+20
+
+Proof. Note the definition of Vs′,a′,πe in (11). Equation (A.2) states that
+Vs′,a′,πe(s, a) − LπeVs′,a′,πe(s, a) = 1(s′,a′)(s, a).
+Put it into (A.4), it follows that, for every (s′, a′) ∈ S0 × A,
+L(w, Vs′,a′,πe) =
+�
+(s,a)∈S0×A
+�
+w πe
+πb (s, a) − w(s, a)
+�
+1(s′,a′)(s, a)dπb(s, a)
+= dπe(s′, a′) − w(s′, a′)dπb(s′, a′).
+When {±Vs′,a′,πe : (s′, a′) ∈ S0 × A} ⊆ Q, the derised result follows from taking maximum of L
+over q ∈ Q.
+For the second assertion, consider particularly the function ±Vs′,a′,πe/dπb(s′, a′). If
+{±Vs′,a′,πe/dπb(s′, a′) : (s′, a′) ∈ S0 × A} ⊆ Q,
+then maxq∈Q |L(w, q)| ≥ ||w πe
+πb − w∗||∞. Therefore,
+min
+w∈W max
+q∈Q |L(w, q)| = max
+q∈Q |L(w∗, q)| ≥ ||w πe
+πb − w∗||∞.
+This completes the proof.
+□
+B. Proof of Theorems in Section 4
+To prove Theorem 4.1 and Theorem 4.2, we give the definition of pseudo-dimension.
+Definition B.1. (Anthony et al., 1999) Let F be a set of functions mapping from a domain
+X to R and suppose that S = {x1, x2, . . . , xm} ⊆ X. Then S is pseudo-shattered by F if there
+are real number r1, r2, . . . , rm such that for each b ∈ {0, 1}m there is a function fb in F with
+sgn (fb (xi) − ri) = bi for 1 ≤ i ≤ m. We say that r = (r1, r2, . . . , rm) witnesses the shattering.
+Definition B.2. (Anthony et al., 1999) Suppose that F is a set of functions from a domain X
+to R. Then F has pseudo-dimension d if d is the maximum cardinality of a subset S of X that is
+pseudo-shattered by F. If no such maximum exists, we say that F has infinite pseudo-dimension.
+The pseudo-dimension of F is denoted Pdim(F).
+Let us define the empirical ℓ1-covering numbers of a function class F with respect to a data
+set {Z1, . . . , Zn}. In particular, we define the set F ({Zi}) ⊂ Rn as follows:
+F ({Zi}) = {(f (Z1) , . . . , f (Zn)) | f ∈ F} .
+We then define the empirical ℓ1-covering number as N (ϵ; F ({Zi})) as the smallest number N of
+balls of radius ϵ required to cover F ({Zi}). In this definition, distances are measured in terms
+21
+
+of the empirical ℓ1-norm
+∥f − g∥F({Zi}) := 1
+n
+n
+�
+i=1
+|f (Zi) − g (Zi)| .
+We quote the following two results for easy reference later on.
+Lemma B.1. (Pollard, 2012) Let F be a permissible class of Z → [−M, M] functions and
+{Zi}n
+i=1 are i.i.d. samples from some distribution P. Then, for any given ϵ > 0,
+P
+�
+sup
+f∈F
+�� 1
+n
+n
+�
+i=1
+f
+�
+Zi
+�
+− Ef(Z)
+�� > ϵ
+�
+≤ 8E
+�
+N (ϵ; F ({Zi}))
+�
+exp
+� −nϵ2
+512M2
+�
+.
+(B.1)
+The covering number can then be bounded in terms of the function class’s pseudo-dimension:
+Lemma B.2. (Corollary 3 (Haussler, 1995)) For any set Z, any points {zi}n
+i=1 ⊆ Z, any class
+F of functions on Z taking values in [0, M] with pseudo-dimension DF < ∞, and any ϵ > 0,
+N (ϵ; F ({zi})) ≤ e (DF + 1)
+�2eM
+ϵ
+�DF
+.
+(B.2)
+Recall that in Theorem 4.1, we address that the functional classes W and Q have finite
+pseudo-dimensions DW and DQ, Qπe ∈ Q, and there exists constants K0, K1 such that for
+any w ∈ W, ∥w∥2 ≤ K0, ∥q∥∞ ≤ K1. In addition, we assume that there exists λ0 > 0 such
+that Eµ,πb(eλ0T ) = M0 < ∞. Besides them, we also remind here that our data consist of i.i.d
+samples from M = (S, A, R, P, µ) under the policy πb. The proof of Theorem 4.1 proceeds in
+the following 3 parts: decomposition, evaluation, and optimization.
+B.1 Decomposition
+Note that the estimator of Rπe based on the truncated sample is
+ˆRm = 1
+m
+m
+�
+i=1
+�
+(s,a)∈S0×A
+ˆwm(s, a)ˆri(s, a) ˆdi
+πb(s, a).
+It is easy to see that
+ˆRm − Rπe = I1 + I2 + I3,
+(B.3)
+where
+I1 =
+�
+(s,a)∈S0×A
+[ ˆwm(s, a) − w πe
+πb (s, a)]R(s, a)dπb(s, a),
+I2 = 1
+m
+m
+�
+i=1
+�
+(s,a)∈S0×A
+ˆwm(s, a)R(s, a)
+�
+ˆdi
+πb(s, a) − dπb(s, a)
+�
+,
+22
+
+and
+I3 = 1
+m
+m
+�
+i=1
+�
+(s,a)∈S0×A
+ˆwm(s, a)
+�
+ˆri(s, a) − R(s, a)
+�
+ˆdi
+πb(s, a).
+By (A.2) and (A.4),
+I1 =
+�
+(s,a)∈S0×A
+[ ˆwm(s, a) − w πe
+πb (s, a)][Qπe(s, a) − (LπeQπe)(s, a)]dπb(s, a) = −L( ˆwm, Qπe).
+Because Qπe ∈ Q, it follows that
+I2
+1 =
+�
+L( ˆwm, Qπe) − ˆLm( ˆwm, Qπe) + ˆLm( ˆwm, Qπe)
+�2
+≤ 2
+�
+L( ˆwm, Qπe) − ˆLm( ˆwm, Qπe)
+�2 + 2 max
+q∈Q
+ˆL2
+m( ˆwm, q),
+where ˆLm is defined in (13) and we have used the fact that
+ˆL2
+m( ˆwm, Qπe) ≤ max
+q∈Q
+ˆL2
+m( ˆwm, q) = min
+w∈W max
+q∈Q
+ˆL2
+m(w, q).
+Using the fact
+max
+q∈Q
+ˆL2
+m( ˆwm, q) ≤ 2 max
+q∈Q L2( ˆwm, q) + 2 max
+q∈Q
+�
+ˆLm( ˆwm, q) − L( ˆwm, q)
+�2
+,
+we further have
+I2
+1 ≤ 6I2
+11 + 4 min
+w∈W max
+q∈Q L(w, q)2,
+(B.4)
+where
+I2
+11 =
+sup
+w∈W,q∈Q
+|L(w, q) − ˆLm(w, q)|2.
+Substituting (12) into the expression of ˆLm(w, q) in (13), it follows that
+ˆLm(w, q) − Eµ(q(s, πe)) = 1
+m
+m
+�
+i=1
+li−1
+�
+t=0
+w(si
+t, ai
+t)(q(si
+t+1, πe) − q(si
+t, ai
+t)),
+where q(ξ, πe) = 0 is assumed. Similarly,
+L(w, q) − Eµ(q(s, πe)) = Eµ,πb
+� T−1
+�
+t=0
+w(st, at)(q(st+1, πe) − q(st, at))
+�
+.
+Define
+˜L(w, q) = Eµ,πb
+� T∧Hβ−1
+�
+t=0
+w(st, at)(q(st+1, πe) − q(st, at))
+�
+23
+
+and
+˜Lm(w, q) = 1
+m
+m
+�
+i=1
+Ti∧Hβ−1
+�
+t=0
+w(si
+t, ai
+t)
+�
+q(si
+t+1, πe) − q(si
+t, ai
+t)
+�
+,
+where Hβ is a constant specified later. Then
+|ˆLm(w, q) − L(w, q)| = |ˆLm(w, q) − Eµ(q(s, πe)) − (L(w, q) − Eµ(q(s, πe)))|
+≤ | ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)| + |˜Lm(w, q) − ˜L(w, q)|
++|˜L(w, q) − (L(w, q) − Eµ(q(s, πe)))|.
+(B.5)
+B.2 Evaluation
+For any β ≥ M0e−λ0H, let Hβ = min{k : M0e−λ0k ≤ β} = ⌈ln(M0/β)/λ0⌉, where ⌈x⌉ is the
+minimum integer no less than x. Obviously, Hβ ≤ H and M0e−λ0Hβ ≤ β.
+(B.2.1) To get the upper bound of E(I2
+11)
+An upper bound of E(I2
+11) is derived via a sequence of auxiliary results.
+Lemma B.3. With the constants K0 and K1, for any β ≥ M0e−λ0H,
+E(
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)|2) ≤ 4K2
+0K2
+1
+λ2
+0
+�
+β2 + 2β
+m
+�
+.
+Proof. For any w ∈ W and q ∈ Q,
+|ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)| =
+��� 1
+m
+m
+�
+i=1
+li−1
+�
+t=Ti∧Hβ
+w(si
+t, ai
+t)
+�
+q(si
+t+1, πe) − q(si
+t, ai
+t)
+�
+1{Ti>Hβ}
+���.
+Recall the notation li = Ti ∧ H. Since w and q are bounded by K0 and K1, respectively,
+|ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)| ≤ 2K0K1
+m
+m
+�
+i=1
+(Ti − Hβ)1{Ti>Hβ}.
+Therefore,
+E(
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)|2)
+≤ 4K2
+0K2
+1
+m2
+�
+�
+m
+�
+i,j=1,i̸=j
+E[(Ti − Hβ)(Tj − Hβ)1{Ti,Tj>Hβ}] +
+m
+�
+i=1
+E[(Ti − Hβ)21{Ti>Hβ}]
+�
+�
+= 4K2
+0K2
+1
+m2
+�
+m(m − 1)E2[(T1 − Hβ)1{T1>Hβ}] + mE[(T1 − Hβ)21{T1>Hβ}]
+�
+,
+(B.6)
+where the equality follows from the i.i.d. property of trajectories. By further the inequality
+24
+
+x ∨ x2/2 ≤ ex for every x > 0,
+E[(T1 − Hβ)1{T1>Hβ}] ≤ 1
+λ0
+E
+�
+eλ0(T1−Hβ)1{T1>Hβ}
+�
+≤ M0e−λ0Hβ
+λ0
+≤ β
+λ0
+,
+(B.7)
+and
+E[(Ti − Hβ)21{Ti>Hβ}] ≤ 2
+λ2
+0
+E(eλ0(T1−Hβ)1{T1>Hβ}) ≤ 2M0e−λ0Hβ
+λ2
+0
+≤ 2β
+λ2
+0
+,
+(B.8)
+substituting (B.7) and (B.8) into (B.6) leads to the desired result.
+□
+Lemma B.4. There exists a constant C1 independent of m, H such that for m ≥ 2,
+E(
+sup
+w∈W,q∈Q
+|˜Lm(w, q) − ˜L(w, q)|2) ≤ C1H2
+β
+ln m
+m .
+Proof. For a representative trajectory Z of the form (6), denote by
+˜wq(Z) =
+T∧Hβ−1
+�
+t=0
+w(st, at)(q(st+1, πe) − q(st, at))
+so that
+˜Lm(w, q) − ˜L(w, q) = 1
+m
+m
+�
+i=1
+˜wq(Zi) − E( ˜wq(Z)).
+It is easy to see that | ˜wq| ≤ 2K0K1Hβ.
+Denote by H = { ˜wq(Z) : w ∈ W, q ∈ Q}. The distance
+in H can be bounded by
+1
+m
+m
+�
+i=1
+���
+Ti∧Hβ−1
+�
+t=0
+w1(si
+t, ai
+t)(q1(si
+t+1, πe) − q1(si
+t, ai
+t))
+−
+Ti∧Hβ−1
+�
+t=0
+w2(si
+t, ai
+t)(q2(si
+t+1, πe) − q2(si
+t, ai
+t))
+���
+≤ 2K1Hβ||w1 − w2||∞ + 2K0Hβ∥q1 − q2∥∞.
+As a result,
+N1(2Hβ(K1ϵ1 + K0ϵ2), H, {Zi}m
+i=1) ≤ N1(ϵ1, W, {Zi}m
+i=1)N1(ϵ2, Q, {Zi}m
+i=1).
+Then a direct application of Lemma B.2 shows that
+N1(2Hβ(K1ϵ1 + K0ϵ2), H, {Zi}m
+i=1) ≤ e2(DW + 1)(DQ + 1)
+�4eK0
+ϵ1
+�DW �4eK1
+ϵ2
+�DQ
+.
+25
+
+Taking ϵ1 =
+ϵ
+32K1Hβ and ϵ2 =
+ϵ
+32K0Hβ , we have
+N1
+� ϵ
+8, H, {Zi}m
+i=1
+�
+≤
+M
+ϵDW+DQ ,
+where M = e2(DW + 1)(DQ + 1)(128eK0K1Hβ)DW+DQ. By Pollard’s tail inequality (Lemma
+B.1),
+P
+�
+sup
+w∈W,q∈Q
+�����
+1
+m
+m
+�
+i=1
+hw,q(Zi) − Ehw,q(Z)
+����� > ϵ
+�
+≤
+8M
+ϵDW+DQ exp
+�
+−mϵ2
+2048(K0K1Hβ)2
+�
+.
+(B.9)
+For any m > 1, let x0 = (32K0K1Hβ)2(DW+DQ) ln m
+m
+. Then, by (B.9), we have that
+E(
+sup
+w∈W,q∈Q
+|˜Lm(w, q) − ˜L(w, q)|2)
+≤
+� ∞
+0
+P(
+sup
+w∈W,q∈Q
+|˜Lm(w, q) − ˜L(w, q)| ≥ √x)dx.
+≤
+� x0
+0
+dx +
+� ∞
+x0
+8M
+x(DW+DQ)/2 exp
+�
+−mx
+2048(K0K1Hβ)2
+�
+dx
+≤ x0 +
+8M
+x(DW+DQ)/2
+0
+� ∞
+x0
+exp
+�
+−mx
+2048(K0K1Hβ)2
+�
+dx
+= (32K0K1Hβ)2
+m
+�
+(DW + DQ) ln m +
+8M
+[(32K0K1Hβ)2 ln m(DW + DQ)](DW+DQ)/2
+�
+,
+which implies the desired result.
+□
+Lemma B.5. Assume that W and Q have finite pseudo-dimensions DW and DQ.
+If there
+exists constants K0, K1 such that for any w ∈ W, ∥w∥∞ ≤ K0, ∥q∥∞ ≤ K1, then there exists a
+constant C3 independent of m, H such that for m ≥ 2,
+E(
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − L(w, q)|2) ≤ C3
+�
+β2 + (1 − ln β + ln2 β)ln m
+m
++ β
+m
+�
+.
+Proof. By (B.5),
+E[
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − L(w, q)|2] ≤ 3E(
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)|2)
++3E(
+sup
+w∈W,q∈Q
+|˜L(w, q) − (L(w, q) − Eµ(q(s, πe)))|2)
++3
+sup
+w∈W,q∈Q
+|˜Lm(w, q) − ˜L(w, q)|2.
+(B.10)
+26
+
+Note that for any w ∈ W,
+|˜L(w, q) − (L(w, q) − Eµ(q(s, πe)))| = E
+� T−1
+�
+T∧Hβ
+w(st, at)(q(st+1, πe) − q(st, at))
+�
+≤ 2K0K1E((T − Hβ)1{T≥Hβ})
+≤ 2K0K1E
+�
+eλ0(T−Hβ)
+λ0
+1{T≥Hβ}
+�
+≤ 2K0K1
+λ0
+β.
+(B.11)
+Substituting (B.11) into (B.10) and applying Lemma B.3 and Lemma B.4, we get that
+E(
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − L(w, q)|2) ≤ 3M2
+3 µ2 + 3
+�
+M2
+3 β2 + 2M2
+3
+m β
+�
++ 3C1H2
+β
+ln m
+m ,
+where M3 = 2K0K1/λ0. Since Hβ ≤ ln(M0/β)
+λ0
++ 1, it follows from (B.10) that
+E(
+sup
+w∈W,q∈Q
+|ˆLm(w, q) − L(w, q)|2)
+≤ 3C1
+λ2
+0
+�
+(ln β)2 − 2(ln M0 + λ0) ln β + (ln M0 + λ0)2� ln m
+m
++ 6M2
+3 β2 + 6M2
+3
+m β
+≤ 3C1
+λ2
+0
+�
+(ln β)2 − 2C2 ln β + C2
+2
+� ln m
+m
++ 6M2
+3 β2 + 6M2
+3 β
+m
+,
+where C2 = ln M0 + λ0. Let
+C3 = max
+�3C1
+λ2
+0
+, 3C1
+λ2
+0
+C2, 3C1
+λ2
+0
+C2
+2, 6M2
+3
+�
+,
+we can readily get the desired result.
+□
+(B.2.2) To get the upper bound of E(I2
+2) and E(I2
+3)
+Lemma B.6. Assume that there exists constant K0 such that for any w ∈ W, ∥w∥2 ≤ K0.
+Then
+E(I2
+2) ≤ 2R2
+maxK2
+0M0
+λ2
+0
+( 1
+m + M0e−2λ0H),
+where Rmax .= sup |R(s, a)|.
+Proof. Define a truncated occupancy measure ˜dπb(s, a) = Eµ,πb(�T∧H−1
+t=0
+1(s,a)(st, at)). Then
+I2 = 1
+m
+m
+�
+i=1
+�
+(s,a)∈S0×A
+ˆwm(s, a)R(s, a)[ ˆdi
+πb(s, a) − ˜dπb(s, a)]
++
+�
+(s,a)∈S0×A
+ˆwm(s, a)R(s, a)[ ˜dπb(s, a) − dπb(s, a)]
+27
+
+and hence
+I2
+2 ≤ 2
+�
+�
+(s,a)∈S0×A
+ˆwm(s, a)R(s, a)
+� 1
+m
+m
+�
+i=1
+ˆdi
+πb(s, a) − ˜dπb(s, a)
+��2
++2
+�
+�
+(s,a)∈S0×A
+ˆwm(s, a)R(s, a)
+�
+˜dπb(s, a) − dπb(s, a)
+��2
+≤ 2
+�
+(s,a)∈S0×A
+ˆwm(s, a)2R(s, a)2
+�
+(s,a)∈S0×A
+� 1
+m
+m
+�
+i=1
+ˆdi
+πb(s, a) − ˜dπb(s, a)
+�2
++2
+�
+�
+(s,a)∈S0×A
+ˆwm(s, a)R(s, a)Eµ,πb(
+T−1
+�
+t=T∧H
+1(s.a)(st, at))
+�2
+,
+where the last inequality follows from H¨older’s inequality. Invoking the bounds from ˆwm and R,
+we get that
+E(I2
+2) ≤ 2R2
+maxK2
+0
+�
+�
+(s,a)∈S0×A
+E
+� 1
+m
+m
+�
+i=1
+ˆdi
+πb(s, a) − ˜dπb(s, a)
+�2
++
+�
+Eµ,πb
+�
+�
+(s,a)∈S0×A
+T−1
+�
+t=T∧H
+1(s,a)(st, at)
+��2�
+≤ 2R2
+maxK2
+0
+� 1
+m
+�
+(s,a)∈S0×A
+Var( ˆdi
+πb(s, a))) +
+�
+Eµ,πb((T − H)1{T>H})
+�2�
+,
+where the first inequality is due to ˆwm ∈ W and the second inequality follows from the fact that
+di
+πb(s, a), 1 ≤ i ≤ m, are i.i.d and have expectation ˜dπb(s, a). Observing that
+Var( ˆdi
+πb(s, a))) = Varµ,πb(
+T∧H−1
+�
+t=0
+1(s,a)(st, at)) ≤ Eµ,πb
+�� T∧H−1
+�
+t=0
+1(s,a)(st, at)
+�2�
+≤ Eµ,πb
+�
+(T ∧ H)
+T∧H−1
+�
+t=0
+1(s,a)(st, at)
+�
+,
+we obtain that
+�
+(s,a)∈S0×A
+Var( ˆdi
+πb(s, a))) ≤ Eµ,πb
+�
+(T ∧ H)
+�
+(s,a)∈S0×A
+T∧H−1
+�
+t=0
+1(s,a)(st, at)
+�
+≤ Eµ,πb
+�
+T 2�
+.
+Since Eµ,πb
+�
+T 2�
+≤ 2M0
+λ2
+0
+and Eµ,πb((T − H)1{T>H}) ≤ M0
+λ0 e−λ0H, we arrive at the conclusion. □
+To estimate E(I2
+3), we have the following result.
+Lemma B.7. Assume that there exists constant K0 such that for any w ∈ W, ∥w∥2 ≤ K0.
+Then
+E(I2
+3) ≤ R2
+maxK2
+0M0
+λ0m
+.
+28
+
+Proof. Noting that
+I3 =
+�
+(s,a)∈S0×A
+ˆwm(s, a)
+� 1
+m
+m
+�
+i=1
+(ˆri(s, a) − R(s, a)) ˆdi
+πb(s, a)
+�
+,
+applying the H¨older inequality, we have that
+I2
+3 ≤
+�
+(s,a)∈S0×A
+ˆw2
+m(s, a)
+�
+(s,a)∈S0×A
+� 1
+m
+m
+�
+i=1
+(ˆri(s, a) − R(s, a)) ˆdi
+πb(s, a)
+�2
+≤ K2
+0
+�
+(s,a)∈S0×A
+� 1
+m
+m
+�
+i=1
+(ˆri(s, a) − R(s, a)) ˆdi
+πb(s, a)
+�2
+.
+Therefore,
+E[I2
+3] ≤ K2
+0E
+�
+�
+�
+(s,a)∈S0×A
+E
+�� 1
+m
+m
+�
+i=1
+(ˆri(s, a) − R(s, a)) ˆdi
+πb(s, a)
+�2��� ˆdi
+πb(s, a), i = 1, · · · , m
+��
+� .
+Because ˆdi
+πb(s, a), i = 1, · · · , m are i.i.d and ri(s, a) follows distribution R(s, a) and is indepen-
+dent of ˆdi
+πb(s, a), i = 1, · · · , m, it follows that
+E
+�� 1
+m
+m
+�
+i=1
+(ˆri(s, a) − R(s, a)) ˆdi
+πb(s, a)
+�2��� ˆdi
+πb(s, a), i = 1, · · · , m
+�
+= 1
+m2
+m
+�
+i=1
+( ˆdi
+πb(s, a))2Var
+�
+ˆri(s, a)
+��� ˆdi
+πb(s, a)
+�
+= 1
+m2
+m
+�
+i=1
+( ˆdi
+πb(s, a))2Var
+��li−1
+t=0 ri
+t1(s,a)(si
+t, ai
+t)
+ˆdiπb(s, a)
+��� ˆdi
+πb(s, a)
+�
+.
+When ˆdi
+πb(s, a) is given, �li−1
+t=0 ri
+t1(s,a)(si
+t, ai
+t) is the sum of ˆdi
+πb(s, a) random variables who are
+independent with the same distribution R(s, a). Hence
+Var
+�
+ˆri(s, a)
+��� ˆdi
+πb(s, a)
+�
+= VarR(s,a)(r)
+ˆdiπb(s, a)
+≤
+R2
+max
+ˆdiπb(s, a)
+.
+Therefore
+E(I2
+3) ≤ R2
+maxK2
+0
+m2
+E
+�
+�
+(s,a)∈S0×A
+m
+�
+i=1
+ˆdi
+πb(s, a)
+�
+= R2
+maxK2
+0
+m2
+E
+� m
+�
+i=1
+�
+(s,a)∈S0×A
+ˆdi
+πb(s, a)
+�
+= R2
+maxK2
+0
+m2
+E
+� m
+�
+i=1
+(li − 1)
+�
+≤ R2
+maxK2
+0
+m
+Eµ,πb(T),
+which implies the desired result, since Eµ,πb(T) ≤ M0
+λ0 .
+□
+29
+
+B.3 Optimization
+Based on the above discussion, we get that for any truncation level H and β such that
+M0e−λ0H ≤ β,
+E(| ˆRm − Rπe|2) ≤ 2E(I2
+1) + 4E(I2
+2) + 4E(I2
+3)
+≤ 8 min
+w∈W max
+q∈Q L(w, q)2 + 12E(I2
+11) + 4E(I2
+2) + 4E(I2
+3)
+≤ 8 min
+w∈W max
+q∈Q L(w, q)2 + 12C3
+�
+β2 + (1 − ln β + (ln β)2)ln m
+m
++ β
+m
+�
++4R2
+maxK2
+0M0
+mλ0
++ 8K2
+0R2
+maxM0
+λ2
+0
+( 1
+m + M0e−2λ0H),
+where the first inequality follows from (B.3) and the simple inequality (a + b)2 ≤ 2a2 + 2b2, the
+second inequality follows from (B.4) and the last inequality is due to Lemma B.5-Lemma B.7.
+Letting
+C4 = max{24C3 + 4R2
+maxK2
+0M0
+λ0
++ 8K2
+0R2
+maxM0
+λ2
+0
+, 8K2
+0R2
+maxM2
+0
+λ2
+0
+},
+we have that, for any truncation level H and β ≥ M0e−λ0H,
+E(| ˆRm − Rπe|2) ≤ 8 min
+w∈W max
+q∈Q L(w, q)2 + C4
+�
+G(β, m) + e−2λ0H�
+,
+(B.12)
+where
+G(x, m) := x2 + (1 − ln x + ln2 x)ln m
+m ,
+for x ∈ (0, +∞), m ≥ 1. Note that
+G′
+x(x, m) = 2x + ln m(2 ln x − 1)
+mx
+,
+G′′
+xx(x, m) = 2 + ln m(3 − 2 ln x)
+mx2
+> 0,
+which combining the fact G′
+x(1, m) > 1 and
+lim
+x→0+ G′
+x(x, m) = −∞ implies that there exists a
+unique Hm ∈ (0, 1) such that
+2mH2
+m + 2 ln m ln Hm − ln m = 0,
+which implies G′
+x(Hm, m) = 0 and hence for all x ∈ (0, +∞),
+G(Hm, m) ≤ G(x, m).
+Moreover, G(x, m) is decreasing for x ∈ (0, Hm) while increasing for x ∈ (Hm, +∞).
+Lemma B.8. For any m ≥ e,
+�
+ln m/(2m) ≤ Hm ≤
+�
+e ln2 m/m and there exists constants
+0 < k1 < k2 such that
+k1
+ln3 m
+m
+≤ G(Hm, m) ≤ k2
+ln3 m
+m
+.
+30
+
+Proof. From G′
+µ(Hm, m) = 0, we know that Hm is a solution of
+Hm(x) := 2mx2 + 2 ln m ln x − ln m = 0.
+Note that Hm(x) is increasing on (0, 1] and
+Hm
+��
+ln m
+2m
+�
+= 2 ln m ln
+��
+ln m
+2m
+�
+< 0,
+Hm
+�
+�
+�
+e ln2 m
+m
+�
+� = (2e − 1) ln2 m + 2 ln m ln ln m > 0.
+We have that
+�
+ln m/(2m) ≤ Hm ≤
+�
+e ln2 m/m.
+To prove the second assertion, we note that
+G(Hm, m) ≤ G
+�
+�
+�
+ln3 m
+m
+, m
+�
+� ≤ ln3 m
+m
++ (1 + 1
+2 ln m + 1
+4 ln2 m)ln m
+m
+≤ 3ln3 m
+m
+.
+On the other hand, when x <
+�
+ln3 m/m,
+G(x, m) ≥ (1 − ln x + ln2 x)ln m
+m
+≥ 1
+4
+�
+ln ln3 m
+m
+�2 ln m
+m
+≥ ln3 m
+4m
+�
+1 − 3ln ln m
+ln m
+�2
+≥ (e3 − 9)2
+4e6
+ln3 m
+m
+,
+for m ≥ ee3. Noting that ln3 m/m > 0 for all m > e, we can find a constant k′ such that
+G(x, m) ≥ k′ ln3 m
+m
+,
+for all x ∈ (0,
+�
+ln3 m/m) and m > e. Moreover, when x ≥
+�
+ln3 m/m,
+G(x, m) > x2 ≥ ln3 m
+m
+.
+Consequently, there exists a constant k1 such that G(Hm, m) ≥ k1 ln3 m/m.
+□
+Now we are at the position to finish the proof of Theorem 4.1.
+Proof. From (B.12) and Lemma B.8, it follows that if Hm ≥ M0e−λ0H, there is a constant C
+31
+
+independent of m, H such that
+E(| ˆRm − Rπe|2) ≤ 8
+min
+w∈W,q∈Q L2(w, q) + C4
+�
+min
+β≥M0e−λ0H G(β, m) + e−2λ0H�
+= 8
+min
+w∈W,q∈Q L2(w, q) + C4(G(Hm, m) + e−2λ0H)
+≤ 8
+min
+w∈W,q∈Q L2(w, q) + C ln3 m
+m
+,
+since Hm ≥ M0e−λ0H implies that
+e−2λ0H ≤ H2
+m
+M0
+≤
+e
+M0
+ln2 m
+m
+.
+When Hm < M0e−λ0H,
+E(| ˆRm − Rπe|2) ≤ 8
+min
+w∈W,q∈Q L2(w, q) + C4(G(M0e−λ0H, m) + e−2λ0H)
+= 8
+min
+w∈W,q∈Q L2(w, q) + C4
+�
+(1 + M2
+0 )e−2λ0H + (1 − ln M0 + λ0H + (ln M0 − λ0H)2)ln m
+m
+�
+≤ 8
+min
+w∈W,q∈Q L2(w, q) + C
+�
+e−2λ0H + H2 ln m
+m
+�
+,
+for some constant C independent of m, H.
+□
+From Theorem 4.1, it is easy to get Therorem 4.2. We briefly state the proof as follows.
+Proof of Theorem 4.2. When Qπe ̸∈ Q, (B.4) does not hold but can be adjusted as follows.
+Since
+I2
+1 = L2( ˆwm, Qπe) = (L( ˆwm, Qπe − q) + L( ˆwm, q))2 ≤ 2(L2( ˆwm, q) + L2( ˆwm, Qπe − q)),
+for any q ∈ Q, we have that,
+I2
+1 ≤ 2 max
+q∈Q L2( ˆwm, q) + 2 min
+q∈Q L2( ˆwm, Qπe − q).
+≤ 2 max
+q∈Q (L( ˆwm, q) − ˆLm( ˆwm, q) + ˆLm( ˆwm, q))2 + 2 max
+w∈W min
+q∈Q L2(w, Qπe − q)
+≤ 4 max
+q∈Q (L( ˆwm, q) − ˆLm( ˆwm, q))2 + 4 max
+q∈Q
+ˆL2
+m(w, q) + 2 max
+w∈W min
+q∈Q L2(w, Qπe − q),
+32
+
+for any w ∈ W. Consequently,
+I2
+1 ≤ 4
+max
+w∈W,q∈Q(L( ˆwm, q) − ˆLm( ˆwm, q))2 + 8 max
+q∈Q (ˆLm(w, q) − L(w, q))2
++8 max
+q∈Q L2(w, q) + 2 max
+w∈W min
+q∈Q L2(w, Qπe − q)
+≤ 4
+max
+w∈W,q∈Q(L( ˆwm, q) − ˆLm( ˆwm, q))2 + 8
+max
+w∈W,q∈Q(ˆLm(w, q) − L(w, q))2
++8 min
+w∈W max
+q∈Q L2(w, q) + 2 max
+w∈W min
+q∈Q L2(w, Qπe − q)
+= 12
+max
+w∈W,q∈Q(L( ˆwm, q) − ˆLm( ˆwm, q))2 + 8 min
+w∈W max
+q∈Q L2(w, q) + 2 max
+w∈W min
+q∈Q L2(w, Qπe − q),
+and therefore
+E(I2
+1) ≤ 12E(
+max
+w∈W,q∈Q(L( ˆwm, q) − ˆLm( ˆwm, q))2)
++8 min
+w∈W max
+q∈Q L2(w, q) + 2 max
+w∈W min
+q∈Q L2(w, Qπe − q).
+Using this inequality to replace (B.4) and then repeating the discussion in Theorem 4.1, we can
+readily get the desired result.
+□
+C. Algorithm Supplement
+In the algorithm, we give the following notation:
+G =
+�
+���
+0 . . . 0
+... ...
+0
+0
+�
+���
+nh×nh
+,
+Frequency = [0, . . . , 0]⊤
+nh,
+auxi = [0, . . . , 0]⊤
+nh,
+ˆµ = [0, . . . , 0]⊤
+nh, X represents absorbing state
+set, Y = {h × i + j, i ∈ X, j ∈ A}.
+The Algorithm 1 summarizes the pseudo-codes of our MWLA algorithm applied to the taxi
+environment in Section 6.
+The principle of the algorithm is explained in Example 3.1. Here, for the convenience of
+computations, we set
+w(s, a) =
+1⊤
+{s,a}u
+ˆdπb(s, a)
+and q(s, a) = 1⊤
+{s,a}v,
+∀s ∈ S0, a ∈ A.
+We also introduce a regularization factor λ > 0 which helps us find the unique solution of
+the constrained quadratic programming problem arg min
+u≥0 ∥ ( ˆG + λI)⊤u + ˆb ∥2.
+When λ is
+sufficient small, the solution is an approximation of −( ˆG+)⊤b where ˆG+ is the Moore-Penrose
+pseudo-inverse of ˆG. In our experiments, λ is set to be 0.001.
+33
+
+Algorithm 1 Tabular case
+Input: Off-policy data D = {si
+0, ai
+0, ri
+0, si
+1, · · · , si
+Ti∧H−1, ai
+Ti∧H−1, ri
+Ti∧H−1, si
+Ti∧H}m
+i=1 from the
+behavior policy πb; a target policy π for which we want to estimate the expected return.
+1:
+Estimate the initial state distribution ˆµ(s) =
+1
+m
+m
+�
+i=1
+1{si
+0=s}, where 1{·} is an indicative
+function.
+2: for episode in D do
+3:
+for s,a,s
+′,r in episode do
+4:
+cur = h × s + a,
+5:
+G[cur, h × s
+′ : h × (s
+′ + 1)]+ = π[s
+′, :],
+6:
+G[cur, cur]− = 1.0,
+7:
+Frequency[cur]+ = 1.
+end for
+8: auxi = �
+s,a ˆµ(s)π(a|s)1(s, a).
+9: tvalid = where(Frequency > 0) indicates the index of an element whose value is greater
+than 0.
+10: ˆdπb = delete(Frequency, Y, 0), delete the row corresponding to the absorbing state from
+Frequency.
+11: tvalid1 = where( ˆdπb > 0).
+12: ˆdπb = ˆdπb/m.
+13: ˆG = delete(G, Y, 0), delete the row corresponding to the absorbing state from G.
+14: ˆG = delete( ˆG, Y, 1), delete the column corresponding to the absorbing state from ˆG.
+15: ˆG[:, tvalid1] = ˆG[:, tvalid1]/(m × ˆdπb[tvalid1]).
+16: ˆb = delete(auxi, Y, 0).
+17: Compute ˆu = arg min
+u≥0 ∥ ( ˆG + λI)⊤u + ˆb ∥2, where
+ˆG = 1
+m
+m
+�
+i=1
+Ti∧H−1
+�
+t=0
+1(si
+t,ai
+t)
+� �
+a∈A π(a|si
+t+1)1⊤
+(si
+t+1,a) − 1⊤
+(si
+t,ai
+t)
+�
+ˆdπb(si
+t, ai
+t)
+.
+ˆb =
+�
+(s,a)∈S0×A
+ˆµ(s)π(a|s)1(s,a),
+λ is a regularization factor, I denotes an identity matrix.
+18: Parameterize w(tvalid) =
+ˆu(tvalid1)
+ˆdπb(tvalid1).
+19: for episode in D do
+20:
+for s, a, s
+′, r in episode do
+21:
+cur = h × s + a,
+22:
+ˆRm = 1
+m
+m
+�
+i=1
+Ti∧H−1
+�
+t=0
+w(cur) × r.
+end for
+Output: ˆRm.
+34
+
diff --git a/Y9E1T4oBgHgl3EQfcQTz/content/tmp_files/load_file.txt b/Y9E1T4oBgHgl3EQfcQTz/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..934824debd639ea57ef915a1b05dff9a013d9c3a
--- /dev/null
+++ b/Y9E1T4oBgHgl3EQfcQTz/content/tmp_files/load_file.txt
@@ -0,0 +1,1006 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf,len=1005
+page_content='Minimax Weight Learning for Absorbing MDPs Fengying Li Yuqiang Li Xianyi Wu School of Statistics, KLATASDS-MOE, East China Normal University, Shanghai 200062, PR China Abstract Reinforcement learning policy evaluation problems are often modeled as finite or discounted/averaged infinite- horizon MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In this paper, we study undiscounted off-policy policy evaluation for absorbing MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Given the dataset consisting of the i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d episodes with a given truncation level, we propose a so-called MWLA algorithm to directly estimate the expected return via the importance ratio of the state-action occupancy measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The Mean Square Error (MSE) bound for the MWLA method is investigated and the dependence of statistical errors on the data size and the truncation level are analyzed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' With an episodic taxi environment, computational experiments illustrate the performance of the MWLA algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Keywords: Absorbing MDP, Off-policy, Minimax weight learning, Policy evaluation, Occupancy measure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Introduction Off-policy evaluation (OPE) in reinforcement learning means to estimate expected returns of target policies with data collected by behavior policies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' It is very crucial in circumstances where deploying new strategies is expensive, risky or even dangerous, such as medicine (Murphy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2001), education (Mandel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2014), economics (Hirano et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2003), and recommender systems (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The existing off-policy evaluation is mostly based on importance sam- pling techniques, and thus suffers from high variance exponentially increasing in time horizon, known as “the curse of the horizon” (Jiang and Li, 2016;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Recently, a promis- ing idea of using Marginalized Importance Sampling (MIS) is proposed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For example, for an infinite-horizon discounted MDP, Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2018) computes the importance weights regarding the state distribution by solving a minimax optimization problem, and proposes a method to estimate the expected return estimation and Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020) proposes a Minimax Weight Learning (MWL) algorithm that directly estimates the ratio of the state-action distribution without reference to the information of behavior policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In many reinforcement learning applications, the objectives of learning are to achieve some prescribed goals so that the processes are terminated at certain finite but random times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Exam- ples can be found in many robotic applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In situations as such, it is then not appropriate to use finite-time and infinite-time to model the environments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' MDPs with absorbing states, with the absorption to reflect the termination of processes, are thus suitable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Moreover, from Preprint submitted to ****** January 10, 2023 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='03183v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='LG] 9 Jan 2023 a theoretical perspective, absorbing MDPs extend infinite-horizon discounted MDP processes (Altman, 1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The theory of controllable absorbing MDPs has been widely studied and understood.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A stop- ping time to reach the absorbing state is discussed in Chatterjee et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2008) and Iida and Mori (1996)), which depends on the state and action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The minimization of the expected undiscounted cost until the state enters the absorbing set is studied in, for example, pursuit problems, tran- sient programming and first pass problems (Eaton and Zadeh, 1962;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Derman, 1970;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Kushner, 1971).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Other research efforts include stochastic shortest path problem (Bertsekas and Tsitsiklis, 1991), the control-to-exit time problem (Kesten and Spitzer, 1975;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Borkar, 1988), and so on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' There are also quite a few of problems that can be included in the category of absorbing MDPs, including board games (a game terminates once the winner is determined), trips through a maze, and dialog systems (a session terminates when the conversation is concluded) (Jiang, 2017), to name just a few.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We aim to extend MWL method aforestated to the problems of OPE for absorbing MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Since the absorbing MDPs have indefinite-horizon, and some of them are short, it is natural to assume that data consists of trajectories, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' each trajectory is treated as a single data point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' However, in practice many trajectories are not recorded completely because they are too long, expensive or other reasons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In other words, the collected trajectory may be truncated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Naturally, an interesting and important problem in theory and practice is to quantify the possible errors resulting from the truncated data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Though researchers frequently need to handle the OPE in MIS experiments where many benchmark environments are indeed episodic and have random trajectory lengths, to our knowledge, there is few literature to discuss the absorbing MDPs and study their errors of OPE from the data truncation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In this paper, we propose a new but MWL-like algorithm (MWLA for short) with data set consisting of the truncated trajectories of an absorbing MDP, or in other words, the truncated episode data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We derive the estimation of the expected total return.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' An upper bound of the MSE of the MWLA algorithm is established in Theorems 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2, showing that the MSE mainly consists of three parts: statistical error, approximation error, and optimization error.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We explicitly evaluate the statistical errors by the truncation level and data size, and more importantly, we get the uniformly bound of MSE by optimizing the truncations when the truncation level is relatively large.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Some numerical experiments in the episodic taxi environment to show the effectivity of our algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The left is organized as follows: Section 2 introduces the model and specifies some basic settings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The MWLA algorithm and its theoretical guarantees with unknown behavior policy is presented in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A parallel version of MWLA, referred to MSWLA there, is also discussed for absorbing MDP with known behavior policies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In Section 4, under the assumption that the data consists of i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' episodes, MSE bounds for the MWLA method are given in Theorems 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When the function classes are VC classes, compared with Theorem 9 in Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020), it is found that our statistical error is related to the truncation length H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The relevant work is discussed in Section 5 in more detail so as to clarify their connection to and differences 2 between the current work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In Section 6, a computer experiment is reported under the episodic taxi environment, compared with on-policy, naive-average, and MSWLA methods, estimations of return and MSE are given for different episode numbers and different truncation lengths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' All theoretical proofs and the pseudo-code of the algorithm are deferred to Appendix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Basic setting An MDP is a controllable rewarded Markov process that is represented by a standard tuple M = (S, A, R, P, µ) of a state space S, an action space A, a reward distribution R which maps a state-action pair (s, a) to a probability distribution R(s, a) with expectation R(s, a), a transfer probability function P : (s, a, s′) ∈ S×A×S → P(s′|s, a) ∈ [0, 1] and an initial state distribution µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The space S × A is assumed enumerable and R bounded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A policy π := π(a|s) is a time-homogeneous mapping from S to the family of all distributions on A, under which the agent interacts with the environment consecutively: starting with an initial state s0 ∼ µ, at any integer time t ≥ 0, an action at ∼ π(·|st) is sampled, a scalar reward rt from the distribution R(st, at) is collected, and a next state st+1 ∼ P(·|st, at) is then assigned by the environment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The probability distribution generated by M under a policy π and an initial distribution µ is denoted by Pµ,π and uses Eµ,π for its expectation operation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When the initial state s0 = i, the probability distribution and expectation are indicated by Pi,π and Ei,π respectively, so that Pµ,π = � i∈S µ(i)Pi,π and Eµ,π = � i∈S µ(i)Ei,π.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We also use P(s,a),π and E(s,a),π to indicate the probability and expectation generated by M when it is initialized by the state-action pair (s, a) and then follows policy π.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' An absorbing state, which is denoted by ξ, in S is such that r(ξ, a) = 0 and P(ξ|ξ, a) = 1 for all a ∈ A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Without loss of generality, the absorbing state is assumed to be unique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For a trajectory, denote the terminal time by T .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= min{n ≥ 1, sn = ξ}, where and in the paper .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= reads “defined as”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' An MDP is absorbing if Pi,π(T < ∞) = 1 for all states i ̸= ξ and all policies π.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Denote S0 = S \\ {ξ}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We need the following assumption on T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' sup(s,a)∈S0×A,π E(s,a),π(T) < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The expected return of a rewarded Markov process depends only on the transition probability and the mean rewards rather than the distributions of the rewards at any state-action pair.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For an MDP, the expected return under a policy π is Rπ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= Eµ,π �T−1 � t=0 rt � = ∞ � t=0 Eµ,π [R(st, at)] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (1) Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' With the expected return defined in (1) as the objective, the MDP models with absorbing states are more general than the infinite-horizon MDPs M = (S, A, r, P, µ, γ) under discounted expected return Rπ = Eµ,π[ ∞ � t=0 γtrt], where γ is a discount factor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' This can be done by introducing a dummy absorbing state κ such that the probability of transferring to κ is 1−γ from any state action pair (s, a) because they have the same Bellman optimality equations (Altman, 3 1999).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' However, it worth noting that, unlike in the standard discounted infinite-horizon MDPs, in this artificially constructed MDP, as the survival probability of the system, the parameter γ is unknown and needs to be estimated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Especially, for an arbitrary but fixed state-action pair (s, a) ∈ S0×A, regarding the indication function 1(s,a) as a special mean reward function, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', collecting a unit reward at the state-action pair (s, a) and zero otherwise, we can introduce the so-called occupancy measure as dπ(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= Eµ,π � T−1 � t=0 1(s,a)(st, at) � = ∞ � t=0 Pµ,π(st = s, at = a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2) Note that dπ(s, a) < ∞ by Assumptions 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Here, note that dπ(s, a) implicitly depends on the initial distribution µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Conceptually, dπ can be retrieved from the commonly known Chapmann- Kolmogorov equation dπ(s, a) = µ(s)π(a|s) + ∞ � t=1 Pµ,π(st = s, at = a) = µ(s)π(a|s) + ∞ � t=1 � (s′,a′)∈S0×A Pµ,π(st−1 = s′, at−1 = a′)P(s|s′, a′)π(s|a) = µ(s)π(a|s) + � (s′,a′)∈S0×A P(s|s′, a′)π(s|a)dπ(s′, a′), for all (s, a) ∈ S0 × A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (3) Conversely, with the measure dπ(s, a), it follows that Rπ = � (s,a)∈S0×A R(s, a)dπ(s, a) is an integration of the mean reward function with respect to the occupancy measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' If we define a function Φπ(q) for any q ∈ B(S0 × A) by Φπ(q) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= Eµ,π[ T−1 � t=0 q(st, at)], where B(S0×A) denotes the class of bounded functions on S0×A, with q(ξ, a) = 0 for all a ∈ A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We can readily see that Rπ = Φπ(R) and dπ(s, a) = Φπ(1(s,a)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Furthermore, a simple recursive argument shows that Φπ(q) = � (s,a)∈S0×A q(s, a)dπ(s, a) = � s∈S0 µ(s)q(s, π) + � (s,a,s′)∈S0×A×S0 P(s′|s, a)q(s′, π)dπ(s, a), 4 where for any function q, q(s, π) is the shorthand for � a∈A π(a|s)q(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A direct result of the equality is � (s,a)∈S0×A � �q(s, a) − � s′∈S0 P(s′|s, a)q(s′, π) � � dπ(s, a) = � s′∈S0 µ(s′)q(s′, π).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (4) Denote by dπ(s, a, s′) = dπ(s, a)P(s′|s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then equation (4) can be rewritten as � (s,a,s′)∈S0×A×S0 q(s′, π)dπ(s, a, s′) − � (s,a)∈S0×A q(s, a)dπ(s, a) + Es∼µ � q(s, π) � = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (5) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Minimax Weight Learning for absorbing MDP (MWLA) In the rest of this paper, we aim to estimate the expected return of a target policy πe under a given initial distribution µ, by using a set of offline trajectories τ i, i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , m collected from a possibly different and unknown behavior policy πb and truncated at a level H a priori specified.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' More clearly, let Z = (s0, a0, s1, a1, · · · , sT−1, aT−1) and Zi = (si 0, ai 0, si 1, ai 1, · · · , si Ti−1, ai Ti−1), i = 1, 2 · · · , m (6) be a representative episodes of an absorbing MDP with probability distribution Pµ,πb and its i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' copies, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The dataset D contains m i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' trajectories {τ i, i = 1, · · · , m}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Each τ i is a realization of Zi but only has the first li = Ti ∧ H consecutive sample transitions, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' τ i = � si 0, ai 0, ri 0, si 1, ai 1, ri 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , si li−1, ai li−1, ri li−1, si Ti∧H � , i = 1, 2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , m, where Ti denotes the absorbing time of episode i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Note that Ti may not be observed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' But Ti, i = 1, 2, · · · , m, are i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d with the distribution Pµ,πb(T = k) for each positive integer k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The goal of this paper is to estimate the expected return Rπe by using the data aforedescribed and analyze or control the errors caused by the truncation level H specified beforehand.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For the infinite-horizon discounted MDPs, Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020) proposes a Minimax Weight Learning (MWL) algorithm to handle the estimation of the expected discounted return, which is agnostic to the knowledge of πb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Their method uses a discriminator function class Q to learn the importance weight w (see equation (7) below) on state-action pairs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' One of their important tools is the (normalized) discounted occupancy which can be approximated well considering the given discount factor γ and the suitable dataset (for example, the dataset consisting of i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' tuples (s, a, r, s′)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' However in our setting, the normalized occupancy is invalid since the reward is not discounted.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In this section, we extend the MWL method to the absorbing MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Our method is es- sentially based on another occupancy measure defined by the first equation in (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The corre- spondingly developed algorithm, as what is indicated in the title of this section, is referred to as MWLA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 5 For any (s, a) ∈ S0 × A, let w πe πb (s, a) := dπe(s, a) dπb(s, a), almost everywhere dπb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (7) Observe that Rπe = � (s,a)∈S0×A R(s, a)dπe(s, a) = � (s,a)∈S0×A w πe πb (s, a)R(s, a)dπb(s, a) = Φπb(w πe πb R), if dπe(s, a) > 0 implies dπb(s, a) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' With dπb, R and w πe πb being estimated by ˆdπb, ˆR and ˆw πe πb , respectively, a plug-in idea suggests that Rπe can be simply estimated by ˆΦπb( ˆw πe πb ˆR) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= � (s,a)∈S0×A ˆw πe πb (s, a) ˆR(s, a) ˆdπb(s, a), in which the key is to estimate w πe πb .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We need the following technical assumption.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' There exists a constant Cw > 0, such that sup(s,a)∈S0×A w πe πb (s, a) ≤ Cw.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' With equality (5), we formally introduce a loss function L(w, q) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= � (s,a,s′)∈S0×A×S0 w(s, a)q(s′, πe)dπb(s, a, s′) − � (s,a)∈S0×A w(s, a)q(s, a)dπb(s, a) + Es∼µ � q(s, πe) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (8) Obviously, B(S0 × A) × B(S0 × A) ⊂ D(L) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= � (w, q) : L(w, q) < ∞ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Recall that identity (5) simply states that L(w πe πb , q) = 0 for all q ∈ B(S0 × A).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Conversely, by taking a family of particular functions {q(s, a) = 1(¯s,¯a)(s, a) : (¯s, ¯a) ∈ S0 × A}, as what has been done in (3), we have the following result on the uniqueness of the solution to this system of equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assume dπe is the unique solution to the systems of equations on q, q(s, a) = µ(s)πe(a|s) + � (s′,a′)∈S0×A P(s|s′, a′)πe(s|a)q(s′, a′), (s, a) ∈ S0 × A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (9) If Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 holds and dπe(s, a) > 0 implies dπb(s, a) > 0 for all (s, a) ∈ S0 × A, then 6 w = w πe πb is the unique bounded solution to the system of equations L(w, q) = 0 for each q ∈ l2(S0 × A) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= � � �g : � (s,a)∈S0×A g2(s, a) < ∞ � � � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 simply states that w πe πb = argmin w max q∈l2(S0×A) L(w, q)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (10) In order to retrieve the solution to (10), we introduce two function classes: W : S0 × A → R as working class of w πe πb , and Q : S0 × A → R to be treated as discriminators, then use w∗(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= argmin w∈W max q∈Q L(w, q)2 to approximate w πe πb .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The theorem below will be helpful in bounding the estimation error of occupancy measure ratio by means of the mini-max loss via the identification function class Q chosen properly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The following bounds hold: (1) max q∈Q |L(w, q)| ≥∥ dπe − wdπb ∥∞, if {±Vs′,a′,πe : (s′, a′) ∈ S0 × A} ⊆ Q;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2) min w∈W max q∈Q |L(w, q)| ≥∥ w πe πb − w∗ ∥∞, if {±Vs′,a′,πe/dπb(s′, a′) : (s′, a′) ∈ S0 × A} ⊆ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' where ∥ · ∥∞ denotes the supremum norm and Vs,a,πe(s′, a′) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= ∞ � t=0 P(s,a),πe(st = s′, at = a′), ∀ (s, a) ∈ S0 × A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (11) In order to construct estimators for dπb, R and w πe πb with the dataset of m i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' trajectories τ i, for all (s, a, s′) ∈ S0 × A × S0, define ˆdi πb(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= li−1 � t=0 1(s,a)(si t, ai t), ˆdi πb(s, a, s′) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= li−1 � t=0 1(s,a,s′)(si t, ai t, si t+1) (12) and ˆri(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= �li−1 t=0 ri t1(s,a)(si t, ai t) ˆdiπb(s, a) if ˆdi πb(s, a) > 0 and 0 otherwise to be the empirical occupancy measures and rewards from the single i-th episode, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 7 Let ˆRm(s, a) = 1 m m � i=1 ˆri(s, a), ˆdm(s, a) = 1 m m � i=1 ˆdi πb(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In addition, for any w, q ∈ B(S0 × A), introduce the empirical loss function ˆLm(w, q) = 1 m m � i=1 � (s,a,s′)∈S0×A×S0 w(s, a)q(s′, πe) ˆdi πb(s, a, s′) − 1 m m � i=1 � (s,a)∈S0×A w(s, a)q(s, a) ˆdi πb(s, a) + Es∼µ � q(s, πe) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (13) An estimator of w πe πb can then be defined as ˆwm(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= argmin w∈W max q∈Q ˆLm(w, q)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Naturally, we estimate Rπe by ˆRπe .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= � (s,a)∈S0×A ˆwm(s, a) ˆRm(s, a) ˆdm(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' This estimation procedure is referred to as an MWLA algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Unlike the MWL algorithm in Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020), the estimators defined here are not based on the (s, a, r, s′) tuple data but on truncated episodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Clearly, the performance of the estimation depends on the sample size m and the truncation level H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We need to understand how the errors vary as m and H, which can help us to understand better the effects of truncating episodes and/or find a suitable level H to balance the errors caused by the truncation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We discuss this problem in next section, which also is one of the main contributions of this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Below is an example of the MWLA algorithm applied to the tabular MDP with an absorbing state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Example 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Consider an absorbing tabular MDP with S0 = {0, 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , n−1}, A = {0, 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , h− 1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The function classes are W = Q = � ga(k, l) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= akh+l, k ∈ S0, l ∈ A : a = (a0, · · · , anh−1)⊤ ∈ Rnh� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For every 0 ≤ k ≤ n− 1, 0 ≤ l ≤ h− 1, denote by 1(k,l) the nh-dimensional column vector whose (kh + l)-th component is 1 and the others are 0, let 1(k,πe) = �h−1 l=0 πe(l|k)1(k,l).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' taking 8 w = gu and q = gv the empirical loss function is ˆLm(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) = 1 m m � i=1 � (k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='l,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='v)∈S0×A×S0 gu(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' l)gv(v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' πe) ˆdi πb(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' l,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' v) − 1 m m � i=1 � (k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='l)∈S0×A gv(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' l) ˆdi πb(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' l) + � k∈S0 µ(k)gv(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' πe) = (u⊤ ˆA + b⊤)v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' where ˆA = 1 m m � i=1 n−1 � k=0 h−1 � l=0 n−1 � v=0 1(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='l)1⊤ (v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='πe) ˆdi πb(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' l,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' v) − 1 m m � i=1 n−1 � k=0 h−1 � l=0 1(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='l)1⊤ (k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='l) ˆdi πb(k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' l) = 1 m m � i=1 Ti∧H−1 � t=0 1(si t,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='ai t) � � a∈A πe(a|si t+1)1⊤ (si t+1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a) − 1⊤ (si t,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='ai t) � is an nh × nh matrix and b = � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A µ(s)πe(a|s)1(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a) is an nh-dimensional vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' There- fore, ˆL2 m(w, q) = v⊤(ˆA⊤u + b)(ˆA⊤u + b)⊤v = ∥v∥ 2∥ˆA⊤u + b∥ 2 2, where ∥ · ∥2 denotes the Euclidean norm on Rnh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The estimator ˆwm is ereadily seen as the follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (1) When the matrix ˆA is invertible, ˆwm = gˆu, where ˆu = −(ˆA⊤)−1b, since ˆLm(gˆu, q) ≡ 0 for any q ∈ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2) When the matrix ˆA is not invertible, but the equation b = −ˆA⊤u (14) is consistent, ˆwm = gˆu where ˆu = −(ˆA+)⊤b is the minimum norm least square solution of (14), because we have ˆLm(gˆu, q) ≡ 0 for any q ∈ Q (ˆA+ is the Moore-Penrose inverse of ˆA).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (3) When equation (14) is inconsistent, ˆwm = gˆu where u = −(ˆA+)⊤b is the minimum norm least square solution of (14), because we also have ˆL2 m(w, q) ≥ ˆL2(gˆu, q) for any q ∈ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' If we define dπ(s) = Φπ(1{s}), then dπ(s, a) = dπ(s)π(a|s) and from (5), we have that � (s,a,s′)∈S0×A×S0 q(s′, π)dπ(s)π(a|s)P(s′|s, a) − � s∈S0 q(s, π)dπ(s) + Es∼µ � q(s, π) � = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 9 For a given target policy πe, simply denote q(s, πe) by q(s), so that the equation above can be rewritten as � (s,a,s′)∈S0×A×S0 w πe πb (s)q(s′)πe(a|s) πb(a|s)dπb(s, a, s′) − � s∈S0 w πe πb (s)q(s)dπb(s) + Es∼µ � q(s) � = 0, where w πe πb (s) = dπe(s) dπb(s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' With this equation, when the behavior policy πb is known, we can construct a corresponding estimate of the value function based on the minimax optimization problem: min w∈Ws max q∈Qs � � (s,a,s′)∈S0×A×S0 w πe πb (s)q(s′)πe(a|s) πb(a|s)dπb(s, a, s′)− � s∈S0 w πe πb (s)q(s)dπb(s)+Es∼µ � q(s) ��2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For convenience, we refer to the method as the MSWLA algorithm which is essentially an extension of the method discussed in Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' By similar arguments in Example 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1, in the tabular case where S0 = {0, 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , n − 1} and the function classes Ws and Qs are Rn, the empirical loss function for the MSWLA algorithm is ˆLm(w, q) = (u⊤ ˆA + b⊤)v for any w ∈ Ws, q ∈ Qs, where ˆA = 1 m m � i=1 Ti∧H−1 � t=0 1{si t} �πe(ai t|si t) πb(ai t|si t)1⊤ {si t+1} − 1⊤ {si t} � , b = � s∈S0 µ(s)1{s}, and for any s ∈ S0, 1{s} is the n-dimensional column vector whose s-th entry is 1 and other elements are 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' MSE of estimated return In this section, we discuss how the error of ˆRπe varies with the episodic size m and the truncation level H in terms of the mean squared error (MSE).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' To state our main results, we need the so-called state-action function Qπe(s, a) which is defined by Qπe(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= E(s,a),πe( T−1 � t=0 r(st, at)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Denote by Hm the unique solution to the equation 2mx2 + 2 ln m ln x − ln m = 0, x > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The following theorems state the results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Suppose that 10 1) there exists constants K0, K1 such that for any w ∈ W, q ∈ Q, ∥w∥2 := � � (s,a)∈S0×A w2(s, a) �1/2 ≤ K0, ∥q∥∞ := sup (s,a)∈S0×A |q(s, a)| ≤ K1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 2) W and Q have finite pseudo-dimensions DW and DQ, respectively;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 3) Assumptions 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 hold;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 4) Qπe ∈ Q;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 5) There exists λ0 > 0 such that Eµ,πb(eλ0T ) = M0 < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then we have the following: 1) When M0e−λ0H > Hm, there exists a constant C independent of H, m, such that E � ( ˆRm − Rπe)2� ≤ C � e−2λ0H + H2 ln m m � + 8 min w∈W max q∈Q L(w, q)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 2) When M0e−λ0H ≤ Hm, there exists a constant C independent of H, m, such that E � ( ˆRm − Rπe)2� ≤ C ln3 m m + 8 min w∈W max q∈Q L(w, q)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Especially, if w πe πb ∈ W and M0e−λ0H ≤ Hm, then E � ( ˆRm − Rπe)2� ≤ C ln3 m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Suppose the assumptions in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 hold and m ≥ e but Qπe ̸∈ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (1) When M0e−λ0H > Hm, there exists a constant C independent of H, m, such that E � ( ˆRm − Rπe)2� ≤ C � e−2λ0H + H2 ln m m � + 16 min w∈W max q∈Q L(w, q)2 + 4 max w∈W min q∈Q L2(w, Qπe − q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2) When M0e−λ0H ≤ Hm, there exists a constant C independent of H, m, such that E � ( ˆRm − Rπe)2� ≤ C ln3 m m + 16 min w∈W max q∈Q L(w, q)2 + 4 max w∈W min q∈Q L2(w, Qπe − q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Obviously, the additional term max w∈W min q∈Q L(w, Qπe − q)2 becomes 0 when Qπe in the closure of Q under the metric || · ||∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 and Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 provide upper bounds of MSE, which are related to the truncation level H and the number m of the episodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' On one hand, for a small truncation level H (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', M0e−λ0H > Hm), the estimate errors include four terms: the pure truncation term e−2λ0H, the mixing term H2 ln m/m generated by the randomness of sampling, the approximation error 11 min w∈W max q∈Q L2(w, q), and the optimization error max w∈W min q∈Q L2(w, Qπe −q), in which the first two are caused by the randomness of statistics and the other two are from the approximation of the two function classes W and Q to wπ/πb and Qπ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' On the other hand, when the truncation level H is relatively large (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', M0e−λ0H ≤ Hm), the pure truncation term e−2λ0H and the mixing term H2 ln m/m can be dominated by C ln3 m/m which is dependent of the truncation level H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' This observation also approves in some sense that our MWLA algorithm can eliminate the curse of the horizon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In the following are more remarks on the results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Consider the case Qπe ∈ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For the infinite horizon MDP with m i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' tuples (si, ai, s′ i, ri), the error bound of the MWL method consists of a statistical error ln m m +R2 m(W, Q) and an approximation error min w∈W max q∈Q L(w, q)2, where Rm(W, Q) is the Rademacher complexity of the function class �� s, a, s′� �→ |w(s, a)(q(s ′, π) − q(s, a))| : w ∈ W, q ∈ Q � , as given in Theorem 9 of Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Let DW,DQ be the VC-subgraph dimension (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' pseudo-dimension) of W, Q, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Because Rm(W, Q) = O( � max (DW,DQ) m ) (Corollary 1 of Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2021), the statistical error is dominated by ln m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In our method with m i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' episodes, the MSE bound also includes an approximation error min w∈W max q∈Q L(w, q)2 and a statistical error.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When M0e−λ0H ≤ Hm, the statistical error is bounded by ln3 m m which includes an extra factor ln2 m in form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For m > e, one has � ln m/(2m) < Hm < ln m � e/m (Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='8 in Appendix).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Therefore, when M0e−λ0H > Hm, it follows that H ≤ ln M0+ln 2/2+ln m/2 and H2 ln m m ≤ C ln3 m m for some constant C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Whatever H is, the bounds in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 are both less than C(e−2λ0H + ln3 m/m) + 8 min w∈W max q∈Q L(w, q)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In the tabular setting, if we take W = {w : ∥w∥2 ≤ K0} and Q = {q : ∥q∥∞ < K1}, where K0 is a constant larger than Cw in Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1, then all assumptions in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Hence, E � ( ˆRm − Rπe)2� ≤ C � e−2λ0H + ln3 m m � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Connections to related work The research of the off-policy evaluation, which has been mainly modeled as an infinite and fixed finite horizon MDP, can be divided into two categories according to whether the behavior policy is known.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When the behavior policy is known, Importance Sampling (IS) is a method for reweighting rewards according to its likelihood ratio of πe over πb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Since the ratio is computed by a cumulative 12 product of the importance weight over action πe(a|s) πb(a|s) at each time step (Precup, 2000), the IS method suffers from a variance that is exponentially increasing in time horizon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' To reduce that extremely high variance, a series of off-policy estimation methods are proposed based on IS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For example, the Weighted Importance Sampling method, Stepwise Importance Sampling method, and Doubly Robust(DR) method can reduce the variance to certain degree (Cassel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 1976;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Robins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 1994;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Robins and Rotnitzky, 1995;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Bang and Robins, 2005).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' However, the exponential variance of IS-based methods cannot be significantly improved when the MDP has a high stochasticity (Jiang and Li, 2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The marginalized importance sampling method proves a promising improvement over IS by successfully avoiding the trouble of exponential variance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For example, for a finite-horizon inhomogeneous MDP, compared to weighting the whole trajectory, using a ratio wt(s) πe(a|s) πb(a|s) with wt(s) = dπe,t(s) dπb,t(s) to reweight the rewards r (Xie et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2019) gives rise to a lower variance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In an infinite horizon setting, based on a discounted stationary distribution, Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2018) proposes to use the ratio w πe πb (s) · πe(a|s) πb(a|s) with w πe πb (s) = dπe,γ(s) dπb,γ(s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The ratio w πe πb (s) is then estimated by a minimax procedure with two function approximators: one to model a weight function w πe πb (s), and the other to model V πb, as a discriminator class for distribution learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A further effort is made on the case of unknown behavior policies by Hanna et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2019), showing that the importance sampling method with history-based behavior policy estimation has lower asymptotic variance than the one with a known behavior strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The fitted Q- iteration,which uses dynamic programming techniques to fit Qπe directly from the data, can overcome the curse of dimensionality, with a const of assuming that the function class contains Qπe and is closed under the bellman update Bπe, so as to avoid a high bias, see Ernst et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2005) and Le et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020) proposes the MWL algorithm by means of estimating marginalized importance weight w πe πb (s, a) = dπe,γ(s,a) dπb(s,a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A Dualdice algorithm is further proposed to estimate the discounted stationary distribution ratios (Nachum et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2019a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Nachum et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2019b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Nachum and Dai, 2020) where the loss function can be considered as a derivative of the loss function in Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Jiang and Huang (Jiang and Huang, 2020) briefly discuss about how to derive the equivalent of MWL/MQL as well as their value interval for finite-horizon MDPs and average-reward MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In reinforcement learning, there are only a few efforts made on absorbing MDP, though many benchmark environments are indeed episodic and have random horizons, such as board games (a game terminates once the winner is determined), trips through a maze, and dialog systems (a session terminates when the conversation is concluded) (Jiang, 2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Researchers often handle absorbing MDPs as a special case of finite-horizon MDPs by padding all trajectories (with random lengths) to the same length with absorbing states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Another way to handle it practically is to use the infinite-horizon setup (with a sufficiently large discount factor γ), and whenever a trajectory terminates, we imagine it continuous to infinity with absorbing states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' However, when the random horizons are not bounded and the random episodes are not observed completely, especially, accompanied by the non-discounted rewards, new issues will arise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For example, how do the unobserved trajectories affect the results?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' As our results show, this problem, which is by 13 no means trivial, is essentially neglected when we simply apply the two ways mentioned above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In this paper we employ a direct way to handle the OPE for absorbing MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Our contri- butions are two folds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We introduce the MWLA algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The MWLA algorithm proposed in this paper is a variant of the MWL to fit the random horizon and truncated episodic data modeled by absorbing MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The difference is that the MWL algorithm uses basically (s, a, r, s′)-tuple data and the MWLA algorithm uses episodic data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We explicitly evaluate the errors caused by the truncation level and data size, and more importantly, we get the uniform bound of MSE by optimizing the truncations when the trunca- tion level is relatively large.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Experiments 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Setting The environment Taxi (Dietterich, 2000) is a two-dimensional grid world that simulates a taxi moving along a grid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The experiment is conducted with a 5 × 5 grid as shown in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The four corners are marked as R(ed), B(lue), G(reen), and Y(ellow).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Initially, the taxi randomly chooses a corner to wait for a passenger, who appears or disappears with probability at each of the four corners, and that passenger wishes to be transported to one of the four corners (also chosen randomly).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The taxi must pick up the passenger and drops him off at a destination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' An episode ends once a passenger is successfully dropped off at his destination.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Figure 1: Taxi Grid The elements of the model are as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' There are a total of 2000 states (25 × 24 × 5), including 25 taxi locations, 24 passenger appearance status, and 5 taxi status (empty or one of 4 destinations).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' There are four navigation actions that move the taxi one square North, South, East, or West, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Each action is deterministic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Besides, we can only take 3 and 2 actions when the taxi is at the boundary and the corner, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When a passenger is picked up in Taxi, it receives a reward of -1 at every time step, but a reward of 0 is collected if she is successfully dropped off at the right place.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A reward of -2 at each time step is collected if the taxi is empty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' As in Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2018), we first run a Q-learning with 400,000 iterations to produce a policy and then run 60,000 iterations to produce another policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The two policies are then regularized such that the probability of crossing any boundary of the grid is 0 and the sums of the probabilities of the actions are 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The first policy is treated as the target policy πe and the second as the policy π+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The behaviors are πb = απe + (1 − α)π+ with α ∈ {0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 14 4 R D 3 D 2 1 0 Y B 0 2 3 4Taxi environment above differs from the ones analyzed by others in the following aspects: (1) Initial Location.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' While the taxi starts in a randomly-chosen square in Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2018) and Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2020)), it starts in a randomly-chosen corner here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (2) Actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In previous research (Dietterich, 2000;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2020), there are other two actions: a Pickup action and a Putdown action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In the experiment, the true expected return of the target policy is approximated by a set of 2 × 106 on-policy Monte-Carlo trajectories, truncated at H = 500 to assure that most trials stop at the absorbing state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Results Provided here are the experiment results for MWLA and MSWLA, together with an on- policy algorithm and a naive averaging baseline algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The on-policy algorithm estimates the expected return by the direct average over a set of trajectories generated by the target policy itself, and the naive average baseline algorithm does it by a direct average over another set of trajectories generated by the behavior policy, all truncated at H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We also show the results of MWL applied to the data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The first experiment is on the MSEs of the four methods applied to m = 15000, 20000, 30000, 40000 and 50000 trajectories with policy mixing ratio α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 and varying truncation level H = {20, 50, 100, 150, 200}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A total of 100 duplicates of every parameter setting are generated with different random seeds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Two strategies of visualization are employed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In the upper panel of Figure 2, every graph corresponds to a fixed episode size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' It appears that the MSE of MWLA and MSWLA decreases at the beginning and then varies slowly when the truncation level increases, MWLA is better than MSWLA to a moderate degree, and both are significantly lower than the on-policy and naive averaging baseline algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The lower panel of Figure 2 provides a different perspective for the results, in which every graph corresponds to a fixed H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Figure 3 visualizes the results in the same setting as in Figure 2 but now we use the policy ratio α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 and it does show similar results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The estimated returns are depicted in Figures 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The estimates by both MSWLA and MWLA approach the true value as the number of trajectories increases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' MSWLA is slightly better than MWLA in mean but worse in fluctuation, so that gives rise to a slightly higher MSE as shown in Figure 2 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Conclusions In this paper, we study the off-policy policy evaluation problem in reinforcement learning for absorbing MDPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We propose the MWLA algorithm, which extends the MWL (Uehara et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 2020) algorithm to the episode data with truncation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assuming the collected data are i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' episodes with a given truncation level, we obtain the estimator of the expected total return Rπe and theoretically study their MSE under certain conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We evaluate the statistical errors by the data size and the truncation level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' If the behavior policy is known, we also briefly introduce the MSWLA algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' At last, in the episodic taxi environment, our method is compared to 15 Agend: The horizontal axis indicates the truncation level H and the vertical the logarithm of the MSE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Agend: The horizontal axis indicates the number of trajectories and the vertical the MSE, both are scaled in logarithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Figure 2: MSE of the four algorithms: α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 Agend: The horizontal axis indicates the truncation level H and the vertical the logarithm of the MSE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Agend: The horizontal axis indicates the number of trajectories and the vertical the MSE, both are scaled in logarithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Figure 3: MSE of the four algorithms: α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 16 On-Policy Naive Average MSWLA MWLA 3 3 log10MSE 2 - 2 0 1 0 50 100150 200 50 100150 200 50 100150 200 50 100150 200 50 100150 200 (a)m=15000 (b) m=20000 (c) m=30000 (d) m=40000 (e) m=50000On-Policy Naive Average MSWLA MWLA 2 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 (a) logTrajecories, H=20 (b) logTrajecories, H=50 (c) logTrajecories, H=100 (d) logTrajecories, H=150 (e) logTrajecories, H=200On-Policy ¥一 Naive Average MSWLA MWLA 3 3 log10MSE 2 - 2 1 0 1 1 1 50 100 150 50 100150 200 50 100150 200 50 100 150 200 50 100150 200 (a)m=15000 (b) m=20000 (c) m=30000 (d) m=40000 (e) m=50000On-Policy Naive Average MSWLA MWLA 2 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 (a) logTrajecories, H=20 (b) logTrajecories, H=50 (c) logTrajecories, H=100 (d) logTrajecories, H=150 (e) logTrajecories, H=200α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 Figure 4: Estimated expected return.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' the on-policy, naive-average, and MSWLA algorithm, the estimation of return and mean squared errors are given for different number of trajectories and different truncation lengths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' It is shown that the MWLA method has the lower MSE as the number of episodes and truncation length increase, significantly improving the accuracy of policy evaluation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' References Eitan Altman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Constrained Markov decision processes: stochastic modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Routledge, 1999.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Martin Anthony, Peter L Bartlett, Peter L Bartlett, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Neural network learning: Theoretical foundations, volume 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' cambridge university press Cambridge, 1999.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Heejung Bang and James M Robins.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Doubly robust estimation in missing data and causal inference models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Biometrics, 61(4):962–973, 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Dimitri P Bertsekas and John N Tsitsiklis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' An analysis of stochastic shortest path problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Mathematics of Operations Research, 16(3):580–595, 1991.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Vivek S Borkar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A convex analytic approach to markov decision processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Probability Theory and Related Fields, 78(4):583–602, 1988.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Claes M Cassel, Carl E S¨arndal, and Jan H Wretman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Some results on generalized difference estimation and generalized regression estimation for finite populations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Biometrika, 63(3):615–620, 1976.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Debasish Chatterjee, Eugenio Cinquemani, Giorgos Chaloulos, and John Lygeros.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Stochastic control up to a hitting time: optimality and rolling-horizon implementation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' arXiv preprint arXiv:0806.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3008, 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Cyrus Derman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Finite state markovian decision processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Technical report, 1970.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Thomas G Dietterich.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Hierarchical reinforcement learning with the maxq value function decomposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of artificial intelligence research, 13:227–303, 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Jo H Eaton and LA Zadeh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Optimal pursuit strategies in discrete-state probabilistic systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 1962.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 17 MSWLA MWLA True 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='25 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='20 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='0 I Return 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='30 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='25 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='30 xpected 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='30 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='35 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='35 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='35 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='40 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='40 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='40 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='45 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 (a) logTrajectores, H=50 (b) logTrajectores, H=100 (c) logTrajectores, H=150 (d) logTrajectores, H=200MSWLA MWLA True 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='9 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='25 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='275 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='0 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='30 IReturn 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='300 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='30 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='325 Expected 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='35 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='350 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='35 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='40 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='375 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='40 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='400 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='45 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='425 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6 (a) logTrajectores, H=50 (b) logTrajectores, H=100 (c) logTrajectores, H=150 (d) logTrajectores, H=200Damien Ernst, Pierre Geurts, and Louis Wehenkel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Tree-based batch mode reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of Machine Learning Research, 6:503–556, 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Josiah Hanna, Scott Niekum, and Peter Stone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Importance sampling policy evaluation with an estimated behavior policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 2605–2613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' PMLR, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' David Haussler.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Sphere packing numbers for subsets of the boolean n-cube with bounded vapnik-chervonenkis dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of Combinatorial Theory, Series A, 69(2):217–232, 1995.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Keisuke Hirano, Guido W Imbens, and Geert Ridder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Efficient estimation of average treatment effects using the estimated propensity score.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Econometrica, 71(4):1161–1189, 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Tetsuo Iida and Masao Mori.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Markov decision processes with random horizon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of the Operations Research Society of Japan, 39(4):592–603, 1996.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Nan Jiang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' A theory of model selection in reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' PhD thesis, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Nan Jiang and Jiawei Huang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Minimax value interval for off-policy evaluation and policy optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In 34th Conference on Neural Information Processing Systems, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Nan Jiang and Lihong Li.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Doubly robust off-policy value evaluation for reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In Interna- tional Conference on Machine Learning, pages 652–661.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' PMLR, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Harry Kesten and Frank Spitzer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Controlled markov chains.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The Annals of Probability, pages 32–40, 1975.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' HJ Kushner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Introduction to stochastic control.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' holt, rinehart and wilson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' New York, 1971.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Hoang Le, Cameron Voloshin, and Yisong Yue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Batch policy learning under constraints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 3703–3712.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' PMLR, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Lihong Li, Wei Chu, John Langford, and Xuanhui Wang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Unbiased offline evaluation of contextual-bandit- based news article recommendation algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In Proceedings of the fourth ACM international conference on Web search and data mining, pages 297–306, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Lihong Li, R´emi Munos, and Csaba Szepesv´ari.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Toward minimax off-policy value estimation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In Artificial Intelligence and Statistics, pages 608–616.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' PMLR, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Qiang Liu, Lihong Li, Ziyang Tang, and Dengyong Zhou.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Breaking the curse of horizon: Infinite-horizon off-policy estimation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 31, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Travis Mandel, Yun-En Liu, Sergey Levine, Emma Brunskill, and Zoran Popovic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Offline policy evaluation across representations with applications to educational games.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In AAMAS, volume 1077, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Susan A Murphy, Mark J van der Laan, James M Robins, and Conduct Problems Prevention Research Group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Marginal mean models for dynamic regimes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of the American Statistical Association, 96 (456):1410–1423, 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Ofir Nachum and Bo Dai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Reinforcement learning via fenchel-rockafellar duality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' arXiv preprint arXiv:2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='01866, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Ofir Nachum, Yinlam Chow, Bo Dai, and Lihong Li.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Dualdice: Behavior-agnostic estimation of discounted stationary distribution corrections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 32, 2019a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Ofir Nachum, Bo Dai, Ilya Kostrikov, Yinlam Chow, Lihong Li, and Dale Schuurmans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Algaedice: Policy gradient from arbitrary experience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' arXiv preprint arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='02074, 2019b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' David Pollard.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Convergence of stochastic processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Springer Science & Business Media, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Doina Precup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Eligibility traces for off-policy policy evaluation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Computer Science Department Faculty Publication Series, page 80, 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' James M Robins and Andrea Rotnitzky.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Semiparametric efficiency in multivariate regression models with missing data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of the American Statistical Association, 90(429):122–129, 1995.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' James M Robins, Andrea Rotnitzky, and Lue Ping Zhao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Estimation of regression coefficients when some regressors are not always observed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Journal of the American statistical Association, 89(427):846–866, 1994.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 18 Masatoshi Uehara, Jiawei Huang, and Nan Jiang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Minimax weight and q-function learning for off-policy evaluation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In International Conference on Machine Learning, pages 9659–9668.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' PMLR, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Masatoshi Uehara, Masaaki Imaizumi, Nan Jiang, Nathan Kallus, Wen Sun, and Tengyang Xie.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Finite sample analysis of minimax offline reinforcement learning: Completeness, fast rates and first-order efficiency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' arXiv preprint arXiv:2102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='02981, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Tengyang Xie, Yifei Ma, and Yu-Xiang Wang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Towards optimal off-policy evaluation for reinforcement learning with marginalized importance sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 32, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 19 Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof of Theorems in Section 3 The proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 is read as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' It suffices to prove the uniqueness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For any w ∈ B(S0 × A), let Fw(s, a) = µ(s)πe(a|s) + � (¯s,¯a)∈S0×A w(¯s, ¯a)P(s|¯s, ¯a)πe(a|s)dπb(¯s, ¯a) − w(s, a)dπb(s, a), for any (s, a) ∈ S0 × A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 implies that Fw ∈ l2(S0 × A).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Moreover, it is easy to check that L(w, q) = ⟨Fw, q⟩, where ⟨·, ·⟩ is the inner product in the Hilbert space l2(S0 × A).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' From the assumption L(w, q) = 0, ∀q ∈ l2(X × A), we know Fw ≡ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The assumption that dπe is the unique solution of (9) implies wdπb = dπe, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' w = w πe πb .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ In order to prove Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2, for any function q on S0 × A, we define a one-step forward operator under policy π, writing Lπ, by (Lπq)(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= � (s′,a′)∈S0×A q(s′, a′)π(a′|s′)P(s′|s, a) = Eπ [q(st+1, at+1)|st = s, at = a] , (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1) Then, the so-called state-action function Qπ(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= E(s,a),π( T−1 � t=0 r(st, at)) = R(s, a) + � s′∈S0 P(s′|s, a)Es′,π �T−1 � t=0 rt � = R(s, a) + � (s′,a′)∈S0×A P(s′|s, a)π(a′|s′)Qπ(s′, a′) = R(s, a) + (LπQπ)(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2) and the loss function L(w, q) = � s,a∈S0×A w(s, a) [Lπeq(s, a) − q(s, a)] dπb(s, a) + Es∼µ � q(s, πe) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3) By Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1, we can further change the form of L(w, q) as L(w, q) = � (s,a)∈S0×A � wπ/πb(s, a) − w(s, a) � [q(s, a) − (Lπeq)(s, a)] dπb(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4) Now we give the proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 20 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Note the definition of Vs′,a′,πe in (11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Equation (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2) states that Vs′,a′,πe(s, a) − LπeVs′,a′,πe(s, a) = 1(s′,a′)(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Put it into (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4), it follows that, for every (s′, a′) ∈ S0 × A, L(w, Vs′,a′,πe) = � (s,a)∈S0×A � w πe πb (s, a) − w(s, a) � 1(s′,a′)(s, a)dπb(s, a) = dπe(s′, a′) − w(s′, a′)dπb(s′, a′).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When {±Vs′,a′,πe : (s′, a′) ∈ S0 × A} ⊆ Q, the derised result follows from taking maximum of L over q ∈ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For the second assertion, consider particularly the function ±Vs′,a′,πe/dπb(s′, a′).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' If {±Vs′,a′,πe/dπb(s′, a′) : (s′, a′) ∈ S0 × A} ⊆ Q, then maxq∈Q |L(w, q)| ≥ ||w πe πb − w∗||∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Therefore, min w∈W max q∈Q |L(w, q)| = max q∈Q |L(w∗, q)| ≥ ||w πe πb − w∗||∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof of Theorems in Section 4 To prove Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 and Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2, we give the definition of pseudo-dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Definition B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (Anthony et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 1999) Let F be a set of functions mapping from a domain X to R and suppose that S = {x1, x2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , xm} ⊆ X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then S is pseudo-shattered by F if there are real number r1, r2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , rm such that for each b ∈ {0, 1}m there is a function fb in F with sgn (fb (xi) − ri) = bi for 1 ≤ i ≤ m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We say that r = (r1, r2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , rm) witnesses the shattering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Definition B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (Anthony et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=', 1999) Suppose that F is a set of functions from a domain X to R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then F has pseudo-dimension d if d is the maximum cardinality of a subset S of X that is pseudo-shattered by F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' If no such maximum exists, we say that F has infinite pseudo-dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The pseudo-dimension of F is denoted Pdim(F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Let us define the empirical ℓ1-covering numbers of a function class F with respect to a data set {Z1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , Zn}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In particular, we define the set F ({Zi}) ⊂ Rn as follows: F ({Zi}) = {(f (Z1) , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , f (Zn)) | f ∈ F} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We then define the empirical ℓ1-covering number as N (ϵ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' F ({Zi})) as the smallest number N of balls of radius ϵ required to cover F ({Zi}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In this definition, distances are measured in terms 21 of the empirical ℓ1-norm ∥f − g∥F({Zi}) := 1 n n � i=1 |f (Zi) − g (Zi)| .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We quote the following two results for easy reference later on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (Pollard, 2012) Let F be a permissible class of Z → [−M, M] functions and {Zi}n i=1 are i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' samples from some distribution P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then, for any given ϵ > 0, P � sup f∈F �� 1 n n � i=1 f � Zi � − Ef(Z) �� > ϵ � ≤ 8E � N (ϵ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' F ({Zi})) � exp � −nϵ2 512M2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1) The covering number can then be bounded in terms of the function class’s pseudo-dimension: Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (Corollary 3 (Haussler, 1995)) For any set Z, any points {zi}n i=1 ⊆ Z, any class F of functions on Z taking values in [0, M] with pseudo-dimension DF < ∞, and any ϵ > 0, N (ϵ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' F ({zi})) ≤ e (DF + 1) �2eM ϵ �DF .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2) Recall that in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1, we address that the functional classes W and Q have finite pseudo-dimensions DW and DQ, Qπe ∈ Q, and there exists constants K0, K1 such that for any w ∈ W, ∥w∥2 ≤ K0, ∥q∥∞ ≤ K1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In addition, we assume that there exists λ0 > 0 such that Eµ,πb(eλ0T ) = M0 < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Besides them, we also remind here that our data consist of i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d samples from M = (S, A, R, P, µ) under the policy πb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 proceeds in the following 3 parts: decomposition, evaluation, and optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1 Decomposition Note that the estimator of Rπe based on the truncated sample is ˆRm = 1 m m � i=1 � (s,a)∈S0×A ˆwm(s, a)ˆri(s, a) ˆdi πb(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' It is easy to see that ˆRm − Rπe = I1 + I2 + I3, (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3) where I1 = � (s,a)∈S0×A [ ˆwm(s, a) − w πe πb (s, a)]R(s, a)dπb(s, a), I2 = 1 m m � i=1 � (s,a)∈S0×A ˆwm(s, a)R(s, a) � ˆdi πb(s, a) − dπb(s, a) � , 22 and I3 = 1 m m � i=1 � (s,a)∈S0×A ˆwm(s, a) � ˆri(s, a) − R(s, a) � ˆdi πb(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' By (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2) and (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4), I1 = � (s,a)∈S0×A [ ˆwm(s, a) − w πe πb (s, a)][Qπe(s, a) − (LπeQπe)(s, a)]dπb(s, a) = −L( ˆwm, Qπe).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Because Qπe ∈ Q, it follows that I2 1 = � L( ˆwm, Qπe) − ˆLm( ˆwm, Qπe) + ˆLm( ˆwm, Qπe) �2 ≤ 2 � L( ˆwm, Qπe) − ˆLm( ˆwm, Qπe) �2 + 2 max q∈Q ˆL2 m( ˆwm, q), where ˆLm is defined in (13) and we have used the fact that ˆL2 m( ˆwm, Qπe) ≤ max q∈Q ˆL2 m( ˆwm, q) = min w∈W max q∈Q ˆL2 m(w, q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Using the fact max q∈Q ˆL2 m( ˆwm, q) ≤ 2 max q∈Q L2( ˆwm, q) + 2 max q∈Q � ˆLm( ˆwm, q) − L( ˆwm, q) �2 , we further have I2 1 ≤ 6I2 11 + 4 min w∈W max q∈Q L(w, q)2, (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4) where I2 11 = sup w∈W,q∈Q |L(w, q) − ˆLm(w, q)|2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Substituting (12) into the expression of ˆLm(w, q) in (13), it follows that ˆLm(w, q) − Eµ(q(s, πe)) = 1 m m � i=1 li−1 � t=0 w(si t, ai t)(q(si t+1, πe) − q(si t, ai t)), where q(ξ, πe) = 0 is assumed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Similarly, L(w, q) − Eµ(q(s, πe)) = Eµ,πb � T−1 � t=0 w(st, at)(q(st+1, πe) − q(st, at)) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Define ˜L(w, q) = Eµ,πb � T∧Hβ−1 � t=0 w(st, at)(q(st+1, πe) − q(st, at)) � 23 and ˜Lm(w, q) = 1 m m � i=1 Ti∧Hβ−1 � t=0 w(si t, ai t) � q(si t+1, πe) − q(si t, ai t) � , where Hβ is a constant specified later.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then |ˆLm(w, q) − L(w, q)| = |ˆLm(w, q) − Eµ(q(s, πe)) − (L(w, q) − Eµ(q(s, πe)))| ≤ | ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)| + |˜Lm(w, q) − ˜L(w, q)| +|˜L(w, q) − (L(w, q) − Eµ(q(s, πe)))|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='5) B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 Evaluation For any β ≥ M0e−λ0H, let Hβ = min{k : M0e−λ0k ≤ β} = ⌈ln(M0/β)/λ0⌉, where ⌈x⌉ is the minimum integer no less than x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Obviously, Hβ ≤ H and M0e−λ0Hβ ≤ β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1) To get the upper bound of E(I2 11) An upper bound of E(I2 11) is derived via a sequence of auxiliary results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' With the constants K0 and K1, for any β ≥ M0e−λ0H, E( sup w∈W,q∈Q |ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)|2) ≤ 4K2 0K2 1 λ2 0 � β2 + 2β m � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For any w ∈ W and q ∈ Q, |ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)| = ��� 1 m m � i=1 li−1 � t=Ti∧Hβ w(si t, ai t) � q(si t+1, πe) − q(si t, ai t) � 1{Ti>Hβ} ���.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Recall the notation li = Ti ∧ H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Since w and q are bounded by K0 and K1, respectively, |ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)| ≤ 2K0K1 m m � i=1 (Ti − Hβ)1{Ti>Hβ}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Therefore, E( sup w∈W,q∈Q |ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)|2) ≤ 4K2 0K2 1 m2 � � m � i,j=1,i̸=j E[(Ti − Hβ)(Tj − Hβ)1{Ti,Tj>Hβ}] + m � i=1 E[(Ti − Hβ)21{Ti>Hβ}] � � = 4K2 0K2 1 m2 � m(m − 1)E2[(T1 − Hβ)1{T1>Hβ}] + mE[(T1 − Hβ)21{T1>Hβ}] � , (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6) where the equality follows from the i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' property of trajectories.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' By further the inequality 24 x ∨ x2/2 ≤ ex for every x > 0, E[(T1 − Hβ)1{T1>Hβ}] ≤ 1 λ0 E � eλ0(T1−Hβ)1{T1>Hβ} � ≤ M0e−λ0Hβ λ0 ≤ β λ0 , (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='7) and E[(Ti − Hβ)21{Ti>Hβ}] ≤ 2 λ2 0 E(eλ0(T1−Hβ)1{T1>Hβ}) ≤ 2M0e−λ0Hβ λ2 0 ≤ 2β λ2 0 , (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='8) substituting (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='7) and (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='8) into (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6) leads to the desired result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' There exists a constant C1 independent of m, H such that for m ≥ 2, E( sup w∈W,q∈Q |˜Lm(w, q) − ˜L(w, q)|2) ≤ C1H2 β ln m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For a representative trajectory Z of the form (6), denote by ˜wq(Z) = T∧Hβ−1 � t=0 w(st, at)(q(st+1, πe) − q(st, at)) so that ˜Lm(w, q) − ˜L(w, q) = 1 m m � i=1 ˜wq(Zi) − E( ˜wq(Z)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' It is easy to see that | ˜wq| ≤ 2K0K1Hβ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Denote by H = { ˜wq(Z) : w ∈ W, q ∈ Q}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The distance in H can be bounded by 1 m m � i=1 ��� Ti∧Hβ−1 � t=0 w1(si t, ai t)(q1(si t+1, πe) − q1(si t, ai t)) − Ti∧Hβ−1 � t=0 w2(si t, ai t)(q2(si t+1, πe) − q2(si t, ai t)) ��� ≤ 2K1Hβ||w1 − w2||∞ + 2K0Hβ∥q1 − q2∥∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' As a result, N1(2Hβ(K1ϵ1 + K0ϵ2), H, {Zi}m i=1) ≤ N1(ϵ1, W, {Zi}m i=1)N1(ϵ2, Q, {Zi}m i=1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then a direct application of Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2 shows that N1(2Hβ(K1ϵ1 + K0ϵ2), H, {Zi}m i=1) ≤ e2(DW + 1)(DQ + 1) �4eK0 ϵ1 �DW �4eK1 ϵ2 �DQ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 25 Taking ϵ1 = ϵ 32K1Hβ and ϵ2 = ϵ 32K0Hβ , we have N1 � ϵ 8, H, {Zi}m i=1 � ≤ M ϵDW+DQ , where M = e2(DW + 1)(DQ + 1)(128eK0K1Hβ)DW+DQ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' By Pollard’s tail inequality (Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1), P � sup w∈W,q∈Q ����� 1 m m � i=1 hw,q(Zi) − Ehw,q(Z) ����� > ϵ � ≤ 8M ϵDW+DQ exp � −mϵ2 2048(K0K1Hβ)2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='9) For any m > 1, let x0 = (32K0K1Hβ)2(DW+DQ) ln m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then, by (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='9), we have that E( sup w∈W,q∈Q |˜Lm(w, q) − ˜L(w, q)|2) ≤ � ∞ 0 P( sup w∈W,q∈Q |˜Lm(w, q) − ˜L(w, q)| ≥ √x)dx.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' ≤ � x0 0 dx + � ∞ x0 8M x(DW+DQ)/2 exp � −mx 2048(K0K1Hβ)2 � dx ≤ x0 + 8M x(DW+DQ)/2 0 � ∞ x0 exp � −mx 2048(K0K1Hβ)2 � dx = (32K0K1Hβ)2 m � (DW + DQ) ln m + 8M [(32K0K1Hβ)2 ln m(DW + DQ)](DW+DQ)/2 � , which implies the desired result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assume that W and Q have finite pseudo-dimensions DW and DQ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' If there exists constants K0, K1 such that for any w ∈ W, ∥w∥∞ ≤ K0, ∥q∥∞ ≤ K1, then there exists a constant C3 independent of m, H such that for m ≥ 2, E( sup w∈W,q∈Q |ˆLm(w, q) − L(w, q)|2) ≤ C3 � β2 + (1 − ln β + ln2 β)ln m m + β m � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' By (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='5), E[ sup w∈W,q∈Q |ˆLm(w, q) − L(w, q)|2] ≤ 3E( sup w∈W,q∈Q |ˆLm(w, q) − Eµ(q(s, πe)) − ˜Lm(w, q)|2) +3E( sup w∈W,q∈Q |˜L(w, q) − (L(w, q) − Eµ(q(s, πe)))|2) +3 sup w∈W,q∈Q |˜Lm(w, q) − ˜L(w, q)|2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='10) 26 Note that for any w ∈ W, |˜L(w, q) − (L(w, q) − Eµ(q(s, πe)))| = E � T−1 � T∧Hβ w(st, at)(q(st+1, πe) − q(st, at)) � ≤ 2K0K1E((T − Hβ)1{T≥Hβ}) ≤ 2K0K1E � eλ0(T−Hβ) λ0 1{T≥Hβ} � ≤ 2K0K1 λ0 β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='11) Substituting (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='11) into (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='10) and applying Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3 and Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4, we get that E( sup w∈W,q∈Q |ˆLm(w, q) − L(w, q)|2) ≤ 3M2 3 µ2 + 3 � M2 3 β2 + 2M2 3 m β � + 3C1H2 β ln m m , where M3 = 2K0K1/λ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Since Hβ ≤ ln(M0/β) λ0 + 1, it follows from (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='10) that E( sup w∈W,q∈Q |ˆLm(w, q) − L(w, q)|2) ≤ 3C1 λ2 0 � (ln β)2 − 2(ln M0 + λ0) ln β + (ln M0 + λ0)2� ln m m + 6M2 3 β2 + 6M2 3 m β ≤ 3C1 λ2 0 � (ln β)2 − 2C2 ln β + C2 2 � ln m m + 6M2 3 β2 + 6M2 3 β m , where C2 = ln M0 + λ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Let C3 = max �3C1 λ2 0 , 3C1 λ2 0 C2, 3C1 λ2 0 C2 2, 6M2 3 � , we can readily get the desired result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2) To get the upper bound of E(I2 2) and E(I2 3) Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assume that there exists constant K0 such that for any w ∈ W, ∥w∥2 ≤ K0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then E(I2 2) ≤ 2R2 maxK2 0M0 λ2 0 ( 1 m + M0e−2λ0H), where Rmax .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='= sup |R(s, a)|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Define a truncated occupancy measure ˜dπb(s, a) = Eµ,πb(�T∧H−1 t=0 1(s,a)(st, at)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then I2 = 1 m m � i=1 � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A ˆwm(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)R(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)[ ˆdi πb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) − ˜dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)] + � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A ˆwm(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)R(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)[ ˜dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) − dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)] 27 and hence I2 2 ≤ 2 � � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A ˆwm(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)R(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) � 1 m m � i=1 ˆdi πb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) − ˜dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) ��2 +2 � � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A ˆwm(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)R(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) � ˜dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) − dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) ��2 ≤ 2 � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A ˆwm(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)2R(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)2 � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A � 1 m m � i=1 ˆdi πb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) − ˜dπb(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a) �2 +2 � � (s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)∈S0×A ˆwm(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)R(s,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a)Eµ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='πb( T−1 � t=T∧H 1(s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='a)(st, at)) �2 , where the last inequality follows from H¨older’s inequality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Invoking the bounds from ˆwm and R, we get that E(I2 2) ≤ 2R2 maxK2 0 � � (s,a)∈S0×A E � 1 m m � i=1 ˆdi πb(s, a) − ˜dπb(s, a) �2 + � Eµ,πb � � (s,a)∈S0×A T−1 � t=T∧H 1(s,a)(st, at) ��2� ≤ 2R2 maxK2 0 � 1 m � (s,a)∈S0×A Var( ˆdi πb(s, a))) + � Eµ,πb((T − H)1{T>H}) �2� , where the first inequality is due to ˆwm ∈ W and the second inequality follows from the fact that di πb(s, a), 1 ≤ i ≤ m, are i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d and have expectation ˜dπb(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Observing that Var( ˆdi πb(s, a))) = Varµ,πb( T∧H−1 � t=0 1(s,a)(st, at)) ≤ Eµ,πb �� T∧H−1 � t=0 1(s,a)(st, at) �2� ≤ Eµ,πb � (T ∧ H) T∧H−1 � t=0 1(s,a)(st, at) � , we obtain that � (s,a)∈S0×A Var( ˆdi πb(s, a))) ≤ Eµ,πb � (T ∧ H) � (s,a)∈S0×A T∧H−1 � t=0 1(s,a)(st, at) � ≤ Eµ,πb � T 2� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Since Eµ,πb � T 2� ≤ 2M0 λ2 0 and Eµ,πb((T − H)1{T>H}) ≤ M0 λ0 e−λ0H, we arrive at the conclusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ To estimate E(I2 3), we have the following result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Assume that there exists constant K0 such that for any w ∈ W, ∥w∥2 ≤ K0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Then E(I2 3) ≤ R2 maxK2 0M0 λ0m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 28 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Noting that I3 = � (s,a)∈S0×A ˆwm(s, a) � 1 m m � i=1 (ˆri(s, a) − R(s, a)) ˆdi πb(s, a) � , applying the H¨older inequality, we have that I2 3 ≤ � (s,a)∈S0×A ˆw2 m(s, a) � (s,a)∈S0×A � 1 m m � i=1 (ˆri(s, a) − R(s, a)) ˆdi πb(s, a) �2 ≤ K2 0 � (s,a)∈S0×A � 1 m m � i=1 (ˆri(s, a) − R(s, a)) ˆdi πb(s, a) �2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Therefore, E[I2 3] ≤ K2 0E � � � (s,a)∈S0×A E �� 1 m m � i=1 (ˆri(s, a) − R(s, a)) ˆdi πb(s, a) �2��� ˆdi πb(s, a), i = 1, · · · , m �� � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Because ˆdi πb(s, a), i = 1, · · · , m are i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='d and ri(s, a) follows distribution R(s, a) and is indepen- dent of ˆdi πb(s, a), i = 1, · · · , m, it follows that E �� 1 m m � i=1 (ˆri(s, a) − R(s, a)) ˆdi πb(s, a) �2��� ˆdi πb(s, a), i = 1, · · · , m � = 1 m2 m � i=1 ( ˆdi πb(s, a))2Var � ˆri(s, a) ��� ˆdi πb(s, a) � = 1 m2 m � i=1 ( ˆdi πb(s, a))2Var ��li−1 t=0 ri t1(s,a)(si t, ai t) ˆdiπb(s, a) ��� ˆdi πb(s, a) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When ˆdi πb(s, a) is given, �li−1 t=0 ri t1(s,a)(si t, ai t) is the sum of ˆdi πb(s, a) random variables who are independent with the same distribution R(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Hence Var � ˆri(s, a) ��� ˆdi πb(s, a) � = VarR(s,a)(r) ˆdiπb(s, a) ≤ R2 max ˆdiπb(s, a) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Therefore E(I2 3) ≤ R2 maxK2 0 m2 E � � (s,a)∈S0×A m � i=1 ˆdi πb(s, a) � = R2 maxK2 0 m2 E � m � i=1 � (s,a)∈S0×A ˆdi πb(s, a) � = R2 maxK2 0 m2 E � m � i=1 (li − 1) � ≤ R2 maxK2 0 m Eµ,πb(T), which implies the desired result, since Eµ,πb(T) ≤ M0 λ0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ 29 B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3 Optimization Based on the above discussion, we get that for any truncation level H and β such that M0e−λ0H ≤ β, E(| ˆRm − Rπe|2) ≤ 2E(I2 1) + 4E(I2 2) + 4E(I2 3) ≤ 8 min w∈W max q∈Q L(w, q)2 + 12E(I2 11) + 4E(I2 2) + 4E(I2 3) ≤ 8 min w∈W max q∈Q L(w, q)2 + 12C3 � β2 + (1 − ln β + (ln β)2)ln m m + β m � +4R2 maxK2 0M0 mλ0 + 8K2 0R2 maxM0 λ2 0 ( 1 m + M0e−2λ0H), where the first inequality follows from (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='3) and the simple inequality (a + b)2 ≤ 2a2 + 2b2, the second inequality follows from (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4) and the last inequality is due to Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='5-Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Letting C4 = max{24C3 + 4R2 maxK2 0M0 λ0 + 8K2 0R2 maxM0 λ2 0 , 8K2 0R2 maxM2 0 λ2 0 }, we have that, for any truncation level H and β ≥ M0e−λ0H, E(| ˆRm − Rπe|2) ≤ 8 min w∈W max q∈Q L(w, q)2 + C4 � G(β, m) + e−2λ0H� , (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='12) where G(x, m) := x2 + (1 − ln x + ln2 x)ln m m , for x ∈ (0, +∞), m ≥ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Note that G′ x(x, m) = 2x + ln m(2 ln x − 1) mx , G′′ xx(x, m) = 2 + ln m(3 − 2 ln x) mx2 > 0, which combining the fact G′ x(1, m) > 1 and lim x→0+ G′ x(x, m) = −∞ implies that there exists a unique Hm ∈ (0, 1) such that 2mH2 m + 2 ln m ln Hm − ln m = 0, which implies G′ x(Hm, m) = 0 and hence for all x ∈ (0, +∞), G(Hm, m) ≤ G(x, m).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Moreover, G(x, m) is decreasing for x ∈ (0, Hm) while increasing for x ∈ (Hm, +∞).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' For any m ≥ e, � ln m/(2m) ≤ Hm ≤ � e ln2 m/m and there exists constants 0 < k1 < k2 such that k1 ln3 m m ≤ G(Hm, m) ≤ k2 ln3 m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 30 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' From G′ µ(Hm, m) = 0, we know that Hm is a solution of Hm(x) := 2mx2 + 2 ln m ln x − ln m = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Note that Hm(x) is increasing on (0, 1] and Hm �� ln m 2m � = 2 ln m ln �� ln m 2m � < 0, Hm � � � e ln2 m m � � = (2e − 1) ln2 m + 2 ln m ln ln m > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We have that � ln m/(2m) ≤ Hm ≤ � e ln2 m/m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' To prove the second assertion, we note that G(Hm, m) ≤ G � � � ln3 m m , m � � ≤ ln3 m m + (1 + 1 2 ln m + 1 4 ln2 m)ln m m ≤ 3ln3 m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' On the other hand, when x < � ln3 m/m, G(x, m) ≥ (1 − ln x + ln2 x)ln m m ≥ 1 4 � ln ln3 m m �2 ln m m ≥ ln3 m 4m � 1 − 3ln ln m ln m �2 ≥ (e3 − 9)2 4e6 ln3 m m , for m ≥ ee3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Noting that ln3 m/m > 0 for all m > e, we can find a constant k′ such that G(x, m) ≥ k′ ln3 m m , for all x ∈ (0, � ln3 m/m) and m > e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Moreover, when x ≥ � ln3 m/m, G(x, m) > x2 ≥ ln3 m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Consequently, there exists a constant k1 such that G(Hm, m) ≥ k1 ln3 m/m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ Now we are at the position to finish the proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' From (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='12) and Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='8, it follows that if Hm ≥ M0e−λ0H, there is a constant C 31 independent of m, H such that E(| ˆRm − Rπe|2) ≤ 8 min w∈W,q∈Q L2(w, q) + C4 � min β≥M0e−λ0H G(β, m) + e−2λ0H� = 8 min w∈W,q∈Q L2(w, q) + C4(G(Hm, m) + e−2λ0H) ≤ 8 min w∈W,q∈Q L2(w, q) + C ln3 m m , since Hm ≥ M0e−λ0H implies that e−2λ0H ≤ H2 m M0 ≤ e M0 ln2 m m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When Hm < M0e−λ0H, E(| ˆRm − Rπe|2) ≤ 8 min w∈W,q∈Q L2(w, q) + C4(G(M0e−λ0H, m) + e−2λ0H) = 8 min w∈W,q∈Q L2(w, q) + C4 � (1 + M2 0 )e−2λ0H + (1 − ln M0 + λ0H + (ln M0 − λ0H)2)ln m m � ≤ 8 min w∈W,q∈Q L2(w, q) + C � e−2λ0H + H2 ln m m � , for some constant C independent of m, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ From Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1, it is easy to get Therorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We briefly state the proof as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Proof of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When Qπe ̸∈ Q, (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4) does not hold but can be adjusted as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Since I2 1 = L2( ˆwm, Qπe) = (L( ˆwm, Qπe − q) + L( ˆwm, q))2 ≤ 2(L2( ˆwm, q) + L2( ˆwm, Qπe − q)), for any q ∈ Q, we have that, I2 1 ≤ 2 max q∈Q L2( ˆwm, q) + 2 min q∈Q L2( ˆwm, Qπe − q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' ≤ 2 max q∈Q (L( ˆwm, q) − ˆLm( ˆwm, q) + ˆLm( ˆwm, q))2 + 2 max w∈W min q∈Q L2(w, Qπe − q) ≤ 4 max q∈Q (L( ˆwm, q) − ˆLm( ˆwm, q))2 + 4 max q∈Q ˆL2 m(w, q) + 2 max w∈W min q∈Q L2(w, Qπe − q), 32 for any w ∈ W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Consequently,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' I2 1 ≤ 4 max w∈W,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='q∈Q(L( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) − ˆLm( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q))2 + 8 max q∈Q (ˆLm(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) − L(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q))2 +8 max q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) + 2 max w∈W min q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Qπe − q) ≤ 4 max w∈W,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='q∈Q(L( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) − ˆLm( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q))2 + 8 max w∈W,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='q∈Q(ˆLm(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) − L(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q))2 +8 min w∈W max q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) + 2 max w∈W min q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Qπe − q) = 12 max w∈W,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='q∈Q(L( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) − ˆLm( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q))2 + 8 min w∈W max q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) + 2 max w∈W min q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Qπe − q),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' and therefore E(I2 1) ≤ 12E( max w∈W,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='q∈Q(L( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) − ˆLm( ˆwm,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q))2) +8 min w∈W max q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' q) + 2 max w∈W min q∈Q L2(w,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Qπe − q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Using this inequality to replace (B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='4) and then repeating the discussion in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1, we can readily get the desired result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' □ C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Algorithm Supplement In the algorithm, we give the following notation: G = � ��� 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 0 0 � ��� nh×nh , Frequency = [0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , 0]⊤ nh, auxi = [0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , 0]⊤ nh, ˆµ = [0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' , 0]⊤ nh, X represents absorbing state set, Y = {h × i + j, i ∈ X, j ∈ A}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The Algorithm 1 summarizes the pseudo-codes of our MWLA algorithm applied to the taxi environment in Section 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' The principle of the algorithm is explained in Example 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' Here, for the convenience of computations, we set w(s, a) = 1⊤ {s,a}u ˆdπb(s, a) and q(s, a) = 1⊤ {s,a}v, ∀s ∈ S0, a ∈ A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' We also introduce a regularization factor λ > 0 which helps us find the unique solution of the constrained quadratic programming problem arg min u≥0 ∥ ( ˆG + λI)⊤u + ˆb ∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' When λ is sufficient small, the solution is an approximation of −( ˆG+)⊤b where ˆG+ is the Moore-Penrose pseudo-inverse of ˆG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' In our experiments, λ is set to be 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 33 Algorithm 1 Tabular case Input: Off-policy data D = {si 0, ai 0, ri 0, si 1, · · · , si Ti∧H−1, ai Ti∧H−1, ri Ti∧H−1, si Ti∧H}m i=1 from the behavior policy πb;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' a target policy π for which we want to estimate the expected return.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 1: Estimate the initial state distribution ˆµ(s) = 1 m m � i=1 1{si 0=s}, where 1{·} is an indicative function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 2: for episode in D do 3: for s,a,s ′,r in episode do 4: cur = h × s + a, 5: G[cur, h × s ′ : h × (s ′ + 1)]+ = π[s ′, :], 6: G[cur, cur]− = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content='0, 7: Frequency[cur]+ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' end for 8: auxi = � s,a ˆµ(s)π(a|s)1(s, a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 9: tvalid = where(Frequency > 0) indicates the index of an element whose value is greater than 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 10: ˆdπb = delete(Frequency, Y, 0), delete the row corresponding to the absorbing state from Frequency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 11: tvalid1 = where( ˆdπb > 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 12: ˆdπb = ˆdπb/m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 13: ˆG = delete(G, Y, 0), delete the row corresponding to the absorbing state from G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 14: ˆG = delete( ˆG, Y, 1), delete the column corresponding to the absorbing state from ˆG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 15: ˆG[:, tvalid1] = ˆG[:, tvalid1]/(m × ˆdπb[tvalid1]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 16: ˆb = delete(auxi, Y, 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 17: Compute ˆu = arg min u≥0 ∥ ( ˆG + λI)⊤u + ˆb ∥2, where ˆG = 1 m m � i=1 Ti∧H−1 � t=0 1(si t,ai t) � � a∈A π(a|si t+1)1⊤ (si t+1,a) − 1⊤ (si t,ai t) � ˆdπb(si t, ai t) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' ˆb = � (s,a)∈S0×A ˆµ(s)π(a|s)1(s,a), λ is a regularization factor, I denotes an identity matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 18: Parameterize w(tvalid) = ˆu(tvalid1) ˆdπb(tvalid1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 19: for episode in D do 20: for s, a, s ′, r in episode do 21: cur = h × s + a, 22: ˆRm = 1 m m � i=1 Ti∧H−1 � t=0 w(cur) × r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' end for Output: ˆRm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
+page_content=' 34' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/Y9E1T4oBgHgl3EQfcQTz/content/2301.03183v1.pdf'}
diff --git a/YdE0T4oBgHgl3EQfmwE9/content/tmp_files/2301.02502v1.pdf.txt b/YdE0T4oBgHgl3EQfmwE9/content/tmp_files/2301.02502v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..87a196b77d44e1664d6a7dba8bec81e07f25bf9f
--- /dev/null
+++ b/YdE0T4oBgHgl3EQfmwE9/content/tmp_files/2301.02502v1.pdf.txt
@@ -0,0 +1,3380 @@
+Stacking-dependent topological magnons in bilayer CrI3
+M. Soenen,1 C. Bacaksiz,1, 2, 3 R. M. Menezes,1 and M. V. Miloˇsevi´c1, ∗
+1Department of Physics & NANOlab Center of Excellence,
+University of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium
+2Bremen Center for Computational Material Science (BCCMS), Bremen D-28359, Germany
+3Computational Biotechnology, RWTH Aachen University, Worringerweg 3, 52074 Aachen, Germany
+(Dated: January 9, 2023)
+Motivated by the potential of atomically-thin magnets towards tunable high-frequency magnonics,
+we detail the spin-wave dispersion of bilayer CrI3. We demonstrate that the magnonic behavior of the
+bilayer strongly depends on its stacking configuration and the interlayer magnetic ordering, where
+a topological bandgap opens in the dispersion caused by the Dzyaloshinskii-Moriya and Kitaev
+interactions, classifying bilayer CrI3 as a topological magnon insulator. We further reveal that both
+size and topology of the bandgap in a CrI3 bilayer with an antiferromagnetic interlayer ordering are
+tunable by an external magnetic field.
+I.
+INTRODUCTION
+Emergent two-dimensional (2D) magnetic materials1
+provide an exciting platform to study collective spin ex-
+citations, i.e. magnons. CrI3, the archetypal 2D van der
+Waals (vdW) ferromagnet,2 has recently been suggested
+to host magnon modes in the highly sought-after tera-
+hertz (THz) regime,3–5 showing promise for the develop-
+ment of faster and more energy-efficient data processing
+applications.6 Moreover, due to the 2D nature of the ma-
+terial, its spin-wave properties are highly susceptible to
+tuning, e.g. by strain, buckling, defect-engineering, gat-
+ing and/or vdW heterostructuring.5
+Recently, several bulk materials, including CrBr3,7
+CrI3,8,9 CrSiTe310 and CrGeTe3,10 have been identified
+as topological magnon insulators (TMIs), characterized
+by bulk magnon bands with a gap at the Dirac point,
+and topologically protected edge states. The magnonic
+bandgap is attributed to the anti-symmetric exchange in-
+teraction – more often called the Dzyaloshinskii-Moriya
+interaction (DMI)11,12 – arising from the lack of inver-
+sion symmetry between next-nearest-neighboring (NNN)
+Cr atoms. In contrast, bulk CrCl3,13,14 where the DMI is
+weak, is classified as a magnon Dirac material (MDM),
+characterized by a Dirac-point in the dispersion, showing
+a linear band crossing at the Brillouin zone edge.
+However, it remains an open question whether the
+topological features of aforementioned materials will per-
+sist down to the monolayer limit. Early theoretical work
+suggested that honeycomb ferromagnetic (FM) mono-
+layers could be classified as either MDMs or TMIs de-
+pending on whether any NNN DMI is present in the
+material.15–20 Nonetheless, recent work identified Ki-
+taev interactions as an alternative mechanism potentially
+able to open a topological bandgap in FM honeycomb
+materials,21,22 suggesting that the absence of DMI is not
+the sole criterion for predicting the topological proper-
+ties of such materials.
+Similarly, in magnetic honey-
+comb bilayers, a DMI-induced topological behavior of
+magnons is predicted,20,23,24 including the formation of
+Dirac magnon nodal-line loops,23 and the opening of
+a topological bandgap, which contributes to a magnon
+Hall- and a spin Nernst effect.20,24
+A first attempt to characterize the magnonics of mono-
+layer CrI3, using an itinerant fermion description based
+on ab initio calculations, showed the appearance of a
+small, possibly topological, bandgap caused by the spin-
+orbit coupling (SOC),25 suggesting that the material is
+a TMI. However, more work is required before full un-
+derstanding of the magnonics in CrI3 is achieved. In this
+work, we deploy a multi-scale approach, combining ab
+initio calculations with numerical simulations based on
+a Heisenberg model and linear spin-wave theory, to char-
+acterize the magnonic properties of CrI3 monolayers and
+bilayers, to reveal the topological magnon modes present
+in these systems, and that (topological) magnonic prop-
+erties of the bilayer are strongly affected by its stacking
+order and its interlayer magnetic ordering.
+The article is organized as follows. In section II, we de-
+scribe the computational methodology used in this work.
+We discuss the Heisenberg Hamiltonian that models the
+magnetic interactions in CrI3, explain how the parame-
+ters that characterize this Hamiltonian will be derived
+from first-principles and, finally, sketch how the spin-
+wave dispersion is computed.
+Subsequently, in section
+III, we apply this methodology to monolayer CrI3, con-
+firming the presence of a small topological bandgap with
+non-zero Chern numbers in the material’s spin-wave dis-
+persion. Afterwards, in section IV, we consider bilayer
+CrI3 in three different stacking orders, each exhibiting
+significantly different magnonic behavior.
+Specifically,
+we investigate the AA-stacking and AB-stacking (rhom-
+bohedral) discussed in literature,20,23,24 as well as the
+experimentally very relevant AB’-stacking (monoclinic)
+of which the spin-waves have - to the best of our knowl-
+edge - not been theoretically investigated to date. We
+find that all three stacking versions of bilayer CrI3 ex-
+hibit either FM or antiferromagnetic (AFM) interlayer
+ordering, with intralayer ferromagnetism.
+In case of a
+FM interlayer ordering, we observe a bandgap in the
+spin-wave dispersion with stacking-dependent topologi-
+cal properties.
+We attribute the origin of the gap to
+a combination of DMI and Kitaev interactions that are
+arXiv:2301.02502v1 [cond-mat.mtrl-sci] 6 Jan 2023
+
+2
+modulated by the stacking order. Furthermore, we show
+a significant influence of the interlayer magnetic order-
+ing on the resulting magnonic behavior. Specifically, the
+topological nature of the bands becomes trivial in AFM-
+ordered bilayers. Additionally, we show that magnonic
+dispersion of AFM-ordered bilayers is susceptible to tun-
+ing by an external magnetic field, lifting the degener-
+acy between branches, which decreases the size of the
+magnonic bandgap and leads to a non-trivial topology of
+the bands in the AB’-stacking, or introduces nodal-line
+loops in the AA-stacking case. Finally, section V sum-
+marizes our findings and gives an outlook on some future
+challenges and opportunities within the field.
+II.
+COMPUTATIONAL METHODOLOGY
+We model the magnetic interactions of the system un-
+der study using a Heisenberg spin Hamiltonian of the
+following form:
+ˆH = 1
+2
+�
+i,j
+ˆSiJijˆSj +
+�
+i
+ˆSiAiiˆSi + µB
+�
+i
+B · giˆSi, (1)
+in which the spins are three-dimensional (3D) vectors
+ˆSi = ( ˆSx
+i , ˆSy
+i , ˆSz
+i ) expressed in Cartesian coordinates.
+The first- and second term of this Hamiltonian respec-
+tively describe the exchange interaction and the single
+ion anisotropy (SIA), which are characterized by the
+3 × 3 matrices Jij and Aii. The DMI is characterized
+by a vector Dij with components that can be calcu-
+lated from the off-diagonal elements of the exchange ma-
+trix as Dx
+ij = 1
+2(J yz
+ij − J zy
+ij ), Dy
+ij = 1
+2(J zx
+ij − J xz
+ij ) and
+Dz
+ij = 1
+2(J xy
+ij −J yx
+ij ).26,27 Notice that Dij = νij|Dij| with
+νij = −νji = ±1, where the sign of the latter depends on
+the hopping direction of the considered spin pair. The
+exchange term is now written as:
+ˆHex = 1
+2
+�
+i,j
+� �
+α′
+Jα′
+ij ˆSα′
+i ˆSα′
+j + Dij(ˆSi × ˆSj)
+�
+,
+(2)
+with α′ = {α, β, γ} the local eigenbases that diagonalize
+the symmetric part of the exchange matrices. To consider
+the exchange anisotropy, we define the Kitaev constant
+as Kij = Jγ
+ij − Jij with Jij = (Jα
+ij + Jβ
+ij)/2 the isotropic
+exchange constant,28 leading to the following expression
+for the exchange Hamiltonian:
+ˆHex = 1
+2
+�
+i,j
+�
+JijˆSi · ˆSj + Kij ˆSγ
+i ˆSγ
+j + Dij(ˆSi × ˆSj)
+�
+.
+The symmetric SIA-matrix Aii accounts for the interac-
+tion of the magnetic orbitals with the surrounding crystal
+field and contributes to the magnetic anisotropy of the
+material. In crystals with a 3-, 4-, or 6-fold rotational
+symmetry around the out-of-plane axis, most elements
+of the matrix are redundant and the SIA can be char-
+acterized by a single parameter Azz
+ii instead of the full
+SIA-matrix.26,27 The last term of equation (1) accounts
+for the Zeeman interaction when applying an external
+magnetic field B, where gi ≈ 2 is the g-factor, and µB is
+the Bohr magneton. In CrI3, the magnetic dipole-dipole
+interaction is expected to be small in comparison to its
+out-of-plane magnetic anisotropy and will, therefore, not
+be included in the Heisenberg Hamiltonian.5 Finally, also
+notice that CrI3 has a magnetic moment of µ = 3µB per
+chromium atom and, thus, a spin of S = 3/2.
+To obtain the elements of the exchange- and SIA
+matrices, we use the four-state energy mapping (4SM)
+methodology26,27 in which we calculate the energies of
+several spin configurations of the system from first prin-
+ciples using density functional theory (DFT), and map
+these energies on their corresponding Heisenberg Hamil-
+tonians, setting up a system of equations from which the
+magnetic parameters can be derived. The implementa-
+tion of the needed DFT calculations is thoroughly dis-
+cussed in the supplementary material.29
+The spin-wave dispersion relations are calculated nu-
+merically using the open-source code spinW ,39 in which
+we have implemented our Heisenberg Hamiltonian. This
+code is based on linear spin-wave theory, which is a
+good approximation assuming spin fluctuations are small.
+This condition is comfortably satisfied at low tempera-
+tures, significantly below the critical temperature (Curie
+or N´eel) of the long-range magnetic order at hand. Nu-
+merical diagonalization of the Heisenberg Hamiltonian in
+reciprocal space yields the spin-wave dispersion.
+III.
+MONOLAYER
+A.
+Crystal structure and magnetic parameters
+The crystal structure of monolayer CrI3 is depicted in
+Figure 1(a,e). Monolayer CrI3 comprises one honeycomb
+layer of chromium atoms sandwiched between two layers
+of iodine atoms, where each chromium atom is octahe-
+drally coordinated with six iodine atoms, and each iodine
+atom connects two chromium atoms through an ≈ 90◦
+Cr-I-Cr bond. After structural relaxation using DFT, we
+find a in-plane lattice constant of a = 6.919 ˚A.
+To characterize the magnetic interactions in CrI3, we
+perform a 4SM analysis in order to obtain the ele-
+ments of the exchange and SIA matrices.
+In Table I,
+we report the average nearest-neighbor (NN) and next-
+nearest-neighbor (NNN) intralayer exchange, Kitaev and
+DMI parameters for monolayer CrI3. A full summary of
+the exchange parameters of all the individual pairs can
+be found in the supplementary material.29
+From the calculated parameters, it becomes clear that
+both the NN and the NNN exchange interactions are
+anisotropic and FM, with the NN one delivering the dom-
+inant contribution. In agreement with literature,41,42 we
+find that the material’s out-of-plane magnetic anisotropy
+originates mainly from the NN exchange anisotropy, with
+a smaller contribution of ⟨Azz
+ii ⟩ = -0.08 meV due to the
+
+3
+FIG. 1:
+Crystal structure of monolayer and bilayer CrI3. Top view (a-d) and side view (e-h) of monolayer (1L) CrI3
+(a,e), and bilayer (2L) CrI3 with an AB-stacking (b,f), AB’-stacking (c,g) and AA-stacking (d,h). For the sake of clarity, atoms
+of the same type are assigned a different color in the top and the bottom layer. In the bottom (top) layer, the chromium and
+iodine atoms are depicted with blue (dark blue) and yellow (orange) spheres respectively. The unit cell is marked with a solid
+black line. All crystal structures were plotted using VESTA.40 Panel (i) depicts the corresponding first Brillouin zone and
+high-symmetry points for 2D systems with a hexagonal lattice.
+SIA. The SIA is characterized by a single parameter ow-
+ing to the material’s three-fold rotational symmetry. The
+NN interactions deliver no net contribution to the DMI
+since the inversion symmetry of the material is upheld.
+However, this symmetry is not present between NNN
+sites, resulting in a small yet non-zero DMI. Notice that,
+in CrI3, the DMI, the Kitaev interaction and the SIA all
+originate from the large SOC arising due to the heavy I
+ligands.28,41,42
+B.
+Spin-wave dispersion
+Figure 2(a) depicts the spin-wave dispersion of mono-
+layer CrI3 along the high-symmetry directions of the first
+Brillouin zone. Two distinct branches are present, as is
+expected for a unit cell containing two magnetic atoms.
+At the Γ-point, the dispersion is gapped below the lower
+branch due to the magnetic anisotropy of the material.
+The gap has a size of ∆Γ = 0.41 meV and is an essen-
+tial prerequisite for the existence of long-range magnetic
+order in 2D at finite temperatures.41 The latter can be
+TABLE I:
+Magnetic parameters for monolayer CrI3.
+Summary of the most important magnetic parameters in
+monolayer CrI3, including the exchange- and Kitaev constants
+Jij and Kij, and the size of the DMI-vectors |Dij|.
+JNN
+KNN
+|DNN|
+JNNN
+KNNN
+|DNNN|
+(meV)
+(meV)
+(meV)
+(meV)
+(meV)
+(meV)
+-4.35
+1.49
+0.00
+-0.74
+0.17
+0.06
+seen by considering the total number of magnons excited
+at temperature T, which is given by:
+N =
+�
+D(ωk)
+e¯hωk/kBT − 1dωk,
+(3)
+with D(ωk) the magnon density of states, which is con-
+stant in 2D, kB the Boltzmann constant, and ωk the
+spin-wave frequencies. When the dispersion is gapless,
+i.e. in the absence of magnetic anisotropy, this integral
+will diverge for ωk = 0, preventing long-range 2D mag-
+netic order at non-zero temperature in accordance with
+the Mermin-Wagner theorem.43
+The lower energy ‘acoustic’ branch displays quadratic
+behavior near the Γ-point and is associated with an in-
+phase precession of the spins [see Figure 2(b)], while the
+higher energy ‘optical’ branch is associated with an out-
+of-phase precession of the spins [see Figure 2(c)]. The
+two branches meet at the K-point where they are sep-
+arated by a small bandgap of ∆K = 0.15 meV. At the
+K’-point we find a gap of the same size, since the sub-
+lattice symmetry is upheld. The origin of this Dirac gap
+is partially attributed to the NNN DMI and partially to
+the Kitaev interactions.
+At the K-point, the spins will precess at 120◦ angles to
+each other, as is shown in Figure 2(d,e) for respectively
+the lower- and higher branch. If one would only consider
+a purely isotropic NN exchange, these two states would
+be energetically degenerate resulting in a Dirac point.
+However, introducing Kitaev interactions and/or a NNN
+exchange term with a non-zero DMI, lifts the mutual
+degeneracy between the modes resulting in a bandgap.
+More specifically, it’s the out-of-plane component of
+
+1L
+AB-2L
+AB'-2L
+AA-2L
+(a)
+(b)
+(c)
+(d)
+(h)
+(i)
+b'
+K
+ki4
+(a)
+(b)
+(c)
+(d)
+(e)
+FIG. 2:
+Spin-wave dispersion of monolayer CrI3. (a) Spin-wave dispersion along the high-symmetry directions of the
+first Brillouin zone. A small Dirac gap of ∆K = 0.15 meV is present at the K-point. Corresponding Chern numbers are
+indicated for each band. (b) and (c) display the spin-wave modes at the Γ-point for the lower- and higher branch respectively.
+(d) and (e) display a schematic top-view of the spin-wave modes at the K-point for the lower- and higher branch respectively.
+the DMI that lies at the origin of the magnonic bandgap.
+The size of this component intrinsically present in CrI3
+is rather small (|Dz
+ij| = 0.03 meV), also resulting in a
+small bandgap.
+However, external tuning that breaks
+the inversion symmetry, e.g. the presence of a substrate,
+electric gating or (non-uniform) strain, might induce ad-
+ditional DMI that could potentially increase the size of
+the bandgap. In fact, by (artificially) increasing the DMI
+in our simulations, we verified that the bandgap can
+be ’tuned’. As shown in the supplementary material,29
+the bandgap scales almost linearly with the NNN DMI
+when all the other parameters are kept constant. Tun-
+ing the magnonic bandgap in 2D materials under exter-
+nal stimuli poses an interesting direction for future re-
+search, as the size of the bandgap can influence other
+material properties like as the magnon Hall conductiv-
+ity.
+However, note that increasing the DMI may lead
+to non-collinear magnetization textures, e.g.
+spin cy-
+cloids or magnetic skyrmions, which will fundamentally
+change the magnonic behavior in the material.44,45 More-
+over, when the DMI is set to zero in our calculations, the
+bandgap does not fully vanish, suggesting that there is
+a second mechanism at work, which we identify to be
+the Kitaev interaction between NN spins.
+In the sup-
+plementary material,29 we show that one can tune the
+bandgap by artificially changing Kij. However, varying
+the strength of the Kitaev interaction influences the over-
+all shape of the dispersion, whereas changing the DMI
+mainly influences the dispersion around the K-point.
+C.
+Topology
+Non-trivial band topology arises only in systems where
+non-zero Chern numbers predict the existence of edge
+states. The Chern number is a topological invariant with
+an integer value that is defined for the nth band as:
+Cn =
+1
+2πi
+�
+BZ
+Ωnk d2k,
+(4)
+in which the Berry curvature can be calculated as
+Ωnk = i
+�
+n′̸=n
+⟨n |∂k ˆHk |n′ ⟩⟨n′ |∂k ˆHk |n⟩
+(λnk − λn′k)2
+,
+(5)
+with λnk and |n⟩ respectively the eigenvalues and eigen-
+vectors of the Heisenberg Hamiltonian ˆHk in reciprocal
+space.
+For systems that are gapless or show a trivial
+bandgap, the Chern numbers vanish. In this work, we
+calculate Chern numbers according to the link-variable
+method detailed in Ref. [46] for a discretized Brillouin
+zone. Applying this approach to the magnonic dispersion
+of monolayer CrI3, we find Chern numbers of Cn = ±1
+for respectively the upper and lower band, as shown in
+Figure 2, classifying the material as a TMI with a non-
+trivial topological bandgap. We attribute the origin of
+the topology to the breaking of time-reversal symmetry
+due to the spontaneous magnetization of CrI3.47 Thus,
+the topological nature of the bands persists in monolayer
+CrI3, be it with a significantly smaller bandgap compared
+to bulk CrI3.8
+IV.
+BILAYER
+A.
+Crystal structure and magnetic parameters
+Bilayer CrI3 can be constructed by stacking two mono-
+layers on top of each other in a commensurate manner.
+The three different stacking orders that we consider in
+this work are shown in Figure 1. In analogy to Sivadas
+et al.,48 we refer to those stacking orders as AB (rhom-
+bohedral), AB’ (monoclinic), and AA. The former two
+stackings correspond to respectively the low-temperature
+
+20
+一
+1
+1
+-
+1
+Ci = +l
+15
+-
+1
+(meV)
+1
+1
+K
+10
+1
+C2
+1
+E
+-
+1
+1
+5
+1
+1
+一
+1
+1
+一
+1
+0
+I
+K
+M
+TB
+A
+A
+B
+B
+A
+A
+B
+B
+AB
+A
+A
+B
+B
+A
+A
+B
+B
+A5
+TABLE II:
+Structural and magnetic parameters for bilayer CrI3. Summary of the most important stucutural and
+magnetic parameters in bilayer CrI3, including the lattice constant a and interlayer distance d, the average exchange- and
+Kitaev constants ⟨Jij⟩ and ⟨Kij⟩, the average size of the DMI-vectors ⟨|Dij|⟩, the DFT energy difference between the bilayer
+with an AFM and a FM interlayer ordering, and the average SIA.
+a
+d
+⟨JNN⟩
+⟨KNN⟩
+⟨|DNN|⟩
+⟨JNNN⟩
+⟨KNNN⟩
+⟨|DNNN|⟩
+EAFM − EFM
+⟨Azz
+ii ⟩
+(˚A)
+(˚A)
+(meV)
+(meV)
+(meV)
+(meV)
+(meV)
+(meV)
+(meV)
+(meV)
+AB
+6.915
+3.400
+-4.49
+1.45
+0.07
+-0.62
+0.13
+0.03
+12.13
+-0.07
+AB’
+6.914
+3.430
+-4.49
+1.45
+0.07
+-0.64
+0.15
+0.02
+-0.06
+-0.07
+AA
+6.908
+3.505
+-4.42
+1.44
+0.07
+-0.65
+0.15
+0.03
+0.84
+-0.08
+and the high-temperature phases of CrI3.49 In the AB-
+stacking, the layers are stacked in such a way to place
+the chromium atoms in one layer above the void in the
+chromium honeycomb of the adjacent layer, analogously
+to a Bernal-stacked graphene bilayer [Figure 1(b,f)]. The
+AB-stacking can be transformed to an AB’-stacking by
+sliding one of the layers by a third of the lattice vector
+along the zigzag direction [Figure 1(c,g)]. Alternatively,
+by sliding one of the AB-stacked layers by a third of the
+lattice vector along the armchair direction, we obtain an
+AA-stacked bilayer in which each atom in the top layer is
+placed exactly above its bottom layer counterpart [Fig-
+ure 1(d,h)].
+As shown in Table II, the different stacking orders
+show relatively similar lattice constants and interlayer
+distances. However, changes in interatomic distances and
+(super-)superexchange bonding angles result in a differ-
+ent interlayer magnetic coupling, such that the AB and
+AA stackings prefer a FM ordering between the layers
+while the AB’-stacking slightly favors an AFM one. The
+latter is indicated in Table II by the DFT energy differ-
+ence between AFM and FM phases. In agreement with
+earlier work,48,50–52 we find that the overall ground-state
+of the system is a FM-ordered AB-stacked bilayer. In
+the supplementary material,29 we discuss the stacking-
+dependence of the interlayer ordering in more detail.
+In Table II, we also summarize the predominant mag-
+netic parameters for the CrI3 bilayers calculated with the
+4SM method. A full overview of all the calculated param-
+eters for each specific pair can be found in the supple-
+mentary material.29 For all stacking orders, the NN in-
+tralayer exchange interaction is anisotropic and strongly
+FM. This anisotropy, together with the SIA, causes the
+spins to prefer an out-of-plane orientation. Due to the
+rotational symmetry in the AB and AA-stackings, the
+SIA matrix is reduced to only one parameter ⟨Azz
+ii ⟩. In
+the AB’-stacked bilayer, this symmetry is absent requir-
+ing a calculation of the full SIA-matrix, however, ⟨Azz
+ii ⟩
+will still be the dominant parameter, as most of the other
+matrix elements are very small or vanish.
+To quantify the interlayer coupling, we calculated the
+interlayer NN and NNN exchange matrices for all stack-
+ings. For the AB’-stacking, we also calculate the third
+nearest-neighbor (3NN) interlayer exchange, for the other
+stackings this contribution is negligible as is demon-
+strated in the supplementary material.29 In the AB-
+and AA-stacked bilayers, all NN and NNN interlayer ex-
+change interactions are FM. However, the exchange pa-
+rameters for the AB-stacking are significantly stronger
+than for the AA-stacking, resulting in a stronger prefer-
+ence for a FM ordering. In contrast, for the AB’-stacked
+bilayer, there is a competition between the NN exchange
+which is FM and the NNN and 3NN exchange interac-
+tions which are AFM. Overall, this results in a weak AFM
+interlayer ordering, which is in agreement with earlier
+theoretical and experimental studies.2,48,50–57
+Interestingly, the sublattice symmetry is broken in the
+AB- and AB’ stackings, leading to a difference in out-of-
+plane exchange interactions ∆Jzz = |Jzz
+A − Jzz
+B | between
+sublattices A and B of 0.92 meV for the AB-stacking and
+0.04 meV for the AB’-stacking. Note that Jzz
+A and Jzz
+B
+are the sum of the out-of-plane exchange components of
+all interacting spin pairs. The difference ∆Jzz is sub-
+stantial for the AB-stacking because one sublattice has
+six stronger NNN interactions while the other sublattice
+has one weaker NN coupling and only three NNN in-
+teractions.
+For the AA-stacking, there is no exchange
+difference since the sublattice symmetry is preserved.
+The intralayer Kitaev constants in the bilayers are sim-
+ilar in size compared to the monolayer. For the AB- and
+AB’-stacking, the NN Kitaev interaction is anisotropic,
+leading to different values for each bond, which is at-
+tributed to symmetry breaking due to the stacking. Since
+the NN Kitaev interaction is much stronger than the
+NNN and the interlayer ones, it’s the only contribution
+having a significant influence on the spin-wave dispersion.
+Unlike the monolayer system, the NN intralayer DMI is
+now non-zero, and originates from the inversion symme-
+try breaking due to stacking. Similarly to the monolayer
+case, a non-zero NNN DMI arises. In all stackings, the
+interlayer DMI will be very small or completely absent,
+having a limited influence on the dispersion.
+B.
+Spin-wave dispersion of bilayers with FM
+interlayer order
+Using the magnetic parameters calculated with the
+4SM method, we compute the spin-wave dispersion for
+the three stacking orders considered in this work. In Fig-
+ure 3, we display the results for bilayer CrI3 with different
+stackings, all with the FM interlayer ordering. In a bi-
+
+6
+(a) AB
+(b) AB’
+(c) AA
+(d)
+FIG. 3:
+Spin-wave dispersion for bilayer CrI3 with a FM interlayer ordering in different stacking configurations.
+(a,b,c) Spin-wave dispersion along the high-symmetry directions of the first Brillouin zone for respectively the AB, AB’, and AA-
+stacked bilayers with a FM interlayer ordering. In all stackings, a direct magnonic bandgap opens at the K-point. Corresponding
+Chern numbers are indicated for single bands, and composite Chern numbers for degenerate bands. (d) schematically displays
+the corresponding spin-wave modes at the Γ-point for each band.
+layer, the unit cell contains four magnetic atoms, leading
+to four branches in the dispersion, two ‘acoustic’ and two
+‘optical’ ones.
+The corresponding spin-wave modes at
+the Γ-point of each branch are indicated in Figure 3(d).
+The energy difference between these different modes is
+proportional to the strength of the interlayer coupling,
+hence the large separation for the AB-stacking.
+Each stacking has a gap at the Γ-point, signaling that
+FM order is stable in each of them at finite temperatures.
+The gaps have sizes of ∆Γ = 0.44 meV, ∆Γ = 0.34 meV
+and ∆Γ = 0.35 meV for respectively the AB, AB’, and
+AA-stackings.
+In all stackings, we also observe direct
+magnonic bandgaps at the K-point of ∆K = 0.57 meV,
+∆K = 0.21 meV, and ∆K = 0.08 meV for respectively
+the AB, AB’, and AA-stackings. Similarly to the mono-
+layer, we attribute the origin of these gaps to a combina-
+tion of NNN DMI and NN Kitaev interactions.
+Notice that the first Brillouin zone contains two in-
+equivalent high-symmetry points K and K’ [see Fig-
+ure 1(i)]. In the AB and AB’ stackings, where the sub-
+lattice symmetry is broken, we see a different behavior of
+the dispersion at each point. In the former, a bandgap of
+only 0.42 meV opens close to the K’-point (compared to
+∆K = 0.57 meV). In the AB’-stacking, we see an indirect
+band-crossing at the K’-point, as is often seen in semi-
+metals, and thus, no bandgap. On the other hand, in
+the AA-stacked bilayer, the sublattice symmetry is pre-
+served, resulting in exactly the same dispersion at both
+the K- and K’-points.
+C.
+Spin-wave dispersion of bilayers with AFM
+interlayer order
+By comparing the dispersion of the bilayers with FM
+interlayer ordering with the dispersions of the bilayers
+
+n= 2
+n=3
+n =
+n = 4
+B
+B
+B
+B
+B
+B20
+一
+-
+15
+neV)
+一
+-
+C3 = 0
+K
+10
+一
+E
+1
+1
+1
+5
+1
+一
+1
+1
+0
+I
+K
+M
+T20
+-
+1
+-
+1
+1
+15
+-
+-
+(meV)
+-
+1
+10
+K
+1
+E
+一
+1
+一
+1
+一
+1
+5
+1
+1
+1
+一
+1
+一
+1
+0
+1
+K
+M
+T20
+-
+iCi = +l
+一
+1
+一
+1
+一
+15
+I C2
+=-1
+1
+(meV)
+C3 =+1
+K
+10
+-
+E
+-
+1
+1
+1
+1
+1
+5
+一
+1
+-
+C4
+1
+1
+0
+1
+K
+M7
+(a) AB
+(b) AB’
+(c) AA
+(d)
+FIG. 4:
+Spin-wave dispersion for bilayer CrI3 with an AFM interlayer ordering in different stacking configura-
+tions. (a,b,c) Spin-wave dispersion along the high-symmetry directions of the first Brillouin zone for respectively the AB, AB’
+and AA-stacked bilayers with an AFM interlayer ordering. In the AB and AB’-stackings a direct bandgap opens at the K-point,
+meanwhile in the AA-stacked bilayer we observe a Dirac point. Corresponding composite Chern numbers are indicated for the
+bands and are all equal to zero. (d) schematically displays the corresponding spin-wave modes at the Γ-point for each band.
+with AFM interlayer ordering [see Figure 4], it becomes
+clear that there is a strong dependence of the magnonic
+properties of CrI3 on the interlayer ordering. First and
+foremost, notice that for the AA and AB stackings,
+there is a region close to the Γ-point where the acous-
+tic branches are zeroed. Consequently, there is no gap at
+the Γ-point and the integral in equation (3) will diverge,
+signaling that AFM order is unstable in these stackings
+at non-zero temperatures. However, in the AB’-stacking,
+there is a gap of ∆Γ = 0.30 meV, meaning that AFM
+order is stable in the monoclinic phase, which is in agree-
+ment with experimental observations.2,49,53–56 Further,
+also notice that, in contrast to the FM-ordered bilayers,
+we see a degeneracy of the two acoustic branches and
+the two optical branches. Only at the K-point there are
+notable energy differences between the bands.
+The dispersions of different bilayers with AFM inter-
+layer order are characterized by bandgaps of respectively
+∆K = 1.10 meV and ∆K′ = 1.03 meV for the AB-
+stacking, and ∆K = 0.49 meV and ∆K′ = 0.11 meV
+for the AB’-stacking. At the K and K’-points in the AA-
+stacking case, there is no bandgap, but instead one finds
+a Dirac cone combined with two anti-crossing branches.
+D.
+Topology
+When two or more bands are degenerate, crossing or
+touching, it is no longer possible to assign individual
+Chern numbers to each band. Instead, we define a com-
+posite Chern number Cn⊕n′, jointly shared by the degen-
+erate bands, and calculated as detailed in Ref. [58].
+As shown in Figure 3, there is a strong dependence
+of the Chern number on the stacking configuration in
+the FM-ordered bilayers. Although we are expecting a
+non-trivial topology of the bandgaps in the FM bilay-
+
+20
+1
+-
+一
+-
+一
+-
+一
+-
+15
+-
+-
+neV)
+-
+-
+10
+K
+E
+一
+一
+-
+C34 = 0
+1
+-
+1
+-
+5
+1
+-
+-
+1
+-
+1
+-
+1
+-
+0
+K
+M
+T20
+1
+1
+1
+15
+-
+-
+-
+neV)
+1
+1
+(m
+1
+10
+-
+-
+-
+E
+-
+34 = 0
+1
+1
+5
+1
+1
+1
+1
+一
+1
+0
+K
+M20
+-
+1
+1
+一
+-
+一
+-
+一
+-
+15
+一
+-
+-
+neV)
+-
+一
+-
+E
+K
+10
+-
+一
+-
+E
+1
+-
+1
+-
+1
+-
+5
+1
+-
+1
+1
+1
+-
+1
+1
+1
+1
+0
+K
+M
+Tn=2
+n=3
+n =
+n =4
+B
+B
+B
+B
+B
+B8
+ers, caused by the breaking of time-reversal symmetry
+due to the spontaneous magnetization, only the AA-
+stacking shows non-zero Chern numbers. Thus, the AA-
+stacked CrI3 bilayer can be classified as a TMI. In the
+AB’-stacking there is no bandgap at the K’-point, hence,
+the Chern number is undefined and the bands are not
+topological. In the AB-stacking, all Chern numbers are
+equal to zero, meaning that the bandgap is of trivial
+nature.
+We attribute the lack of topology in the lat-
+ter stacking to the exchange difference ∆Jzz, caused by
+the breaking of sublattice symmetry, which is very large
+for the AB-stacking. In the supplementary material,29
+we show that by artificially reducing ∆Jzz in our simula-
+tions, which also decreases the size of the bandgap at the
+K-point and the K’-point, we can induce a topological
+phase transition to a state with non-zero Chern numbers
+of C1⊕2 = +1, C3 = −1 and C4 = 0, which confirms
+the influence that sublattice symmetry can have on the
+topology of magnonic bands.20
+In bilayers with AFM interlayer order, the bands are
+two-by-two degenerate, meaning that one can only define
+composite Chern numbers. In the case of AA-stacking,
+there is no bandgap and, thus, the Chern number is unde-
+fined and the bands display no topological behavior. The
+composite Chern numbers for the other two stackings
+turn out to be zero for all considered bands, which can be
+related to the conservation of effective time-reversal sym-
+metry in AFM materials, as the layers are time-reversed
+copies of each other.47 However, in the next section, we
+will show that breaking this symmetry by an applied
+magnetic field leads to emergent topological states with
+non-zero Chern numbers.
+E.
+Effect of an external magnetic field
+In this section, we explore whether the magnonic dis-
+persion and band topology of bilayer CrI3 can be tuned
+by applying an out-of-plane external magnetic field.
+In the case of a monolayer or the bilayers with FM
+interlayer order, there is only a trivial effect due to an
+applied magnetic field. Namely, the whole dispersion will
+uniformly shift up or down depending on the orientation
+of the applied field with respect to the magnetization.
+In contrast, for bilayers with AFM interlayer order, an
+external magnetic field will lift the degeneracy between
+branches, shifting two branches up and two branches
+down in energy, and leading to additional interesting
+features in the dispersion. Similar band shifts were ob-
+served by Cenker et al.,4 who reported the splitting of
+degenerate Raman peaks in the optical branches of an
+AFM-ordered monoclinic CrI3 bilayer, after applying an
+external magnetic field. However, note that applying a
+magnetic field to AFM-ordered bilayers should be done
+carefully, as the interlayer magnetic state will switch to
+the FM one for sufficiently strong fields. Here, we cal-
+culate the spin-wave dispersion for the different stack-
+ing scenarios under sufficiently small applied field, where
+FIG. 5:
+Spin-wave dispersion for an AB’-stacked CrI3
+bilayer with an AFM interlayer ordering under the
+influence of an external magnetic field. Left and right
+panels compare the dispersion near the K-point and K’-point
+respectively, for an applied magnetic field of B = 1.9 T. Under
+influence of the magnetic field, blue bands have shifted up and
+red bands down in energy.
+AFM interlayer order is safely stable (see supplementary
+material).29 Especially the AFM-AB phase is very sensi-
+tive, and changes to a FM interlayer order even for very
+weak applied field - hence is excluded from our calcula-
+tions in this section.
+In the case of AB’ stacking, the size of the bandgap
+will decrease after applying the magnetic field, reaching
+a minimum of 0.14 meV at the K-point and 0.02 meV
+at the K’-point for a field of B = 1.9 T, as shown in
+Figure 5. At the K’-point, one sees that, as the bands
+approach each other, they do not entirely touch or cross -
+instead we observe band inversion combined with a small
+bandgap. Band inversion is an effect often also present in
+electronic topological insulators,59 and is typically caused
+by the SOC. For fields applied in the opposite direction,
+we see analogous behavior, as now the two other bands
+are shifted upwards and the previous two downwards. For
+fields larger than B = 1.9 T, the AFM interlayer ordering
+changes to a FM one (see supplementary material).29
+In Figure 6, we show the influence of an external mag-
+netic field with a magnitude of B = 1.6 T on the disper-
+sion of an AFM-ordered AA-stacked CrI3 bilayer. Apply-
+ing the field shifts the Dirac node upwards or downwards
+depending on the polarity of the field, which leads to the
+formation of a closed Dirac magnon nodal-line loop at the
+crossover point of the red and blue bands in Figure 6.
+The latter is a closed one-dimensional loop around the
+K-point where two bands cross, exactly analogous to the
+nodal-lines described for Dirac semimetals.60 Decreasing
+the field leads to a smaller shift of the branches, result-
+ing in nodal-line loops with a smaller radius. For fields
+larger than 1.6 T, the AFM interlayer ordering changes
+to a FM one (see supplementary material).29
+As mentioned earlier, in the absence of applied field,
+all AFM bands show composite Chern numbers equal to
+zero, meaning that the bandgaps have a trivial topol-
+
+16
+-
+-
+15
+-
+14
+(meV)
+13
+E
+12
+-
+11
+-
+-
+10
+K16
+15
+-
+14
+(meV)
+W
+13
+K
+E
+12
+-
+-
+11
+-
+-
+10
+1
+K'9
+B = 0.0 T
+B = 1.6 T
+FIG. 6:
+Spin-wave dispersion for an AA-stacked CrI3
+bilayer with an AFM interlayer ordering under the
+influence of an external magnetic field. Left and right
+panels compare the dispersion at the K-point for applied fields
+of respectively B = 0 T and B = 1.6 T. Under influence of
+the magnetic field, blue bands have shifted up and red bands
+down in energy, forming magnon nodal-lines at the crossing
+points of the red and blue curves.
+ogy. Interestingly, after applying the magnetic field on
+the AB’-stacked bilayer, non-zero Chern numbers emerge
+as C1,4 = +1 and C2,3 = −1. In other words, by applying
+a magnetic field, which breaks the effective time-reversal
+symmetry of the material, a topological phase transition
+can be induced. In contrast, for the AA stacking, the
+(composite) Chern numbers remain undefined after ap-
+plying the magnetic field, as the Dirac cone stays present.
+V.
+CONCLUSIONS
+We characterized the magnonic dispersion for intrinsi-
+cally ferromagnetic monolayer and bilayer CrI3 using lin-
+ear spin-wave theory combined with a Heisenberg model
+parameterized from first principles. We showed that the
+monolayer is characterized by a small Dirac-gap in the
+spin-wave dispersion, sourced to a specific combination of
+next-nearest-neighbor (NNN) DMI and nearest-neighbor
+Kitaev interactions. Non-zero Chern numbers are asso-
+ciated with the bands, indicating the topological nature
+of the bandgap, and suggesting that monolayer CrI3 is
+a topological magnon insulator (TMI). In bilayer CrI3,
+still with dominantly ferromagnetic intralayer interac-
+tions, we demonstrated a dependence of the dispersion
+on the geometric stacking order and the interlayer mag-
+netic ordering, opening a bandgap for the AB stacking
+(for both FM and AFM interlayer order), the AB’ stack-
+ing (only AFM), and the AA stacking (only FM), mean-
+while the FM-ordered AB’ stacking shows an indirect
+band crossing, and the AFM-ordered AA stacking ex-
+hibits a Dirac point. Similarly to the monolayer case, we
+identified the DMI and Kitaev interactions as the lead-
+ing causes behind the opening of the bandgap, both being
+modulated by the stacking order. The latter contradicts
+earlier work on bulk CrI3 which claimed that only the
+NNN DMI and, thus, not the Kitaev interaction, lies at
+the origin of the Dirac gap.9 Interestingly, we found that
+the Chern number, and consequently the magnonic band
+topology, depends on the stacking configuration and the
+interlayer magnetic order, vanishing for all studied cases
+except in the FM-ordered AA bilayer. Thus, depending
+on the stacking order and the interlayer magnetic order,
+bilayer CrI3 is classified as either a topological magnon
+insulator, a trivial magnon insulator, or a magnon Dirac
+material. Finally, we showed that the dispersion of the
+bilayers with AFM interlayer order can be tuned by an
+external out-of-plane magnetic field, changing both size
+and topology of the bandgap for the AB’-stacked bilayer,
+and introducing closed nodal-line loops in the dispersion
+of the AA bilayer.
+The here demonstrated presence of tunable bandgaps
+of possibly topological nature in bilayer CrI3 recommend
+it as a TMI that can serve as a platform to investigate
+tunable magnon Hall- and spin Nernst effects in 2D. Our
+results could be verified experimentally by investigating
+the thermal magnon Hall effect in monolayer and bilayer
+CrI3. Both the DMI and Kitaev interactions originate
+from the spin-orbit coupling, which is relatively strong
+in CrI3 and, thus, lies at the origin of the topological
+bandgap.
+If one wants to achieve a gapless spin-wave
+dispersion, we suggest looking at 2D magnets with a
+weaker SOC, e.g. CrBr3 or CrCl3, which are good can-
+didates to host a Dirac point in the monolayer limit. In
+order to further tailor the magnonic bandgap, one can
+induce and tune the DMI in CrI3, or other 2D magnets,
+by external stimuli such as gating, (non-uniform) strain,
+heterostructuring, etc.
+Furthermore, our work demon-
+strates that stacking vdW monolayers, be it in regular
+bilayers or, in future work, multilayers, or even (moir´e)
+heterostructures, poses a viable route to achieve broadly
+tunable magnonic properties in 2D materials and van der
+Waals homo- and heterostructures.
+Acknowledgments
+We thank D. ˇSabani, B. Jorissen and M. Shafiei for use-
+ful discussions. This work was supported by the Research
+Foundation-Flanders (FWO-Vlaanderen) and the Spe-
+cial Research Funds of the University of Antwerp (BOF-
+UA). The computational resources used in this work were
+provided by the VSC (Flemish Supercomputer Center),
+funded by Research Foundation-Flanders (FWO) and the
+Flemish Government – department EWI.
+
+15
+1
+1
+1
+14.5
+1
+1
+1
+14
+1
+1
+1
+13.5
+1
+1
+13
+E
+1
+12.5
+1
+1
+1
+12
+1
+1
+1
+11.5
+1
+1
+1
+11
+K15
+14.5
+14
+13.5
+(meV)
+13
+E
+12.5
+12
+11.5
+1
+11
+K10
+∗ Electronic address: milorad.milosevic@uantwerpen.be
+1 M. Gibertini, M. Koperski, A. F. Morpurgo, and K. S.
+Novoselov, Nature Nanotechnol. 14, 408 (2019).
+2 B. Huang, G. Clark, E. Navarro-Moratalla, D. R. Klein,
+R. Cheng, K. L. Seyler, D. Zhong, E. Schmidgall, M. A.
+McGuire, D. H. Cobden, et al., Nature 546, 270 (2017).
+3 W. Jin, H. H. Kim, Z. Ye, S. Li, P. Rezaie, F. Diaz, S.
+Siddiq, E. Wauer, B. Yang, C. Li, et al., Nature Commun.
+9, 5122 (2018).
+4 J. Cenker, B. Huang, N. Suri, P. Thijssen, A. Miller, T.
+Song, T. Taniguchi, K. Watanabe, M. A. McGuire, D.
+Xiao, and X. Xu, Nature Phys. 17, 20–25 (2021).
+5 R. M. Menezes, D. ˇSabani, C. Bacaksiz, C. C. de Souza
+Silva, and M. V. Miloˇsevi´c, 2D Mater. 9, 025021 (2022).
+6 A. Barman, G. Gubbiotti, S. Ladak, A. O. Adeyeye, M.
+Krawczyk, J. Gr¨afe, C. Adelmann, S. Cotofana, A. Naeemi,
+V. I. Vasyuchka, et al., J. Phys.:
+Condens. Matter 33
+413001 (2021).
+7 Z. Cai, S. Bao, Z.-L. Gu, Y.-P. Gao, Z. Ma, Y. Shangguan,
+W. Si, Z.-Y. Dong, W. Wang, Y. Wu et al., Phys. Rev. B
+104, L020402 (2021).
+8 L. Chen, J.-H. Chung, B. Gao, T. Chen, M. B. Stone, A. I.
+Kolesnikov, Q. Huang, and P. Dai, Phys. Rev. X 8, 041028
+(2018).
+9 L. Chen, J.-H. Chung, M. B. Stone, A. I. Kolesnikov, B.
+Winn, V. O. Garlea, D. L. Abernathy, B. Gao, M. Au-
+gustin, E. J. G. Santos, and P. Dai, Phys. Rev. X 11,
+031047 (2021).
+10 F. Zhu, L. Zhang, X. Wang, F. J. dos Santos, J. Song,
+T. Mueller, K. Schmalzl, W. F. Schmidt, A. Ivanov, J. T.
+Park et al., Sci. Adv. 7, eabi7532 (2021).
+11 I. E. Dzialoshinskii, Sov. Phys. JETP 5, 1259 (1957).
+12 T. Moriya, Phys. Rev. 120, 91 (1960).
+13 L. Chen, M. B. Stone, A. I. Kolesnikov, B. Winn, W. Shon,
+P. Dai, and J.-H. Chung, 2D Mater. 9, 015006 (2022).
+14 J. A. Schneeloch, Y. Tao, Y. Cheng, L. Daemen, G. Xu, Q.
+Zhang, and D. Louca, npj Quantum Mater. 7, 66 (2022).
+15 S. A. Owerre, J. Phys.: Condens. Matter 28 386011 (2016).
+16 S. A. Owerre, J. Appl. Phys. 120, 043903 (2016).
+17 J. Fransson, A. M. Black-Schaffer, and A. V. Balatsky,
+Phys. Rev. B 94, 075401 (2016).
+18 S. A. Owerre, J. Phys. Commun. 1 025007 (2017).
+19 S. S. Pershoguba, S. Banerjee, J. C. Lashley, J. Park, H.
+˚Agren, G. Aeppli, and A. V. Balatsky, Phys. Rev. X 8,
+011010 (2018).
+20 H. Kim, and S. K. Kim, Phys. Rev. B 106, 104430 (2022).
+21 E. Aguilera, R. Jaeschke-Ubiergo, N. Vidal-Silva, L. E. F.
+F. Torres, and A. S. Nunez, Phys. Rev. B 102, 024409
+(2020).
+22 L.-C. Zhang, F. Zhu, D. Go, F. R. Lux, F. J. dos Santos,
+S. Lounis, Y. Su, S. Bl¨ugel, and Y. Mokrousov, Phys. Rev.
+B 103, 134414 (2021).
+23 S. A. Owerre, Sci. Rep. 7, 6931 (2017).
+24 S. A. Owerre, Phys. Rev. B 94, 094405 (2016).
+25 A. T. Costa, D. L. R. Santos, N. M. R. Peres, and J.
+Fern´andez-Rossier, 2D Mater. 7, 045031 (2020).
+26 D. ˇSabani, C. Bacaksiz, and M. V. Miloˇsevi´c, Phys. Rev.
+B 102, 014457 (2020).
+27 H. Xiang,
+C. Lee,
+H.-J. Koo,
+X. Gong,
+and M.-H.
+Whangbo, Dalton Transactions 42, 823 (2013).
+28 C. Xu, J. Feng, H. Xiang, and L. Bellaiche, npj Comput.
+Mater. 4, 1 (2018).
+29 See Supplemental Material at [URL will be inserted by
+publisher] for additional tables, figures and discussions of
+our results, and which cites Refs..2,26,30–38,40,48,50,51,53–56
+30 G. Kresse, and J. Hafner, Phys. Rev. B 47, 558 (1993).
+31 G. Kresse, and J. Furthm¨uller, Computational materials
+science 6, 15 (1996).
+32 G. Kresse, and J. Furthm¨uller, Phys. Rev. B 54, 11169
+(1996).
+33 P. E. Bl¨ochl, Phys. Rev. B 50, 17953 (1994).
+34 J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev.
+Lett. 77, 3865 (1996).
+35 S. Grimme, Journal of Computational Chemistry 27, 1787
+(2006).
+36 S. Steiner, S. Khmelevskyi, M. Marsmann, and G. Kresse,
+Phys. Rev. B 93, 224425 (2016).
+37 S. L. Dudarev, G. A. Botton, S. Y. Sevrasov, C. J.
+Humphreys, and A. P. Sutton, Phys. Rev. B 57, 1505
+(1998).
+38 G. P. M¨uller, M. Hoffmann, C. Dißelkamp, D. Sch¨urhoff,
+S. Mavros, M. Sallermann, N. S. Kiselev, H. J´onsson, and
+S. Bl¨ugel, Phys. Rev. B 99, 224414 (2019).
+39 S. Toth, and B. Lake, J. Phys.:
+Condens. Matter 27,
+166002 (2015).
+40 K. Momma, and F. Izumi, J. Appl. Cryst. 44, 1272 (2011).
+41 J. L. Lado, and J. Fern´andez-Rossier, 2D Mater. 4, 035002
+(2017).
+42 C. Bacaksiz, D. ˇSabani, R. M. Menezes, and M. V.
+Miloˇsevi´c, Phys. Rev. B 103, 125418 (2021).
+43 N. D. Mermin, and H. Wagner, Phys. Rev. Lett. 17, 1133
+(1966).
+44 F. Ma, Y. Zhou, H. B. Braun, and W. S. Lew, Nano Lett.
+15, 4029 (2015).
+45 T. Weber, D. M. Fobes, J. Waizner, P. Steffens, G. S.
+Tucker, M. B¨ohm, L. Beddrich, C. Franz, H. Gabold, R.
+Bewley et al., Science 375, 1025 (2022).
+46 T. Fukui, Y. Hatsugai, and H. Suzuki, J. Phys. Soc. Jpn.
+74, 1674 (2005).
+47 P. A. McClarty, Annu. Rev. Condens. Matter Phys. 13,
+171 (2022).
+48 N. Sivadas, S. Okamoto, X. Xu, C. J. Fennie and D. Xiao,
+Nano Lett. 18, 7658 (2018).
+49 M. A. McGuire, H. Dixit, V. R. Cooper and B. C. Sales,
+Chem. Mater. 27, 612 (2015).
+50 S. W. Jang, M. Y. Jeong, H. Yoon, S. Ryee and M. J. Han,
+Phys. Rev. Materials 3, 031001 (2019).
+51 P. Jiang, C. Wang, D. Chen, Z. Zhong, Z. Yuan, Z.-Y. Lu
+and W. Ji, Phys. Rev. B 99, 144401 (2019).
+52 D. Soriano, C. Cardoso and J. Fern´andez-Rossier, Solid
+State Commun. 299, 113662 (2019).
+53 L. Thiel, Z. Wang, M. A. Tschudin, D. Rohner, I.
+Guti´errez-Lezama, N. Ubrig, M. Gibertini, E. Giannini, A.
+F. Morpurgo and P. Maletinsky, Science 364, 973 (2019).
+54 T. Song, Z. Fei, M. Yankowitz, Z. Lin, Q. Jiang, K.
+Hwangbo, Q. Zhang, B. Sun, T. Taniguchi, K. Watanabe
+et al., Nature Mater. 18, 1298 (2019).
+55 T. Li, S. Jiang, N. Sivadas, Z. Wang, Y. Xu, D. Weber, J.
+E. Goldberger, K. Watanabe, T. Taniguchi, C. J. Fennie
+et al., Nature Mater. 18, 1303 (2019).
+56 W. Chen, Z. Sun, Z. Wang, L. Gu, X. Xu, S. Wu, and C.
+Gao, Science 366, 983 (2019).
+
+11
+57 Y. Jiang, Y. Guo, X. Yan, H. Zeng, L. Lin and X. Mou,
+Nanomat. 11, 2509 (2021).
+58 R. Zhao, G.-D. Xie, M. L. N. Chen, Z. Lan, Z. Huang, and
+W. E. I. Sha, Opt. Express 28, 4638 (2020).
+59 Z. Zhu, Y. Cheng, and U. Schwingenschl¨ogl, Phys. Rev. B
+85, 235401 (2012).
+60 T. O. Wehling, A. M. Black-Schaffer, and A. V. Balatsky,
+Adv. Phys. 76, 1 (2014).
+
+Stacking-dependent topological magnons in bilayer CrI3
+SUPPLEMENTARY MATERIAL
+M. Soenen,1 C. Bacaksiz,1, 2, 3 R. M. Menezes,1 and M. V. Miloˇsevi´c1, ∗
+1Department of Physics & NANOlab Center of Excellence,
+University of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium
+2Bremen Center for Computational Material Science (BCCMS), Bremen D-28359, Germany
+3Computational Biotechnology, RWTH Aachen University, Worringerweg 3, 52074 Aachen, Germany
+(Dated: January 9, 2023)
+I.
+DFT CALCULATIONS
+The DFT calculations are performed within the Vienna ab initio simulation package (VASP)1–3 using a generalized
+gradient approximation (GGA) functional and the projector augmented wave (PAW) method.4 We opt for the Perdew-
+Burke-Ernzerhof (PBE)5 exchange-correlation functional in combination with the D2 method of Grimme6 to account
+for a vdW correction term. A term due to the SOC7 is added ad hoc to the DFT Hamiltonian where necessary. For
+the Brillouin zone integration, we use a Gaussian smearing of 0.01 eV. Due to the periodic boundary consitions, we
+need to use 3 × 3 × 1 supercells in the 4SM calculations in order to calculate the NNN and 3NN interactions. To
+limit the computational cost of the 4SM calculations, we use a plane-wave energy cutoff of 300 eV and a 3×3×1 grid
+for the k-point sampling, for which we deem the systems sufficiently converged within reasonable computing times.
+In VASP, periodic boundary conditions are implemented automatically. Hence, during the calculations, we choose an
+out-of-plane unit cell distance of c = 15 ˚A for the monolayer and c = 26 ˚A for the bilayer ensuring a large enough
+vacuum distance between the materials and their periodic images, effectively minimizing the interaction between
+them. Since the system contains localized (strongly correlated) d-electrons, we implement the GGA+U method in
+the rotationally invariant form proposed by Dudarev et al.,8 i.e. we add an on-site Coulomb interaction of Ueff = U
+- J = 2.8 eV to the d-orbitals of the chromium atoms. To confirm that the chosen Ueff gives the desired qualitative
+description of the magnetic properties of CrI3, we plot the energy difference between the AFM and FM phases as a
+function U - J (with J = 1.1 eV) for all stacking orders [Fig. 1]. For the AB-stacked bilayer (green curve), we find
+a strong preference for a FM interlayer ordering for any value of U - J. Similarly, for the AA-stacking (black curve),
+we find a (smaller) preference for a FM interlayer ordering, except for very big U - J values. For the AB’-phase
+(magneta curve), however, we find a FM interlayer ordering for small U - J values which transitions to an AFM
+ordering for increasing U - J. Similarily to Jiang et al.,9 we choose values of U = 3.9 eV and J = 1.1 eV resulting in
+an effective parameter of U - J = 2.8 eV which slightly favors an AFM-coupled AB’-stacking and a FM-coupled AB-
+and AA-stackings. This agrees with our value of U = 2.8037 eV that we calculated with a linear response method for
+monolayer CrI3.
+FIG. 1: Convergence test for the effective U-parameter used in the DFT calculations. The total energy difference
+between the AFM and FM phases of the three considered stacking orders of bilayer CrI3 as a function of the on-site Coulomb
+interaction U - J (with J fixed to 1.1 eV).
+arXiv:2301.02502v1 [cond-mat.mtrl-sci] 6 Jan 2023
+
+14
+12
+AB
+10
+AB
+(meV)
+AA
+8
+EFM
+6
+4
+0
+-2
+0
+0.5
+1
+1.5
+2
+2.5
+3
+3.5
+4
+U-J (eV)2
+II.
+STACKING DEPENDENCE OF THE INTERLAYER COUPLING
+A.
+Energetic analysis from first-principles
+To map out the stable phases of bilayer CrI3, we plot its energy as the top layer is shifted laterally [Fig. 2 (a,b)].
+The structures are relaxed in the out-of-plane direction to find the optimal interlayer distance. For both an AFM
+[Fig. 2 (a)] and a FM [Fig. 2 (b)] ordered bilayer, we find similar energy profiles with clear minima located at the
+naturally occurring rhombohedral (AB) and monoclinic (AB’) phases, and a local minimum at the AA stacking order,
+hence our choice for these stackings during this work. To determine the the preferential interlayer magnetic order
+for each stacking, we plot the energy difference between the AFM and FM phases in Fig. 2 (c). Stacking orders
+that prefer an AFM configuration show up in blue, phases with a preference for a FM ordering show up in red. In
+the groundstate, the system will take on an AB-stacking combined with a FM interlayer coupling. The AA-stacked
+bilayer also prefers a FM interlayer coupling. Although the energy difference is very small, the AB’-stacked bilayer
+prefers an AFM interlayer coupling. This small energy difference is better illustrated in Fig. 2 (d) which depicts an
+intersection of the AFM and FM energy profiles along the pathway marked with a black dashed line in Fig. 2 (a-c).
+The results reported here are in good agreement with earlier work by Jiang et al.9 and Sivadas et al..10 All results
+presented in Fig. 2 were obtained from first-principles using DFT calculations implemented in VASP.
+(a)
+(b)
+(c)
+(d)
+FIG. 2: Total energy of bilayer CrI3 as a function of a lateral shift with respect to an AB-stacking. The total
+energy is depicted as a function of a lateral shift of the top layer along basis vectors a and b, for a bilayer with AFM (a) and
+FM (b) interlayer ordering respectively. Minima in the energy correspond to the AB, AB’ and AA-stacking orders. The energy
+difference between the AFM and FM phases is shown in (c). Panel (d) shows the energy per Cr atom, with respect to the
+ground-state (AB,FM), along the transition pathway that is marked with a black dashed line in panels (a-c).
+
+Eafm (meV)
+2/3
+AB
+-6.06
+AB'
+AB"
+1/3
+-6.065
+X AA
+X AB
+AB X
+-6.07
+X.AB
+X AB'
+AB'X
+1/3
+b
+2/3
+AB
+-6.075
+×104Efm (meV)
+AB
+-6.06
+2/3
+AB'
+AB'X
+1/3
+-6.065
+X
+AB
+X AA
+X AB
+AB X
+-6.07
+XAB
+X AB
+AB'X
+1/3
+b
+-6.075
+2/3
+AB
+-6.08
+× 10415
+2/3
+AB
+AB'
+ABX
+10
+1/3
+AB
+X AA
+X AB
+AB X
+5
+0
+XAB
+X AB'
+AB'X
+1/3
+b
+0
+2/3
+AB
+X
+1
+-520
+FM
+一AFM
+Cr)
+15
+(mev
+10
+E
+5
+0
+0
+0.2
+0.4
+0.6
+0.8
+1
+x.s3
+B.
+Atomistic spin-dynamics simulations
+The further investigate the dynamic stability of the different phases of bilayer CrI3, both with and without the
+presence of an external magnetic field, we perform atomistic spin dynamics (ASD) simulations, by numerically solving
+the Landau-Lifshitz-Gilbert (LLG) equation:
+∂ˆSi
+∂t = −
+γ
+(1 + α2) µ
+�
+ˆSi × Beff
+i
++ αˆSi ×
+�
+ˆSi × Beff
+i
+��
+,
+(1)
+in which γ is the gyromagnetic ratio, α = 0.001 the damping parameter, µ = 3µB the magnetic moment per Cr atom,
+and Beff
+i
+= −∂ ˆH/∂ˆSi the effective field. The latter is the resulting field due to all the magnetic interactions considered
+in the Heisenberg Hamiltonian (see main text). The simulations are performed at T = 0 K on a 50 × 50 supercell for
+a duration of 105 iterations. The ASD simulations are implemented in the Spirit11 software package which is adapted
+to accommodate the Hamiltonian considered in this work.
+In the abscence of an external magnetic field, our LLG simulations show that a FM interlayer ordering is a
+(meta)stable state for each of the three stacking orders.
+However, an AFM ordering is only stable for the AB’
+and AA-stackings. In the AB-stacking, the structure always relaxes to a FM ordered state, even when the structure
+is initialized with an AFM one. This spontaneous change in magnetization is attributed to the large energy difference
+between FM and AFM phases.
+In Figure 3, we report results for the magnetic field dependence of the AFM interlayer ordering, by plotting the
+magnetization for the top and bottom layer separately as a function of the applied field. The field is oriented in the
+out-of-plane direction. In the AB-stacking [Fig. 3 (b)], the magnetization will always be fully saturated, even in the
+absence of a field. Hence, an AFM interlayer ordering is not stable for this stacking. However, for the AA-stacking
+[Fig. 3 (a)] and the AB’-stacking [Fig. 3 (c)], we see that the AFM ordering remains stable up to relatively high
+fields of respectively B = 1.6 T and B = 1.9 T. For higher applied fields, the magnetization in one of the layers will
+flip (the layer with opposite magnetization to the field), leading to a FM ordering oriented parallel to the field. For a
+field with opposite orientation, we find similar results.
+(a) AA-stacking
+(b) AB-stacking
+(c) AB’-stacking
+FIG. 3: Magnetic field dependence of the interlayer ordering in the AFM ordered CrI3 bilayer for different
+stacking orders. Magnetization of the top layer (red) and bottom layer (blue) of an AFM ordered CrI3 bilayer as a function of
+the applied magnetic field, for respectively the AA-stacking (a), AB-stacking (b) and the AB’-stacking (c). The magnetization
+is normalized with respect to the saturation magnetization.
+
+1.00
+Mz (top)
+Mz (bot)
+0.75
+0.50
+0.25
+0.00
+-0.25
+-0.50
+-0.75
+-1.00
+0.0
+0.5
+1.0
+1.5
+2.0
+2.5
+3.0
+B [T]1.00
+Mz (top)
+Mz (bot)
+0.75
+0.50
+0.25
+/Ms
+0.00
+M
+-0.25
+-0.50
+-0.75
+-1.00
+0.0
+0.5
+1.0
+1.5
+2.0
+2.5
+3.0
+B [T]1.00
+Mz (top)
+Mz (bot)
+0.75
+0.50
+0.25
+0.00
+-0.25
+-0.50
+-0.75
+-1.00
+0.0
+0.5
+1.0
+1.5
+2.0
+2.5
+3.0
+B [T]4
+III.
+MAGNETIC PARAMETERS
+In this section, we present the magnetic parameters – obtained through a 4SM analysis – for all the pairs necessary
+to characterize the magnonic behavior of monolayer and bilayer CrI3. Tables I - IV, contain the magnetic parameters
+for respectively the monolayer [Fig. 4], the AA-stacking [Fig. 5], the AB-stacking [Fig. 6], and the AB’-stacking [Fig.
+7]. For these four systems, we report parameters for the NN and NNN intra- and interlayer exchange and DMI [Tab.
+I - IV], the SIA, and the Kitaev interactions [Tab. V]. The exchange- and DMI parameters are reported in Cartesian
+coordinates [Tab. I - IV], afterwards in Table V, we consider local eigenbases that diagonalize the symmetric exchange
+matrices to quantify the Kitaev interactions.
+In all considered structures, the intralayer exchange is FM and anisotropic with the NN interaction delivering the
+dominant contribution, the NNN exchange also prefers a FM ordering but is weaker than the former. The out-of-plane
+exchange anisotropy, i.e. ∆ = ⟨Jzz
+ij ⟩ − ⟨Jαα
+ij ⟩ with α = {x, y}, is the most important contribution to the magnetic
+anisotropy of CrI3 and causes the spins to prefer an out-of-plane orientation. In monolayer CrI3, the NN DMI is equal
+to zero due to the material’s inversion symmetry. Stacking two layers in a bilayer breaks this inversion symmetry
+leading to a non-zero NN DMI in all stacking orders. Between intralayer NNN spins, there is no inversion symmetry
+resulting, for all structures, in a small non-zero DMI.
+The AB- and AA-stacked bilayers, both show a FM NN and NNN interlayer coupling. The strength of the interlayer
+coupling is significantly stronger for the AB-stacking as it has more interacting pairs that also interact more strongly.
+In the AB’-stacked bilayer, there is competition between the FM NN interlayer coupling and the AFM NNN and 3NN
+interlayer coupling, resulting in an overall preference for an AFM interlayer ordering. For all stackings, the interlayer
+DMI is significantly weaker than the intralayer one, having almost no influence on the material’s properties.
+Notice that, in this work, we only report 3NN parameters for the AB’-stacking. In this stacking, the 3NN interlayer
+interaction is indispensable in order to achieve an overall AFM interlayer ordering as the ground state, in agreement
+with experiment.13–17 In all other cases, we can safely neglect the 3NN interactions since they deliver only minor
+contributions to the overall exchange. To illustrate the latter, we calculate several 3NN parameters to demonstrate
+that their characteristic orders of magnitude are negligible. For example, in the AB-stacked bilayer we found 3NN
+intralayer exchange pairs of (Jxx
+2−7, Jyy
+2−7, Jzz
+2−7) = (0.00, 0.00, 0.02) meV. Similar values were found for the monolayer
+and the other bilayers. Hence, the 3NN intralayer interaction is negligible in all structures. Regarding the 3NN
+interlayer exchange, we found values of (Jxx
+7−16, Jyy
+7−16, Jzz
+7−16) = (0.01, 0.00, 0.00) meV in the AB- and AA-stacking,
+values like this can also safely be neglected without significant influence on further results.
+In principle, the SIA matrix is symmetric resulting in six unique matrix elements that need to be calculated.
+However, notice that, when there is a 3-, 4-, or 6-fold rotational symmetry around the out-of-plane axis, it suffices to
+calculate only one parameter Azz
+ii .12 For the monolayer, the AA-stacking, and the AB-stacking, in which this symmetry
+is upheld, we find values for ⟨Azz
+ii ⟩ of respectively -0.08 meV, -0.08 meV and -0.07 meV. In the AB’-stacked bilayer,
+where this symmetry is absent, we need to calculate the full SIA-matrix, however, all elements vanish except ⟨Azz
+ii ⟩ =
+-0.07 meV and ⟨Ayz
+ii ⟩ = ⟨Azy
+ii ⟩ = 0.02 meV. Hence, for all structures, the SIA shows a preference for an out-of-plane
+orientation of the spins, delivering a small contribution to the material’s magnetic anisotropy.
+To calculate the Kitaev constants, we consider for each spin pair, an eigenbasis {λ, µ, ν} that diagonalizes the
+corresponding symmetric exchange matrix, yielding three eigenvalues Jλ
+ij, Jµ
+ij, and Jν
+ij. In the case that Jλ
+ij ≈ Jµ
+ij ̸= Jν
+ij,
+the exchange constant is defined as Jij = (Jλ
+ij +Jµ
+ij)/2 and the corresponding Kitaev constant as Kij = Jν
+ij −Jij. The
+DMI in the local basis is found by projecting the Cartesian DMI vector along the {λ, µ, ν} directions. We summarize
+the exchange and Kitaev constants in Table V. In all structures, the main contribution to the Kitaev interaction
+comes from the NN intralayer exchange.
+
+5
+TABLE I: Exchange parameters for monolayer CrI3. For each interacting pair of spins, the table contains the elements
+of both the NN and the NNN intralayer exchange matrices Jij, the out-of-plane exchange anisotropy ∆, and the components
+of the corresponding DMI-vectors Dij. All parameters are given in Cartesian coordinates. Corresponding pairs are indicated
+in Figure 4.
+pair
+Jxx
+ij
+Jyy
+ij
+Jzz
+ij
+∆
+Jxy
+ij
+Jyx
+ij
+Jxz
+ij
+Jzx
+ij
+Jyz
+ij
+Jzy
+ij
+|Dx
+ij|
+|Dy
+ij|
+|Dz
+ij|
+|Dij|
+(i-j)
+(meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV)
+NN
+(1-2)
+-4.34
+-3.24
+-3.96
+0.00
+0.00
+0.00
+0.00
+-0.65
+-0.65
+0.00
+0.00
+0.00
+0.00
+(2-3)
+-3.52
+-4.07
+-3.96
+-0.48
+-0.48
+-0.56
+-0.56
+0.33
+0.33
+0.00
+0.00
+0.00
+0.00
+(2-5)
+-3.52
+-4.07
+-3.96
+0.48
+0.48
+0.56
+0.56
+0.33
+0.33
+0.00
+0.00
+0.00
+0.00
+⟨Jij⟩
+-3.79
+-3.79
+-3.96
+-0.17
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+NNN (1-3)
+-0.67
+-0.73
+-0.66
+0.08
+0.03
+0.09
+0.03
+-0.01
+0.09
+0.05
+0.03
+0.03
+0.06
+(1-5)
+-0.67
+-0.73
+-0.66
+-0.08
+-0.03
+-0.09
+-0.03
+-0.01
+0.09
+0.05
+0.03
+0.03
+0.06
+(3-5)
+-0.76
+-0.64
+-0.66
+0.03
+-0.03
+-0.05
+0.05
+-0.08
+-0.08
+0.00
+0.05
+0.03
+0.06
+⟨Jij⟩
+-0.70
+-0.70
+-0.66
+0.04
+0.01
+-0.01
+-0.02
+0.02
+-0.03
+0.03
+0.03
+0.04
+0.03
+0.06
+FIG. 4: Labeled spin sites in monolayer CrI3. Top view of a 2 × 2 × 1 supercell of monolayer CrI3. Chromium atoms are
+represented by green spheres. Iodine atoms and bonds are not shown in the picture for the sake of simplicity. The supercell is
+marked with solid black lines. The crystal structures were drawn using VESTA.19
+
+Cr4
+Cr8
+Cr3
+Cr7
+b
+Cr2
+Cr6
+Cr1
+Cr56
+TABLE II: Exchange parameters for an AA-stacked CrI3 bilayer. For each interacting pair of spins, the table contains
+the elements of both the NN and the NNN intra- and interlayer exchange matrices Jij, the out-of-plane exchange anisotropy ∆,
+and the components of the corresponding DMI-vectors Dij. All parameters are given in Cartesian coordinates. Corresponding
+pairs are indicated in Figure 5.
+pair
+Jxx
+ij
+Jyy
+ij
+Jzz
+ij
+∆
+Jxy
+ij
+Jyx
+ij
+Jxz
+ij
+Jzx
+ij
+Jyz
+ij
+Jzy
+ij
+|Dx
+ij|
+|Dy
+ij|
+|Dz
+ij|
+|Dij|
+(i-j)
+(meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV)
+Intra
+NN
+(1-2)
+-4.40
+-3.34
+-4.05
+0.04
+-0.04
+0.11
+-0.11
+-0.63
+-0.63
+0.00
+0.11
+0.04
+0.12
+(2-3)
+-3.62
+-4.15
+-4.06
+-0.52
+-0.44
+-0.49
+-0.60
+0.41
+0.22
+0.10
+0.06
+0.04
+0.12
+(2-5)
+-3.62
+-4.15
+-4.06
+0.44
+0.52
+0.60
+0.49
+0.22
+0.41
+0.10
+0.05
+0.04
+0.12
+⟨Jij⟩
+-3.88
+-3.88
+-4.05
+-0.17
+-0.01
+0.01
+0.07
+-0.07
+0.00
+0.00
+0.07
+0.07
+0.04
+0.12
+NNN
+(1-3)
+-0.59
+-0.64
+-0.57
+0.10
+0.02
+0.08
+0.03
+0.00
+0.06
+0.03
+0.02
+0.04
+0.06
+(1-5)
+-0.67
+-0.56
+-0.57
+0.04
+-0.05
+-0.04
+0.04
+-0.07
+-0.06
+0.00
+0.04
+0.04
+0.06
+(3-5)
+-0.59
+-0.65
+-0.57
+-0.09
+-0.01
+-0.08
+-0.03
+0.00
+0.07
+0.04
+0.02
+0.04
+0.06
+⟨Jij⟩
+-0.62
+-0.62
+-0.57
+0.05
+0.02
+-0.01
+-0.01
+0.01
+-0.02
+0.02
+0.03
+0.03
+0.04
+0.06
+Inter
+NN
+(1-9)
+-0.12
+-0.12
+-0.04
+0.08
+-0.06
+0.06
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.06
+0.06
+NNN (7-12)
+-0.13
+-0.11
+-0.12
+-0.01
+-0.01
+0.00
+0.00
+-0.01
+-0.01
+0.00
+0.00
+0.00
+0.00
+(7-14)
+-0.13
+-0.12
+-0.12
+0.01
+0.01
+0.01
+0.01
+0.01
+0.01
+0.00
+0.00
+0.00
+0.00
+(7-16)
+-0.11
+-0.13
+-0.11
+-0.01
+-0.01
+-0.01
+-0.01
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+⟨Jij⟩
+-0.12
+-0.12
+-0.12
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+(a) Top view.
+(b) Side view.
+FIG. 5: Labeled spin sites in an AA-stacked CrI3 bilayer. Top view (a) and side view (b) of a 2 × 2 × 1 supercell of an
+AA-stacked CrI3 bilayer. Chromium atoms in the top- and bottom layer are represented by green and blue spheres respectively.
+Iodine atoms and bonds are not shown in the picture for the sake of simplicity. The supercell is marked with solid black lines
+in panel (a). The crystal structures were drawn using VESTA.19
+
+Cr4
+Cr8
+Cr3
+Cr7
+Cr2
+Cr6
+Cr1
+Cr5
+Cr12
+Cr16
+Cr11
+Cr15
+a
+Cr10
+Cr14
+Cr9
+Cr13Cr4
+Cr8
+Cr3
+Cr7
+b
+Cr2
+Cr6
+Cr1
+Cr57
+TABLE III: Exchange parameters for an AB-stacked CrI3 bilayer. For each interacting pair of spins, the table contains
+the elements of both the NN and the NNN intra- and interlayer exchange matrices Jij, the out-of-plane exchange anisotropy
+∆, and the components of the corresponding DMI-vectors Dij. Since the sublattice symmetry is broken, parameters are given
+for the two sublattices (A,B) separately. All parameters are given in Cartesian coordinates. Corresponding pairs are indicated
+in Figure 6.
+pair
+Jxx
+ij
+Jyy
+ij
+Jzz
+ij
+∆
+Jxy
+ij
+Jyx
+ij
+Jxz
+ij
+Jzx
+ij
+Jyz
+ij
+Jzy
+ij
+|Dx
+ij|
+|Dy
+ij|
+|Dz
+ij|
+|Dij|
+(i-j)
+(meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV)
+Intra
+NN
+(1-2)
+-4.63
+-3.56
+-4.25
+-0.04
+0.03
+-0.13
+0.11
+-0.64
+-0.64
+0.00
+0.12
+0.03
+0.13
+(2-3)
+-3.43
+-3.93
+-3.86
+-0.47
+-0.43
+-0.54
+-0.50
+0.26
+0.34
+0.04
+0.02
+0.02
+0.05
+(2-5)
+-3.84
+-4.34
+-4.25
+0.49
+0.53
+0.50
+0.61
+0.42
+0.22
+0.10
+0.05
+0.02
+0.11
+⟨Jij⟩
+-3.96
+-3.94
+-4.12
+-0.16
+-0.01
+0.04
+-0.06
+0.07
+0.02
+-0.02
+0.05
+0.06
+0.02
+0.10
+NNNA (1-3)
+-0.53
+-0.58
+-0.52
+0.08
+0.02
+0.07
+0.02
+0.02
+0.06
+0.02
+0.02
+0.03
+0.04
+(1-5)
+-0.53
+-0.58
+-0.51
+-0.08
+-0.02
+-0.07
+-0.03
+-0.01
+0.05
+0.03
+0.02
+0.03
+0.05
+(3-5)
+-0.61
+-0.50
+-0.50
+0.03
+-0.03
+-0.04
+0.04
+-0.06
+-0.06
+0.00
+0.04
+0.03
+0.05
+⟨Jij⟩
+-0.55
+-0.55
+-0.51
+0.04
+0.01
+-0.01
+-0.01
+0.01
+-0.02
+0.02
+0.02
+0.03
+0.03
+0.05
+NNNB (2-4)
+-0.58
+-0.64
+-0.57
+0.02
+0.08
+0.03
+0.08
+0.07
+0.00
+0.04
+0.03
+0.03
+0.06
+(2-6)
+-0.67
+-0.56
+-0.57
+-0.03
+0.04
+0.04
+-0.04
+-0.06
+-0.07
+0.00
+0.04
+0.03
+0.05
+(4-6)
+-0.59
+-0.64
+-0.57
+-0.02
+-0.09
+-0.03
+-0.08
+0.07
+-0.01
+0.04
+0.02
+0.03
+0.06
+⟨Jij⟩
+-0.61
+-0.61
+-0.57
+0.04
+-0.01
+0.01
+0.01
+-0.01
+0.03
+-0.02
+0.03
+0.03
+0.03
+0.05
+Inter
+NNB
+(2-9)
+-0.26
+-0.27
+-0.30
+-0.02
+0.06
+0.06
+0.01
+0.02
+0.01
+0.01
+0.00
+0.00
+0.00
+0.00
+NNNA (7-10)
+-0.39
+-0.49
+-0.46
+-0.04
+-0.04
+-0.04
+-0.04
+0.05
+0.05
+0.00
+0.00
+0.00
+0.00
+(7-11)
+-0.31
+-0.30
+-0.32
+-0.03
+-0.03
+-0.02
+-0.01
+-0.01
+-0.01
+0.00
+0.00
+0.00
+0.01
+(7-12)
+-0.49
+-0.39
+-0.47
+-0.02
+-0.02
+-0.02
+-0.02
+-0.06
+-0.06
+0.00
+0.00
+0.00
+0.00
+(7-13)
+-0.32
+-0.28
+-0.31
+0.02
+0.02
+0.02
+0.02
+-0.02
+0.00
+0.01
+0.00
+0.00
+0.01
+(7-14)
+-0.43
+-0.45
+-0.47
+0.06
+0.06
+0.06
+0.06
+0.01
+0.01
+0.00
+0.00
+0.00
+0.00
+(7-15)
+-0.28
+-0.33
+-0.31
+0.01
+0.01
+0.01
+0.00
+0.03
+0.02
+0.00
+0.01
+0.00
+0.01
+⟨Jij⟩
+-0.37
+-0.37
+-0.39
+-0.02
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.01
+NNNB (8-12)
+-0.31
+-0.30
+-0.32
+-0.03
+-0.03
+-0.01
+-0.02
+-0.01
+-0.01
+0.00
+0.01
+0.00
+0.01
+(8-14)
+-0.32
+-0.28
+-0.31
+0.02
+0.02
+0.02
+0.02
+0.00
+-0.02
+0.01
+0.00
+0.00
+0.01
+(8-16)
+-0.28
+-0.33
+-0.31
+0.01
+0.01
+0.00
+0.01
+0.02
+0.03
+0.00
+0.01
+0.00
+0.01
+⟨Jij⟩
+-0.30
+-0.30
+-0.32
+-0.02
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.01
+0.00
+0.01
+(a) Top view.
+(b) Side view.
+FIG. 6: Labeled spin sites in an AB-stacked CrI3 bilayer. Top view (a) and side view (b) of a 2 × 2 × 1 supercell of an
+AB-stacked CrI3 bilayer. Chromium atoms in the top- and bottom layer are represented by green and blue spheres respectively.
+Iodine atoms and bonds are not shown in the picture for the sake of simplicity. The supercell is marked with solid black lines
+in panel (a). The crystal structures were drawn using VESTA.19
+
+Cr1
+Gr5
+Gr1
+Cr12
+Cr16
+Cr4
+Cr8
+Cr7
+Cr3
+Cr10
+Cr14
+Cr2
+Cr6
+cH
+Cr5
+ch1Cr1
+Cr5
+Cr1
+Cr4
+Cr8
+Cr3
+Cr7
+Cr3
+Cr2
+Cr6
+Cr1
+Cr5
+Cr1
+C
+Cr12
+Cr16
+Cr11
+Cr15
+Cr10
+Cr14
+Cr9
+Cr138
+TABLE IV: Exchange parameters for an AB’-stacked CrI3 bilayer. For each interacting pair of spins, the table contains
+the elements of both the NN and the NNN intra- and interlayer exchange matrices Jij, the out-of-plane exchange anisotropy
+∆, and the components of the corresponding DMI-vectors Dij. Since the sublattice symmetry is broken, parameters are given
+for the two sublattices (A,B) separately. All parameters are given in Cartesian coordinates. Corresponding pairs are indicated
+in Figure 7.
+pair
+Jxx
+ij
+Jyy
+ij
+Jzz
+ij
+∆
+Jxy
+ij
+Jyx
+ij
+Jxz
+ij
+Jzx
+ij
+Jyz
+ij
+Jzy
+ij
+|Dx
+ij|
+|Dy
+ij|
+|Dz
+ij|
+|Dij|
+(i-j)
+(meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV)
+Intra
+NN
+(1-2)
+-4.63
+-3.54
+-4.25
+0.06
+-0.01
+-0.12
+0.12
+-0.66
+-0.64
+0.01
+0.12
+0.03
+0.12
+(2-3)
+-3.43
+-3.93
+-3.86
+-0.48
+-0.44
+-0.54
+-0.50
+0.26
+0.34
+0.04
+0.02
+0.02
+0.05
+(2-7)
+-3.85
+-4.34
+-4.26
+0.47
+0.54
+0.51
+0.61
+0.43
+0.22
+0.11
+0.05
+0.03
+0.12
+⟨Jij⟩
+-3.97
+-3.94
+-4.12
+-0.15
+0.02
+0.03
+-0.05
+0.08
+0.01
+-0.03
+0.05
+0.06
+0.03
+0.10
+NNNA
+(1-3)
+-0.61
+-0.68
+-0.60
+0.08
+0.02
+0.08
+0.04
+0.00
+0.07
+0.03
+0.02
+0.03
+0.05
+(1-7)
+-0.64
+-0.53
+-0.54
+0.02
+-0.01
+-0.04
+0.03
+-0.07
+-0.07
+0.00
+0.03
+0.02
+0.04
+(3-7)
+-0.57
+-0.62
+-0.54
+-0.10
+0.00
+-0.06
+-0.04
+0.02
+0.05
+0.02
+0.01
+0.05
+0.05
+⟨Jij⟩
+-0.61
+-0.61
+-0.56
+0.05
+0.00
+0.00
+-0.01
+0.01
+-0.02
+0.02
+0.02
+0.02
+0.03
+0.05
+NNNB
+(2-4)
+-0.56
+-0.61
+-0.54
+0.04
+0.07
+0.03
+0.07
+0.07
+0.00
+0.03
+0.02
+0.02
+0.04
+(2-8)
+-0.56
+-0.62
+-0.54
+0.00
+-0.10
+-0.05
+-0.07
+0.04
+0.01
+0.02
+0.01
+0.05
+0.05
+(4-8)
+-0.70
+-0.58
+-0.59
+-0.04
+0.03
+0.04
+-0.03
+-0.07
+-0.08
+0.01
+0.04
+0.04
+0.05
+⟨Jij⟩
+-0.61
+-0.60
+-0.56
+0.05
+0.00
+0.00
+0.01
+-0.01
+0.01
+-0.02
+0.02
+0.02
+0.03
+0.05
+Inter
+NN
+(7-20)
+-0.30
+-0.20
+-0.30
+0.03
+0.03
+-0.01
+0.00
+-0.02
+-0.02
+0.00
+0.00
+0.00
+0.00
+(1-19)
+-0.26
+-0.31
+-0.29
+-0.03
+-0.06
+-0.05
+0.00
+-0.02
+0.06
+0.04
+0.03
+0.01
+0.05
+(2-19)
+-0.26
+-0.27
+-0.28
+0.05
+0.05
+0.02
+0.02
+0.02
+0.02
+0.00
+0.00
+0.00
+0.00
+(2-20)
+-0.26
+-0.29
+-0.29
+-0.05
+-0.04
+0.01
+-0.05
+0.06
+-0.03
+0.04
+0.03
+0.01
+0.05
+NNN
+(9-21)
+0.15
+0.12
+0.13
+-0.03
+-0.02
+0.00
+-0.04
+0.04
+-0.02
+0.03
+0.02
+0.01
+0.04
+(9-26)
+0.15
+0.20
+0.18
+-0.01
+-0.01
+0.00
+0.00
+-0.02
+-0.02
+0.00
+0.00
+0.00
+0.00
+(10-22)
+0.15
+0.12
+0.13
+-0.02
+-0.03
+-0.04
+0.00
+-0.02
+0.04
+0.03
+0.02
+0.01
+0.04
+(10-29)
+0.21
+0.17
+0.19
+0.02
+0.02
+0.02
+0.02
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+3NN
+(9-20)
+0.12
+0.12
+0.12
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+(9-23)
+0.07
+0.07
+0.06
+-0.03
+-0.02
+0.00
+-0.04
+0.04
+-0.02
+0.03
+0.02
+0.01
+0.04
+(9-25)
+0.06
+0.06
+0.06
+-0.06
+0.03
+-0.08
+0.08
+-0.01
+0.00
+0.01
+0.08
+0.05
+0.09
+(9-28)
+0.02
+0.01
+0.01
+-0.01
+-0.01
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+⟨Jij⟩
+0.07
+0.07
+0.07
+-0.03
+0.00
+-0.02
+0.01
+0.01
+0.00
+0.01
+0.02
+0.01
+0.03
+(10-23)
+0.13
+0.13
+0.12
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+(10-24)
+0.07
+0.07
+0.06
+0.03
+-0.03
+-0.03
+0.03
+0.08
+-0.07
+0.07
+0.03
+0.03
+0.09
+(10-26)
+0.06
+0.07
+0.05
+0.03
+-0.03
+0.08
+-0.08
+0.01
+-0.01
+0.01
+0.08
+0.03
+0.09
+(10-33)
+0.02
+0.01
+0.01
+-0.01
+-0.01
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+0.00
+⟨Jij⟩
+0.07
+0.07
+0.06
+0.01
+-0.01
+0.01
+-0.01
+0.02
+-0.02
+0.02
+0.03
+0.02
+0.03
+
+9
+(a) Top view.
+(b) Side view.
+FIG. 7: Labeled spin sites in an AB’-stacked CrI3 bilayer. Top view (a) and side view (b) of a 3 × 3 × 1 supercell
+of an AB’-stacked CrI3 bilayer. Chromium atoms in the top- and bottom layer are represented by green and blue spheres
+respectively. Iodine atoms and bonds are not shown in the picture for the sake of simplicity. The supercell is marked with solid
+black lines in panel (a). The crystal structures were drawn using VESTA.19
+
+cr6
+Cr18
+cr3
+Cr28
+cr3
+Cr9
+cr2
+Cr20
+Cr8
+Cr26
+Cr32Cr6
+Cr12
+Cr18
+Cr5
+Cr11
+Cr17
+Cr4
+Cr10
+Cr16
+Cr9
+Cr2
+Cr8
+Cr1
+Cr7
+Cr13
+Cr24
+Cr30
+cr36
+Cr23
+Cr29
+Cr35
+Cr22
+Cr28
+Cr34
+Cr21
+Cr33
+Cr20
+Cr26
+Cr32
+Cr25
+Cr3110
+TABLE V: Exchange parameters for CrI3 in the {λ, µ, ν} basis. Intra- and interlayer NN and NNN exchange, DMI and
+Kitaev parameters for monolayer (1L) and bilayer (2L) CrI3. The Kitaev interaction for the AB-stacking and the AB’-stacking
+is anisotropic so multiple values are given for each interaction.
+pair(s)
+Jij
+Kij
+|Dij|
+(i-j)
+(meV)
+(meV)
+(meV)
+1L
+Intra
+NN
+All
+-4.35
+1.49
+0.00
+NNN
+All
+-0.74
+0.17
+0.06
+2L-AA
+Intra
+NN
+All
+-4.42
+1.44
+0.12
+NNN
+All
+-0.65
+0.15
+0.06
+Inter
+NN
+All
+-0.12
+0.08
+0.06
+NNN
+All
+-0.13
+0.02
+0.00
+2L-AB
+Intra
+NN
+(1-2)
+-4.62
+1.45
+0.13
+NN
+(2-3)
+-4.20
+1.39
+0.05
+NN
+(2-5)
+-4.64
+1.50
+0.11
+NNNA
+All
+-0.59
+0.15
+0.05
+NNNB
+All
+-0.65
+0.14
+0.06
+Inter
+NNB
+All
+-0.31
+0.11
+0.00
+NNN
+a
+-0.33
+0.05
+0.01
+NNN
+b
+-0.50
+0.15
+0.00
+2L-AB’
+Intra
+NN
+(1-2)
+-4.62
+1.47
+0.12
+NN
+(2-3)
+-4.20
+1.39
+0.05
+NN
+(2-7)
+-4.64
+1.49
+0.12
+NNN
+(1-3), (4-8)
+-0.68
+0.16
+0.04
+NNN
+(1-7), (2-4)
+-0.62
+0.15
+0.04
+NNN
+(3-7), (2-8)
+-0.62
+0.15
+0.05
+Inter
+NN
+(7-20), (2-19)
+-0.31
+0.10
+0.00
+NN
+(1-19), (2-20)
+-0.31
+0.10
+0.05
+NNN
+(9-26), (10-29)
+0.16
+0.06
+0.00
+NNN
+(9-21), (10-22)
+0.11
+0.05
+0.04
+3NN
+(9-20), (10-23)
+0.12
+0.01
+0.00
+3NN
+(9-23), (10-24)
+0.05
+0.06
+0.04
+3NN
+(9-25), (10-26)
+0.05
+0.03
+0.09
+3NN
+(9-28), (10-33)
+0.01
+0.01
+0.00
+a Pairs: (7-11), (7-13), (7-15), (8-12), (8-14), (8-16)
+b Pairs: (7-10), (7-12), (7-14)
+
+11
+IV.
+ARTIFICIAL TUNING OF THE MAGNONIC DISPERSION
+As discussed in the main text, we can ’tune’ the spin-wave dispersion by artificially changing the magnetic pa-
+rameters. As shown in Figure 8(a,b) the size of the magnonic bandgap is tunable by changing the strength of the
+NNN DMI and the Kitaev constant. However, changing the Kitaev constant influences the shape of the dispersion in
+the whole Brillouin zone whereas the NNN DMI mainly effects the dispersion around the K-point and the K’-point.
+Similar bandgap tuning can be observed by artificially changing parameters in bilayer CrI3. In Figure 8(c), we show
+for an FM ordered AB-stacked bilayer that by decreasing the out-of-plane exchange difference between sublattices
+∆Jzz = |Jzz
+A − Jzz
+B |, which is caused by the breaking of sublattice symmetry, we can also decreases the size of the
+magnonic bandgap and induce a topological phase transition from a topologically trivial to a topologically non-trivial
+phase. Similar changes in the bandgap can be observed by changing decreasing ∆Jzz in the AB’-stacking.
+(a)
+(b)
+(c)
+FIG. 8: Artificial tuning of the magnonic dispersion by manually changing the magnetic parameters. (a) Size of
+the magnonic bandgap in monolayer CrI3 plotted as a function of the NNN DMI. (b) Magnonic dispersion of monolayer CrI3
+for different values of the Kitaev constant. (c) Magnonic bandgap in an AB-stacked CrI3 bilayer with a FM interlayer ordering
+as a function of the out-of-plane exchange difference between the two sublattices. Corresponding (composite) Chern numbers
+are indicated for the topological trivial phase (white) and the topological non-trivial phase (red).
+
+1.19
+0.98
+(meV)
+0.77
+0.57
+0.46
+0.15
+0.06
+0
+0.03
+0.06
+0.09
+0.12
+0.15
+0.18
+D2
+(meV)25
+K = 1.49 meV
+-
+-
+K=0.98meV
+1
+20
+K=0.76meV
+-
+-
+-
+一
+≥15
+(me)
+-
+-
+E10
+一
+-
+一
+一
+5
+1
+一
+-
+1
+1
+1
+-
+1
+0
+I
+K
+M0.57
+0.56
+C102 = 0
+C3 = 0
+C4 = 0
+0.54
+K
+0.52
+C1θ2 = +1
+0.5
+C = -1
+0.49
+0=
+0.48
+0.74
+0.77
+0.8
+0.83
+0.86
+0.89
+0.92
+AJ22 (meV)12
+∗ Electronic address: milorad.milosevic@uantwerpen.be
+1 G. Kresse, and J. Hafner, Phys. Rev. B 47, 558 (1993).
+2 G. Kresse, and J. Furthm¨uller, Computational materials science 6, 15 (1996).
+3 G. Kresse, and J. Furthm¨uller, Phys. Rev. B 54, 11169 (1996).
+4 P. E. Bl¨ochl, Phys. Rev. B 50, 17953 (1994).
+5 J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996).
+6 S. Grimme, Journal of Computational Chemistry 27, 1787 (2006).
+7 S. Steiner, S. Khmelevskyi, M. Marsmann, and G. Kresse, Phys. Rev. B 93, 224425 (2016).
+8 S. L. Dudarev, G. A. Botton, S. Y. Sevrasov, C. J. Humphreys, and A. P. Sutton, Phys. Rev. B 57, 1505 (1998).
+9 P. Jiang, C. Wang, D. Chen, Z. Zhong, Z. Yuan, Z.-Y. Lu and W. Ji, Phys. Rev. B 99, 144401 (2019).
+10 N. Sivadas, S. Okamoto, X. Xu, C. J. Fennie and D. Xiao, Nano Lett. 18, 7658 (2018).
+11 G. P. M¨uller, M. Hoffmann, C. Dißelkamp, D. Sch¨urhoff, S. Mavros, M. Sallermann, N. S. Kiselev, H. J´onsson, and S. Bl¨ugel,
+Phys. Rev. B 99, 224414 (2019).
+12 D. ˇSabani, C. Bacaksiz, and M. V. Miloˇsevi´c, Phys. Rev. B 102, 014457 (2020).
+13 B. Huang, G. Clark, E. Navarro-Moratalla, D. R. Klein, R. Cheng, K. L. Seyler, D. Zhong, E. Schmidgall, M. A. McGuire,
+D. H. Cobden, et al., Nature 546, 270 (2017).
+14 L. Thiel, Z. Wang, M. A. Tschudin, D. Rohner, I. Guti´errez-Lezama, N. Ubrig, M. Gibertini, E. Giannini, A. F. Morpurgo
+and P. Maletinsky, Science 364, 973 (2019).
+15 T. Song, Z. Fei, M. Yankowitz, Z. Lin, Q. Jiang, K. Hwangbo, Q. Zhang, B. Sun, T. Taniguchi, K. Watanabe et al., Nature
+Mater. 18, 1298 (2019).
+16 T. Li, S. Jiang, N. Sivadas, Z. Wang, Y. Xu, D. Weber, J. E. Goldberger, K. Watanabe, T. Taniguchi, C. J. Fennie et al.,
+Nature Mater. 18, 1303 (2019).
+17 W. Chen, Z. Sun, Z. Wang, L. Gu, X. Xu, S. Wu, and C. Gao, Science 366, 983 (2019).
+18 S. W. Jang, M. Y. Jeong, H. Yoon, S. Ryee and M. J. Han, Phys. Rev. Materials 3, 031001 (2019).
+19 K. Momma, and F. Izumi, J. Appl. Cryst. 44, 1272 (2011).
+
diff --git a/YdE0T4oBgHgl3EQfmwE9/content/tmp_files/load_file.txt b/YdE0T4oBgHgl3EQfmwE9/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..cd3244b5e23dc63a8173a6e762710f2626819227
--- /dev/null
+++ b/YdE0T4oBgHgl3EQfmwE9/content/tmp_files/load_file.txt
@@ -0,0 +1,2559 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf,len=2558
+page_content='Stacking-dependent topological magnons in bilayer CrI3 M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Soenen,1 C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bacaksiz,1, 2, 3 R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Menezes,1 and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miloˇsevi´c1, ∗ 1Department of Physics & NANOlab Center of Excellence, University of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium 2Bremen Center for Computational Material Science (BCCMS), Bremen D-28359, Germany 3Computational Biotechnology, RWTH Aachen University, Worringerweg 3, 52074 Aachen, Germany (Dated: January 9, 2023) Motivated by the potential of atomically-thin magnets towards tunable high-frequency magnonics, we detail the spin-wave dispersion of bilayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We demonstrate that the magnonic behavior of the bilayer strongly depends on its stacking configuration and the interlayer magnetic ordering, where a topological bandgap opens in the dispersion caused by the Dzyaloshinskii-Moriya and Kitaev interactions, classifying bilayer CrI3 as a topological magnon insulator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We further reveal that both size and topology of the bandgap in a CrI3 bilayer with an antiferromagnetic interlayer ordering are tunable by an external magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' INTRODUCTION Emergent two-dimensional (2D) magnetic materials1 provide an exciting platform to study collective spin ex- citations, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' magnons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' CrI3, the archetypal 2D van der Waals (vdW) ferromagnet,2 has recently been suggested to host magnon modes in the highly sought-after tera- hertz (THz) regime,3–5 showing promise for the develop- ment of faster and more energy-efficient data processing applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 Moreover, due to the 2D nature of the ma- terial, its spin-wave properties are highly susceptible to tuning, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' by strain, buckling, defect-engineering, gat- ing and/or vdW heterostructuring.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 Recently, several bulk materials, including CrBr3,7 CrI3,8,9 CrSiTe310 and CrGeTe3,10 have been identified as topological magnon insulators (TMIs), characterized by bulk magnon bands with a gap at the Dirac point, and topologically protected edge states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The magnonic bandgap is attributed to the anti-symmetric exchange in- teraction – more often called the Dzyaloshinskii-Moriya interaction (DMI)11,12 – arising from the lack of inver- sion symmetry between next-nearest-neighboring (NNN) Cr atoms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In contrast, bulk CrCl3,13,14 where the DMI is weak, is classified as a magnon Dirac material (MDM), characterized by a Dirac-point in the dispersion, showing a linear band crossing at the Brillouin zone edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, it remains an open question whether the topological features of aforementioned materials will per- sist down to the monolayer limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Early theoretical work suggested that honeycomb ferromagnetic (FM) mono- layers could be classified as either MDMs or TMIs de- pending on whether any NNN DMI is present in the material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15–20 Nonetheless, recent work identified Ki- taev interactions as an alternative mechanism potentially able to open a topological bandgap in FM honeycomb materials,21,22 suggesting that the absence of DMI is not the sole criterion for predicting the topological proper- ties of such materials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similarly, in magnetic honey- comb bilayers, a DMI-induced topological behavior of magnons is predicted,20,23,24 including the formation of Dirac magnon nodal-line loops,23 and the opening of a topological bandgap, which contributes to a magnon Hall- and a spin Nernst effect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='20,24 A first attempt to characterize the magnonics of mono- layer CrI3, using an itinerant fermion description based on ab initio calculations, showed the appearance of a small, possibly topological, bandgap caused by the spin- orbit coupling (SOC),25 suggesting that the material is a TMI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, more work is required before full un- derstanding of the magnonics in CrI3 is achieved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In this work, we deploy a multi-scale approach, combining ab initio calculations with numerical simulations based on a Heisenberg model and linear spin-wave theory, to char- acterize the magnonic properties of CrI3 monolayers and bilayers, to reveal the topological magnon modes present in these systems, and that (topological) magnonic prop- erties of the bilayer are strongly affected by its stacking order and its interlayer magnetic ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The article is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In section II, we de- scribe the computational methodology used in this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We discuss the Heisenberg Hamiltonian that models the magnetic interactions in CrI3, explain how the parame- ters that characterize this Hamiltonian will be derived from first-principles and, finally, sketch how the spin- wave dispersion is computed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Subsequently, in section III, we apply this methodology to monolayer CrI3, con- firming the presence of a small topological bandgap with non-zero Chern numbers in the material’s spin-wave dis- persion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Afterwards, in section IV, we consider bilayer CrI3 in three different stacking orders, each exhibiting significantly different magnonic behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Specifically, we investigate the AA-stacking and AB-stacking (rhom- bohedral) discussed in literature,20,23,24 as well as the experimentally very relevant AB’-stacking (monoclinic) of which the spin-waves have - to the best of our knowl- edge - not been theoretically investigated to date.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We find that all three stacking versions of bilayer CrI3 ex- hibit either FM or antiferromagnetic (AFM) interlayer ordering, with intralayer ferromagnetism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In case of a FM interlayer ordering, we observe a bandgap in the spin-wave dispersion with stacking-dependent topologi- cal properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We attribute the origin of the gap to a combination of DMI and Kitaev interactions that are arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02502v1 [cond-mat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='mtrl-sci] 6 Jan 2023 2 modulated by the stacking order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Furthermore, we show a significant influence of the interlayer magnetic order- ing on the resulting magnonic behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Specifically, the topological nature of the bands becomes trivial in AFM- ordered bilayers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Additionally, we show that magnonic dispersion of AFM-ordered bilayers is susceptible to tun- ing by an external magnetic field, lifting the degener- acy between branches, which decreases the size of the magnonic bandgap and leads to a non-trivial topology of the bands in the AB’-stacking, or introduces nodal-line loops in the AA-stacking case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Finally, section V sum- marizes our findings and gives an outlook on some future challenges and opportunities within the field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' COMPUTATIONAL METHODOLOGY We model the magnetic interactions of the system un- der study using a Heisenberg spin Hamiltonian of the following form: ˆH = 1 2 � i,j ˆSiJijˆSj + � i ˆSiAiiˆSi + µB � i B · giˆSi, (1) in which the spins are three-dimensional (3D) vectors ˆSi = ( ˆSx i , ˆSy i , ˆSz i ) expressed in Cartesian coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The first- and second term of this Hamiltonian respec- tively describe the exchange interaction and the single ion anisotropy (SIA), which are characterized by the 3 × 3 matrices Jij and Aii.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The DMI is characterized by a vector Dij with components that can be calcu- lated from the off-diagonal elements of the exchange ma- trix as Dx ij = 1 2(J yz ij − J zy ij ), Dy ij = 1 2(J zx ij − J xz ij ) and Dz ij = 1 2(J xy ij −J yx ij ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26,27 Notice that Dij = νij|Dij| with νij = −νji = ±1, where the sign of the latter depends on the hopping direction of the considered spin pair.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The exchange term is now written as: ˆHex = 1 2 � i,j � � α′ Jα′ ij ˆSα′ i ˆSα′ j + Dij(ˆSi × ˆSj) � , (2) with α′ = {α, β, γ} the local eigenbases that diagonalize the symmetric part of the exchange matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To consider the exchange anisotropy, we define the Kitaev constant as Kij = Jγ ij − Jij with Jij = (Jα ij + Jβ ij)/2 the isotropic exchange constant,28 leading to the following expression for the exchange Hamiltonian: ˆHex = 1 2 � i,j � JijˆSi · ˆSj + Kij ˆSγ i ˆSγ j + Dij(ˆSi × ˆSj) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The symmetric SIA-matrix Aii accounts for the interac- tion of the magnetic orbitals with the surrounding crystal field and contributes to the magnetic anisotropy of the material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In crystals with a 3-, 4-, or 6-fold rotational symmetry around the out-of-plane axis, most elements of the matrix are redundant and the SIA can be char- acterized by a single parameter Azz ii instead of the full SIA-matrix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26,27 The last term of equation (1) accounts for the Zeeman interaction when applying an external magnetic field B, where gi ≈ 2 is the g-factor, and µB is the Bohr magneton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In CrI3, the magnetic dipole-dipole interaction is expected to be small in comparison to its out-of-plane magnetic anisotropy and will, therefore, not be included in the Heisenberg Hamiltonian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 Finally, also notice that CrI3 has a magnetic moment of µ = 3µB per chromium atom and, thus, a spin of S = 3/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To obtain the elements of the exchange- and SIA matrices, we use the four-state energy mapping (4SM) methodology26,27 in which we calculate the energies of several spin configurations of the system from first prin- ciples using density functional theory (DFT), and map these energies on their corresponding Heisenberg Hamil- tonians, setting up a system of equations from which the magnetic parameters can be derived.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The implementa- tion of the needed DFT calculations is thoroughly dis- cussed in the supplementary material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 The spin-wave dispersion relations are calculated nu- merically using the open-source code spinW ,39 in which we have implemented our Heisenberg Hamiltonian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This code is based on linear spin-wave theory, which is a good approximation assuming spin fluctuations are small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This condition is comfortably satisfied at low tempera- tures, significantly below the critical temperature (Curie or N´eel) of the long-range magnetic order at hand.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Nu- merical diagonalization of the Heisenberg Hamiltonian in reciprocal space yields the spin-wave dispersion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' MONOLAYER A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Crystal structure and magnetic parameters The crystal structure of monolayer CrI3 is depicted in Figure 1(a,e).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Monolayer CrI3 comprises one honeycomb layer of chromium atoms sandwiched between two layers of iodine atoms, where each chromium atom is octahe- drally coordinated with six iodine atoms, and each iodine atom connects two chromium atoms through an ≈ 90◦ Cr-I-Cr bond.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' After structural relaxation using DFT, we find a in-plane lattice constant of a = 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='919 ˚A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To characterize the magnetic interactions in CrI3, we perform a 4SM analysis in order to obtain the ele- ments of the exchange and SIA matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In Table I, we report the average nearest-neighbor (NN) and next- nearest-neighbor (NNN) intralayer exchange, Kitaev and DMI parameters for monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A full summary of the exchange parameters of all the individual pairs can be found in the supplementary material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 From the calculated parameters, it becomes clear that both the NN and the NNN exchange interactions are anisotropic and FM, with the NN one delivering the dom- inant contribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In agreement with literature,41,42 we find that the material’s out-of-plane magnetic anisotropy originates mainly from the NN exchange anisotropy, with a smaller contribution of ⟨Azz ii ⟩ = -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 meV due to the 3 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 1: Crystal structure of monolayer and bilayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Top view (a-d) and side view (e-h) of monolayer (1L) CrI3 (a,e), and bilayer (2L) CrI3 with an AB-stacking (b,f), AB’-stacking (c,g) and AA-stacking (d,h).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the sake of clarity, atoms of the same type are assigned a different color in the top and the bottom layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the bottom (top) layer, the chromium and iodine atoms are depicted with blue (dark blue) and yellow (orange) spheres respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The unit cell is marked with a solid black line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' All crystal structures were plotted using VESTA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='40 Panel (i) depicts the corresponding first Brillouin zone and high-symmetry points for 2D systems with a hexagonal lattice.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' SIA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The SIA is characterized by a single parameter ow- ing to the material’s three-fold rotational symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The NN interactions deliver no net contribution to the DMI since the inversion symmetry of the material is upheld.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, this symmetry is not present between NNN sites, resulting in a small yet non-zero DMI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Notice that, in CrI3, the DMI, the Kitaev interaction and the SIA all originate from the large SOC arising due to the heavy I ligands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='28,41,42 B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Spin-wave dispersion Figure 2(a) depicts the spin-wave dispersion of mono- layer CrI3 along the high-symmetry directions of the first Brillouin zone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Two distinct branches are present, as is expected for a unit cell containing two magnetic atoms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' At the Γ-point, the dispersion is gapped below the lower branch due to the magnetic anisotropy of the material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The gap has a size of ∆Γ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='41 meV and is an essen- tial prerequisite for the existence of long-range magnetic order in 2D at finite temperatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='41 The latter can be TABLE I: Magnetic parameters for monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Summary of the most important magnetic parameters in monolayer CrI3, including the exchange- and Kitaev constants Jij and Kij, and the size of the DMI-vectors |Dij|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' JNN KNN |DNN| JNNN KNNN |DNNN| (meV) (meV) (meV) (meV) (meV) (meV) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='35 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='74 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='17 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 seen by considering the total number of magnons excited at temperature T, which is given by: N = � D(ωk) e¯hωk/kBT − 1dωk, (3) with D(ωk) the magnon density of states, which is con- stant in 2D, kB the Boltzmann constant, and ωk the spin-wave frequencies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' When the dispersion is gapless, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' in the absence of magnetic anisotropy, this integral will diverge for ωk = 0, preventing long-range 2D mag- netic order at non-zero temperature in accordance with the Mermin-Wagner theorem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='43 The lower energy ‘acoustic’ branch displays quadratic behavior near the Γ-point and is associated with an in- phase precession of the spins [see Figure 2(b)], while the higher energy ‘optical’ branch is associated with an out- of-phase precession of the spins [see Figure 2(c)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The two branches meet at the K-point where they are sep- arated by a small bandgap of ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 meV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' At the K’-point we find a gap of the same size, since the sub- lattice symmetry is upheld.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The origin of this Dirac gap is partially attributed to the NNN DMI and partially to the Kitaev interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' At the K-point, the spins will precess at 120◦ angles to each other, as is shown in Figure 2(d,e) for respectively the lower- and higher branch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' If one would only consider a purely isotropic NN exchange, these two states would be energetically degenerate resulting in a Dirac point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, introducing Kitaev interactions and/or a NNN exchange term with a non-zero DMI, lifts the mutual degeneracy between the modes resulting in a bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=" More specifically, it’s the out-of-plane component of 1L AB-2L AB'-2L AA-2L (a) (b) (c) (d) (h) (i) b' K ki4 (a) (b) (c) (d) (e) FIG." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2: Spin-wave dispersion of monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a) Spin-wave dispersion along the high-symmetry directions of the first Brillouin zone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A small Dirac gap of ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 meV is present at the K-point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding Chern numbers are indicated for each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (b) and (c) display the spin-wave modes at the Γ-point for the lower- and higher branch respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (d) and (e) display a schematic top-view of the spin-wave modes at the K-point for the lower- and higher branch respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' the DMI that lies at the origin of the magnonic bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The size of this component intrinsically present in CrI3 is rather small (|Dz ij| = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 meV), also resulting in a small bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, external tuning that breaks the inversion symmetry, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' the presence of a substrate, electric gating or (non-uniform) strain, might induce ad- ditional DMI that could potentially increase the size of the bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In fact, by (artificially) increasing the DMI in our simulations, we verified that the bandgap can be ’tuned’.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' As shown in the supplementary material,29 the bandgap scales almost linearly with the NNN DMI when all the other parameters are kept constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Tun- ing the magnonic bandgap in 2D materials under exter- nal stimuli poses an interesting direction for future re- search, as the size of the bandgap can influence other material properties like as the magnon Hall conductiv- ity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, note that increasing the DMI may lead to non-collinear magnetization textures, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' spin cy- cloids or magnetic skyrmions, which will fundamentally change the magnonic behavior in the material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44,45 More- over, when the DMI is set to zero in our calculations, the bandgap does not fully vanish, suggesting that there is a second mechanism at work, which we identify to be the Kitaev interaction between NN spins.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the sup- plementary material,29 we show that one can tune the bandgap by artificially changing Kij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, varying the strength of the Kitaev interaction influences the over- all shape of the dispersion, whereas changing the DMI mainly influences the dispersion around the K-point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Topology Non-trivial band topology arises only in systems where non-zero Chern numbers predict the existence of edge states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The Chern number is a topological invariant with an integer value that is defined for the nth band as: Cn = 1 2πi � BZ Ωnk d2k, (4) in which the Berry curvature can be calculated as Ωnk = i � n′̸=n ⟨n |∂k ˆHk |n′ ⟩⟨n′ |∂k ˆHk |n⟩ (λnk − λn′k)2 , (5) with λnk and |n⟩ respectively the eigenvalues and eigen- vectors of the Heisenberg Hamiltonian ˆHk in reciprocal space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For systems that are gapless or show a trivial bandgap, the Chern numbers vanish.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In this work, we calculate Chern numbers according to the link-variable method detailed in Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' [46] for a discretized Brillouin zone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Applying this approach to the magnonic dispersion of monolayer CrI3, we find Chern numbers of Cn = ±1 for respectively the upper and lower band, as shown in Figure 2, classifying the material as a TMI with a non- trivial topological bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We attribute the origin of the topology to the breaking of time-reversal symmetry due to the spontaneous magnetization of CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 Thus, the topological nature of the bands persists in monolayer CrI3, be it with a significantly smaller bandgap compared to bulk CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='8 IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' BILAYER A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Crystal structure and magnetic parameters Bilayer CrI3 can be constructed by stacking two mono- layers on top of each other in a commensurate manner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The three different stacking orders that we consider in this work are shown in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In analogy to Sivadas et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=',48 we refer to those stacking orders as AB (rhom- bohedral), AB’ (monoclinic), and AA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The former two stackings correspond to respectively the low-temperature 20 一 1 1 1 Ci = +l 15 1 (meV) 1 1 K 10 1 C2 1 E 1 1 5 1 1 一 1 1 一 1 0 I K M TB A A B B A A B B AB A A B B A A B B A5 TABLE II: Structural and magnetic parameters for bilayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Summary of the most important stucutural and magnetic parameters in bilayer CrI3, including the lattice constant a and interlayer distance d, the average exchange- and Kitaev constants ⟨Jij⟩ and ⟨Kij⟩, the average size of the DMI-vectors ⟨|Dij|⟩, the DFT energy difference between the bilayer with an AFM and a FM interlayer ordering, and the average SIA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' a d ⟨JNN⟩ ⟨KNN⟩ ⟨|DNN|⟩ ⟨JNNN⟩ ⟨KNNN⟩ ⟨|DNNN|⟩ EAFM − EFM ⟨Azz ii ⟩ (˚A) (˚A) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) AB 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='915 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='400 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='45 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 AB’ 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='914 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='430 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='45 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 AA 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='908 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='505 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='42 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='84 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 and the high-temperature phases of CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 In the AB- stacking, the layers are stacked in such a way to place the chromium atoms in one layer above the void in the chromium honeycomb of the adjacent layer, analogously to a Bernal-stacked graphene bilayer [Figure 1(b,f)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The AB-stacking can be transformed to an AB’-stacking by sliding one of the layers by a third of the lattice vector along the zigzag direction [Figure 1(c,g)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Alternatively, by sliding one of the AB-stacked layers by a third of the lattice vector along the armchair direction, we obtain an AA-stacked bilayer in which each atom in the top layer is placed exactly above its bottom layer counterpart [Fig- ure 1(d,h)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' As shown in Table II, the different stacking orders show relatively similar lattice constants and interlayer distances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, changes in interatomic distances and (super-)superexchange bonding angles result in a differ- ent interlayer magnetic coupling, such that the AB and AA stackings prefer a FM ordering between the layers while the AB’-stacking slightly favors an AFM one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The latter is indicated in Table II by the DFT energy differ- ence between AFM and FM phases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In agreement with earlier work,48,50–52 we find that the overall ground-state of the system is a FM-ordered AB-stacked bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the supplementary material,29 we discuss the stacking- dependence of the interlayer ordering in more detail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In Table II, we also summarize the predominant mag- netic parameters for the CrI3 bilayers calculated with the 4SM method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A full overview of all the calculated param- eters for each specific pair can be found in the supple- mentary material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 For all stacking orders, the NN in- tralayer exchange interaction is anisotropic and strongly FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This anisotropy, together with the SIA, causes the spins to prefer an out-of-plane orientation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Due to the rotational symmetry in the AB and AA-stackings, the SIA matrix is reduced to only one parameter ⟨Azz ii ⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB’-stacked bilayer, this symmetry is absent requir- ing a calculation of the full SIA-matrix, however, ⟨Azz ii ⟩ will still be the dominant parameter, as most of the other matrix elements are very small or vanish.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To quantify the interlayer coupling, we calculated the interlayer NN and NNN exchange matrices for all stack- ings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the AB’-stacking, we also calculate the third nearest-neighbor (3NN) interlayer exchange, for the other stackings this contribution is negligible as is demon- strated in the supplementary material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 In the AB- and AA-stacked bilayers, all NN and NNN interlayer ex- change interactions are FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, the exchange pa- rameters for the AB-stacking are significantly stronger than for the AA-stacking, resulting in a stronger prefer- ence for a FM ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In contrast, for the AB’-stacked bilayer, there is a competition between the NN exchange which is FM and the NNN and 3NN exchange interac- tions which are AFM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Overall, this results in a weak AFM interlayer ordering, which is in agreement with earlier theoretical and experimental studies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='2,48,50–57 Interestingly, the sublattice symmetry is broken in the AB- and AB’ stackings, leading to a difference in out-of- plane exchange interactions ∆Jzz = |Jzz A − Jzz B | between sublattices A and B of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='92 meV for the AB-stacking and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 meV for the AB’-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Note that Jzz A and Jzz B are the sum of the out-of-plane exchange components of all interacting spin pairs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The difference ∆Jzz is sub- stantial for the AB-stacking because one sublattice has six stronger NNN interactions while the other sublattice has one weaker NN coupling and only three NNN in- teractions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the AA-stacking, there is no exchange difference since the sublattice symmetry is preserved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The intralayer Kitaev constants in the bilayers are sim- ilar in size compared to the monolayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the AB- and AB’-stacking, the NN Kitaev interaction is anisotropic, leading to different values for each bond, which is at- tributed to symmetry breaking due to the stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Since the NN Kitaev interaction is much stronger than the NNN and the interlayer ones, it’s the only contribution having a significant influence on the spin-wave dispersion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Unlike the monolayer system, the NN intralayer DMI is now non-zero, and originates from the inversion symme- try breaking due to stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similarly to the monolayer case, a non-zero NNN DMI arises.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In all stackings, the interlayer DMI will be very small or completely absent, having a limited influence on the dispersion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Spin-wave dispersion of bilayers with FM interlayer order Using the magnetic parameters calculated with the 4SM method, we compute the spin-wave dispersion for the three stacking orders considered in this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In Fig- ure 3, we display the results for bilayer CrI3 with different stackings, all with the FM interlayer ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In a bi- 6 (a) AB (b) AB’ (c) AA (d) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3: Spin-wave dispersion for bilayer CrI3 with a FM interlayer ordering in different stacking configurations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a,b,c) Spin-wave dispersion along the high-symmetry directions of the first Brillouin zone for respectively the AB, AB’, and AA- stacked bilayers with a FM interlayer ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In all stackings, a direct magnonic bandgap opens at the K-point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding Chern numbers are indicated for single bands, and composite Chern numbers for degenerate bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (d) schematically displays the corresponding spin-wave modes at the Γ-point for each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' layer, the unit cell contains four magnetic atoms, leading to four branches in the dispersion, two ‘acoustic’ and two ‘optical’ ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The corresponding spin-wave modes at the Γ-point of each branch are indicated in Figure 3(d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The energy difference between these different modes is proportional to the strength of the interlayer coupling, hence the large separation for the AB-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Each stacking has a gap at the Γ-point, signaling that FM order is stable in each of them at finite temperatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The gaps have sizes of ∆Γ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44 meV, ∆Γ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 meV and ∆Γ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='35 meV for respectively the AB, AB’, and AA-stackings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In all stackings, we also observe direct magnonic bandgaps at the K-point of ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 meV, ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='21 meV, and ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 meV for respectively the AB, AB’, and AA-stackings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similarly to the mono- layer, we attribute the origin of these gaps to a combina- tion of NNN DMI and NN Kitaev interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Notice that the first Brillouin zone contains two in- equivalent high-symmetry points K and K’ [see Fig- ure 1(i)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB and AB’ stackings, where the sub- lattice symmetry is broken, we see a different behavior of the dispersion at each point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the former, a bandgap of only 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='42 meV opens close to the K’-point (compared to ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 meV).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB’-stacking, we see an indirect band-crossing at the K’-point, as is often seen in semi- metals, and thus, no bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' On the other hand, in the AA-stacked bilayer, the sublattice symmetry is pre- served, resulting in exactly the same dispersion at both the K- and K’-points.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='Spin-wave dispersion of bilayers with AFM ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='interlayer order ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='By comparing the dispersion of the bilayers with FM ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='interlayer ordering with the dispersions of the bilayers ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='n= 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='n=3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='n = ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='n = 4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='B ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='B ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='B ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='B ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='B ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='B20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='neV) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 C3 = 0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='K ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='E ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='I ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='K ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='M ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='T20 1 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 (meV) 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='K ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='E ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='K ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='M ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='T20 iCi = +l ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='I C2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='=-1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='(meV) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='C3 =+1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='K ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 E 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='一 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 C4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='K ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='M7 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='(a) AB ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='(b) AB’ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='(c) AA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='(d) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4: Spin-wave dispersion for bilayer CrI3 with an AFM interlayer ordering in different stacking configura- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a,b,c) Spin-wave dispersion along the high-symmetry directions of the first Brillouin zone for respectively the AB, AB’ and AA-stacked bilayers with an AFM interlayer ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB and AB’-stackings a direct bandgap opens at the K-point, meanwhile in the AA-stacked bilayer we observe a Dirac point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding composite Chern numbers are indicated for the bands and are all equal to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (d) schematically displays the corresponding spin-wave modes at the Γ-point for each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' with AFM interlayer ordering [see Figure 4], it becomes clear that there is a strong dependence of the magnonic properties of CrI3 on the interlayer ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' First and foremost, notice that for the AA and AB stackings, there is a region close to the Γ-point where the acous- tic branches are zeroed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Consequently, there is no gap at the Γ-point and the integral in equation (3) will diverge, signaling that AFM order is unstable in these stackings at non-zero temperatures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, in the AB’-stacking, there is a gap of ∆Γ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 meV, meaning that AFM order is stable in the monoclinic phase, which is in agree- ment with experimental observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='2,49,53–56 Further, also notice that, in contrast to the FM-ordered bilayers, we see a degeneracy of the two acoustic branches and the two optical branches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Only at the K-point there are notable energy differences between the bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The dispersions of different bilayers with AFM inter- layer order are characterized by bandgaps of respectively ∆K = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 meV and ∆K′ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 meV for the AB- stacking, and ∆K = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 meV and ∆K′ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 meV for the AB’-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' At the K and K’-points in the AA- stacking case, there is no bandgap, but instead one finds a Dirac cone combined with two anti-crossing branches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Topology When two or more bands are degenerate, crossing or touching, it is no longer possible to assign individual Chern numbers to each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Instead, we define a com- posite Chern number Cn⊕n′, jointly shared by the degen- erate bands, and calculated as detailed in Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' [58].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' As shown in Figure 3, there is a strong dependence of the Chern number on the stacking configuration in the FM-ordered bilayers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Although we are expecting a non-trivial topology of the bandgaps in the FM bilay- 20 1 一 一 一 15 neV) 10 K E 一 一 C34 = 0 1 1 5 1 1 1 1 0 K M T20 1 1 1 15 neV) 1 1 (m 1 10 E 34 = 0 1 1 5 1 1 1 1 一 1 0 K M20 1 1 一 一 一 15 一 neV) 一 E K 10 一 E 1 1 1 5 1 1 1 1 1 1 1 1 0 K M Tn=2 n=3 n = n =4 B B B B B B8 ers,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' caused by the breaking of time-reversal symmetry due to the spontaneous magnetization,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' only the AA- stacking shows non-zero Chern numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Thus, the AA- stacked CrI3 bilayer can be classified as a TMI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB’-stacking there is no bandgap at the K’-point, hence, the Chern number is undefined and the bands are not topological.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB-stacking, all Chern numbers are equal to zero, meaning that the bandgap is of trivial nature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We attribute the lack of topology in the lat- ter stacking to the exchange difference ∆Jzz, caused by the breaking of sublattice symmetry, which is very large for the AB-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the supplementary material,29 we show that by artificially reducing ∆Jzz in our simula- tions, which also decreases the size of the bandgap at the K-point and the K’-point, we can induce a topological phase transition to a state with non-zero Chern numbers of C1⊕2 = +1, C3 = −1 and C4 = 0, which confirms the influence that sublattice symmetry can have on the topology of magnonic bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='20 In bilayers with AFM interlayer order, the bands are two-by-two degenerate, meaning that one can only define composite Chern numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the case of AA-stacking, there is no bandgap and, thus, the Chern number is unde- fined and the bands display no topological behavior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The composite Chern numbers for the other two stackings turn out to be zero for all considered bands, which can be related to the conservation of effective time-reversal sym- metry in AFM materials, as the layers are time-reversed copies of each other.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 However, in the next section, we will show that breaking this symmetry by an applied magnetic field leads to emergent topological states with non-zero Chern numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Effect of an external magnetic field In this section, we explore whether the magnonic dis- persion and band topology of bilayer CrI3 can be tuned by applying an out-of-plane external magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the case of a monolayer or the bilayers with FM interlayer order, there is only a trivial effect due to an applied magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Namely, the whole dispersion will uniformly shift up or down depending on the orientation of the applied field with respect to the magnetization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In contrast, for bilayers with AFM interlayer order, an external magnetic field will lift the degeneracy between branches, shifting two branches up and two branches down in energy, and leading to additional interesting features in the dispersion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similar band shifts were ob- served by Cenker et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=',4 who reported the splitting of degenerate Raman peaks in the optical branches of an AFM-ordered monoclinic CrI3 bilayer, after applying an external magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, note that applying a magnetic field to AFM-ordered bilayers should be done carefully, as the interlayer magnetic state will switch to the FM one for sufficiently strong fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Here, we cal- culate the spin-wave dispersion for the different stack- ing scenarios under sufficiently small applied field, where FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 5: Spin-wave dispersion for an AB’-stacked CrI3 bilayer with an AFM interlayer ordering under the influence of an external magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Left and right panels compare the dispersion near the K-point and K’-point respectively, for an applied magnetic field of B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Under influence of the magnetic field, blue bands have shifted up and red bands down in energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' AFM interlayer order is safely stable (see supplementary material).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 Especially the AFM-AB phase is very sensi- tive, and changes to a FM interlayer order even for very weak applied field - hence is excluded from our calcula- tions in this section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the case of AB’ stacking, the size of the bandgap will decrease after applying the magnetic field, reaching a minimum of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='14 meV at the K-point and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 meV at the K’-point for a field of B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 T, as shown in Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' At the K’-point, one sees that, as the bands approach each other, they do not entirely touch or cross - instead we observe band inversion combined with a small bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Band inversion is an effect often also present in electronic topological insulators,59 and is typically caused by the SOC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For fields applied in the opposite direction, we see analogous behavior, as now the two other bands are shifted upwards and the previous two downwards.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For fields larger than B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 T, the AFM interlayer ordering changes to a FM one (see supplementary material).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 In Figure 6, we show the influence of an external mag- netic field with a magnitude of B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 T on the disper- sion of an AFM-ordered AA-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Apply- ing the field shifts the Dirac node upwards or downwards depending on the polarity of the field, which leads to the formation of a closed Dirac magnon nodal-line loop at the crossover point of the red and blue bands in Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The latter is a closed one-dimensional loop around the K-point where two bands cross, exactly analogous to the nodal-lines described for Dirac semimetals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='60 Decreasing the field leads to a smaller shift of the branches, result- ing in nodal-line loops with a smaller radius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For fields larger than 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 T, the AFM interlayer ordering changes to a FM one (see supplementary material).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content="29 As mentioned earlier, in the absence of applied field, all AFM bands show composite Chern numbers equal to zero, meaning that the bandgaps have a trivial topol- 16 15 14 (meV) 13 E 12 11 10 K16 15 14 (meV) W 13 K E 12 11 10 1 K'9 B = 0." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 T B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 T FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 6: Spin-wave dispersion for an AA-stacked CrI3 bilayer with an AFM interlayer ordering under the influence of an external magnetic field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Left and right panels compare the dispersion at the K-point for applied fields of respectively B = 0 T and B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Under influence of the magnetic field, blue bands have shifted up and red bands down in energy, forming magnon nodal-lines at the crossing points of the red and blue curves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ogy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Interestingly, after applying the magnetic field on the AB’-stacked bilayer, non-zero Chern numbers emerge as C1,4 = +1 and C2,3 = −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In other words, by applying a magnetic field, which breaks the effective time-reversal symmetry of the material, a topological phase transition can be induced.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In contrast, for the AA stacking, the (composite) Chern numbers remain undefined after ap- plying the magnetic field, as the Dirac cone stays present.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' CONCLUSIONS We characterized the magnonic dispersion for intrinsi- cally ferromagnetic monolayer and bilayer CrI3 using lin- ear spin-wave theory combined with a Heisenberg model parameterized from first principles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We showed that the monolayer is characterized by a small Dirac-gap in the spin-wave dispersion, sourced to a specific combination of next-nearest-neighbor (NNN) DMI and nearest-neighbor Kitaev interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Non-zero Chern numbers are asso- ciated with the bands, indicating the topological nature of the bandgap, and suggesting that monolayer CrI3 is a topological magnon insulator (TMI).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In bilayer CrI3, still with dominantly ferromagnetic intralayer interac- tions, we demonstrated a dependence of the dispersion on the geometric stacking order and the interlayer mag- netic ordering, opening a bandgap for the AB stacking (for both FM and AFM interlayer order), the AB’ stack- ing (only AFM), and the AA stacking (only FM), mean- while the FM-ordered AB’ stacking shows an indirect band crossing, and the AFM-ordered AA stacking ex- hibits a Dirac point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similarly to the monolayer case, we identified the DMI and Kitaev interactions as the lead- ing causes behind the opening of the bandgap, both being modulated by the stacking order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The latter contradicts earlier work on bulk CrI3 which claimed that only the NNN DMI and, thus, not the Kitaev interaction, lies at the origin of the Dirac gap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 Interestingly, we found that the Chern number, and consequently the magnonic band topology, depends on the stacking configuration and the interlayer magnetic order, vanishing for all studied cases except in the FM-ordered AA bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Thus, depending on the stacking order and the interlayer magnetic order, bilayer CrI3 is classified as either a topological magnon insulator, a trivial magnon insulator, or a magnon Dirac material.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Finally, we showed that the dispersion of the bilayers with AFM interlayer order can be tuned by an external out-of-plane magnetic field, changing both size and topology of the bandgap for the AB’-stacked bilayer, and introducing closed nodal-line loops in the dispersion of the AA bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The here demonstrated presence of tunable bandgaps of possibly topological nature in bilayer CrI3 recommend it as a TMI that can serve as a platform to investigate tunable magnon Hall- and spin Nernst effects in 2D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Our results could be verified experimentally by investigating the thermal magnon Hall effect in monolayer and bilayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Both the DMI and Kitaev interactions originate from the spin-orbit coupling, which is relatively strong in CrI3 and, thus, lies at the origin of the topological bandgap.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' If one wants to achieve a gapless spin-wave dispersion, we suggest looking at 2D magnets with a weaker SOC, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' CrBr3 or CrCl3, which are good can- didates to host a Dirac point in the monolayer limit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In order to further tailor the magnonic bandgap, one can induce and tune the DMI in CrI3, or other 2D magnets, by external stimuli such as gating, (non-uniform) strain, heterostructuring, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Furthermore, our work demon- strates that stacking vdW monolayers, be it in regular bilayers or, in future work, multilayers, or even (moir´e) heterostructures, poses a viable route to achieve broadly tunable magnonic properties in 2D materials and van der Waals homo- and heterostructures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Acknowledgments We thank D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ˇSabani, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jorissen and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Shafiei for use- ful discussions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This work was supported by the Research Foundation-Flanders (FWO-Vlaanderen) and the Spe- cial Research Funds of the University of Antwerp (BOF- UA).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The computational resources used in this work were provided by the VSC (Flemish Supercomputer Center), funded by Research Foundation-Flanders (FWO) and the Flemish Government – department EWI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 15 1 1 1 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1 1 1 14 1 1 1 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1 1 13 E 1 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1 1 1 12 1 1 1 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1 1 1 11 K15 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 14 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 (meV) 13 E 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 12 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1 11 K10 ∗ Electronic address: milorad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='milosevic@uantwerpen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='be 1 M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gibertini, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Koperski, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Morpurgo, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Novoselov, Nature Nanotechnol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 14, 408 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Huang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Clark, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Navarro-Moratalla, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Klein, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cheng, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Seyler, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhong, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Schmidgall, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' McGuire, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cobden, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature 546, 270 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3 W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jin, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kim, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ye, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Li, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rezaie, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Diaz, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Siddiq, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wauer, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Li, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 9, 5122 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cenker, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Huang, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Suri, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Thijssen, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miller, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Song, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Taniguchi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Watanabe, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' McGuire, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xiao, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, Nature Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 17, 20–25 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 5 R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Menezes, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ˇSabani, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bacaksiz, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' de Souza Silva, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miloˇsevi´c, 2D Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 9, 025021 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 6 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Barman, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gubbiotti, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ladak, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Adeyeye, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Krawczyk, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gr¨afe, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Adelmann, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cotofana, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Naeemi, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Vasyuchka, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' : Condens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Matter 33 413001 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7 Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cai, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bao, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gao, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ma, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Shangguan, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Si, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dong, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 104, L020402 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 8 L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chung, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gao, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Stone, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kolesnikov, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Huang, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dai, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' X 8, 041028 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 9 L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chung, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Stone, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kolesnikov, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Winn, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Garlea, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Abernathy, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gao, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Au- gustin, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Santos, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dai, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' X 11, 031047 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 10 F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhu, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' dos Santos, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Song, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mueller, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Schmalzl, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Schmidt, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ivanov, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Park et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Adv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7, eabi7532 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 11 I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dzialoshinskii, Sov.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' JETP 5, 1259 (1957).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 12 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Moriya, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 120, 91 (1960).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 13 L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Stone, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kolesnikov, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Winn, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Shon, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dai, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chung, 2D Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 9, 015006 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 14 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Schneeloch, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Tao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cheng, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Daemen, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhang, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Louca, npj Quantum Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7, 66 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 15 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Owerre, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' : Condens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Matter 28 386011 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 16 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Owerre, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 120, 043903 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 17 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fransson, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Black-Schaffer, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Balatsky, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 94, 075401 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Owerre, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 1 025007 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 19 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Pershoguba, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Banerjee, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lashley, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Park, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ˚Agren, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Aeppli, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Balatsky, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' X 8, 011010 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 20 H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kim, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kim, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 106, 104430 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 21 E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Aguilera, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jaeschke-Ubiergo, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Vidal-Silva, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Torres, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Nunez, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 102, 024409 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 22 L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhang, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhu, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Go, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lux, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' dos Santos, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lounis, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Su, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bl¨ugel, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mokrousov, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 103, 134414 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 23 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Owerre, Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rep.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7, 6931 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 24 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Owerre, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 94, 094405 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 25 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Costa, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Santos, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Peres, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fern´andez-Rossier, 2D Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7, 045031 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 26 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ˇSabani, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bacaksiz, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miloˇsevi´c, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 102, 014457 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 27 H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xiang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lee, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Koo, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gong, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Whangbo, Dalton Transactions 42, 823 (2013).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 28 C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Feng, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xiang, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bellaiche, npj Comput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4, 1 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 29 See Supplemental Material at [URL will be inserted by publisher] for additional tables, figures and discussions of our results, and which cites Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='.2,26,30–38,40,48,50,51,53–56 30 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hafner, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 47, 558 (1993).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 31 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Furthm¨uller, Computational materials science 6, 15 (1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 32 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Furthm¨uller, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 54, 11169 (1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 33 P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bl¨ochl, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 50, 17953 (1994).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 34 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Perdew, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Burke, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ernzerhof, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 77, 3865 (1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 35 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Grimme, Journal of Computational Chemistry 27, 1787 (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 36 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Steiner, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Khmelevskyi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Marsmann, and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 93, 224425 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 37 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dudarev, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Botton, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sevrasov, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Humphreys, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sutton, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 57, 1505 (1998).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 38 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M¨uller, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hoffmann, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dißelkamp, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sch¨urhoff, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mavros, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sallermann, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kiselev, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J´onsson, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bl¨ugel, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 99, 224414 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 39 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Toth, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lake, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' : Condens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Matter 27, 166002 (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 40 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Momma, and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Izumi, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cryst.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 44, 1272 (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 41 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lado, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fern´andez-Rossier, 2D Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4, 035002 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 42 C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bacaksiz, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ˇSabani, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Menezes, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miloˇsevi´c, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 103, 125418 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 43 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mermin, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wagner, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 17, 1133 (1966).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 44 F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ma, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhou, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Braun, and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lew, Nano Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 15, 4029 (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 45 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Weber, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fobes, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Waizner, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Steffens, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Tucker, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B¨ohm, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Beddrich, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Franz, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gabold, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bewley et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Science 375, 1025 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 46 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fukui, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hatsugai, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Suzuki, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jpn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 74, 1674 (2005).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 47 P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' McClarty, Annu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Condens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Matter Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 13, 171 (2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 48 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sivadas, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Okamoto, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fennie and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xiao, Nano Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18, 7658 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 49 M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' McGuire, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dixit, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cooper and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sales, Chem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 27, 612 (2015).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 50 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jeong, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yoon, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ryee and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Han, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Materials 3, 031001 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 51 P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhong, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yuan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lu and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ji, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 99, 144401 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 52 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Soriano, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cardoso and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fern´andez-Rossier, Solid State Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 299, 113662 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 53 L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Thiel, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Tschudin, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rohner, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Guti´errez-Lezama, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ubrig, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gibertini, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Giannini, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Morpurgo and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Maletinsky, Science 364, 973 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 54 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Song, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fei, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yankowitz, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lin, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hwangbo, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhang, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Taniguchi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Watanabe et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18, 1298 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 55 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Li, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sivadas, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Weber, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Goldberger, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Watanabe, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Taniguchi, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fennie et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18, 1303 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 56 W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sun, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wu, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gao, Science 366, 983 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 11 57 Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Guo, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zeng, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lin and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mou, Nanomat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 11, 2509 (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 58 R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhao, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xie, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Huang, and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sha, Opt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Express 28, 4638 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 59 Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cheng, and U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Schwingenschl¨ogl, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 85, 235401 (2012).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 60 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wehling, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Black-Schaffer, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Balatsky, Adv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 76, 1 (2014).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Stacking-dependent topological magnons in bilayer CrI3 SUPPLEMENTARY MATERIAL M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Soenen,1 C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bacaksiz,1, 2, 3 R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Menezes,1 and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miloˇsevi´c1, ∗ 1Department of Physics & NANOlab Center of Excellence, University of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium 2Bremen Center for Computational Material Science (BCCMS), Bremen D-28359, Germany 3Computational Biotechnology, RWTH Aachen University, Worringerweg 3, 52074 Aachen, Germany (Dated: January 9, 2023) I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' DFT CALCULATIONS The DFT calculations are performed within the Vienna ab initio simulation package (VASP)1–3 using a generalized gradient approximation (GGA) functional and the projector augmented wave (PAW) method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='4 We opt for the Perdew- Burke-Ernzerhof (PBE)5 exchange-correlation functional in combination with the D2 method of Grimme6 to account for a vdW correction term.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A term due to the SOC7 is added ad hoc to the DFT Hamiltonian where necessary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the Brillouin zone integration, we use a Gaussian smearing of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 eV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Due to the periodic boundary consitions, we need to use 3 × 3 × 1 supercells in the 4SM calculations in order to calculate the NNN and 3NN interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To limit the computational cost of the 4SM calculations, we use a plane-wave energy cutoff of 300 eV and a 3×3×1 grid for the k-point sampling, for which we deem the systems sufficiently converged within reasonable computing times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In VASP, periodic boundary conditions are implemented automatically.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hence, during the calculations, we choose an out-of-plane unit cell distance of c = 15 ˚A for the monolayer and c = 26 ˚A for the bilayer ensuring a large enough vacuum distance between the materials and their periodic images, effectively minimizing the interaction between them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Since the system contains localized (strongly correlated) d-electrons, we implement the GGA+U method in the rotationally invariant form proposed by Dudarev et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=',8 i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' we add an on-site Coulomb interaction of Ueff = U J = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='8 eV to the d-orbitals of the chromium atoms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To confirm that the chosen Ueff gives the desired qualitative description of the magnetic properties of CrI3, we plot the energy difference between the AFM and FM phases as a function U - J (with J = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 eV) for all stacking orders [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the AB-stacked bilayer (green curve), we find a strong preference for a FM interlayer ordering for any value of U - J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similarly, for the AA-stacking (black curve), we find a (smaller) preference for a FM interlayer ordering, except for very big U - J values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For the AB’-phase (magneta curve), however, we find a FM interlayer ordering for small U - J values which transitions to an AFM ordering for increasing U - J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similarily to Jiang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=',9 we choose values of U = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 eV and J = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 eV resulting in an effective parameter of U - J = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='8 eV which slightly favors an AFM-coupled AB’-stacking and a FM-coupled AB- and AA-stackings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This agrees with our value of U = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='8037 eV that we calculated with a linear response method for monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 1: Convergence test for the effective U-parameter used in the DFT calculations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The total energy difference between the AFM and FM phases of the three considered stacking orders of bilayer CrI3 as a function of the on-site Coulomb interaction U - J (with J fixed to 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='1 eV).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02502v1 [cond-mat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='mtrl-sci] 6 Jan 2023 14 12 AB 10 AB (meV) AA 8 EFM 6 4 0 2 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 3 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 4 U-J (eV)2 II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' STACKING DEPENDENCE OF THE INTERLAYER COUPLING A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Energetic analysis from first-principles To map out the stable phases of bilayer CrI3, we plot its energy as the top layer is shifted laterally [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 (a,b)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The structures are relaxed in the out-of-plane direction to find the optimal interlayer distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For both an AFM [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 (a)] and a FM [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 (b)] ordered bilayer, we find similar energy profiles with clear minima located at the naturally occurring rhombohedral (AB) and monoclinic (AB’) phases, and a local minimum at the AA stacking order, hence our choice for these stackings during this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To determine the the preferential interlayer magnetic order for each stacking, we plot the energy difference between the AFM and FM phases in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 (c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Stacking orders that prefer an AFM configuration show up in blue, phases with a preference for a FM ordering show up in red.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the groundstate, the system will take on an AB-stacking combined with a FM interlayer coupling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The AA-stacked bilayer also prefers a FM interlayer coupling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Although the energy difference is very small, the AB’-stacked bilayer prefers an AFM interlayer coupling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This small energy difference is better illustrated in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 (d) which depicts an intersection of the AFM and FM energy profiles along the pathway marked with a black dashed line in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 (a-c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The results reported here are in good agreement with earlier work by Jiang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 and Sivadas et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='.10 All results presented in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 were obtained from first-principles using DFT calculations implemented in VASP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a) (b) (c) (d) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2: Total energy of bilayer CrI3 as a function of a lateral shift with respect to an AB-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The total energy is depicted as a function of a lateral shift of the top layer along basis vectors a and b, for a bilayer with AFM (a) and FM (b) interlayer ordering respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Minima in the energy correspond to the AB, AB’ and AA-stacking orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The energy difference between the AFM and FM phases is shown in (c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Panel (d) shows the energy per Cr atom, with respect to the ground-state (AB,FM), along the transition pathway that is marked with a black dashed line in panels (a-c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Eafm (meV) 2/3 AB 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 AB\' AB" 1/3 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='065 X AA X AB AB X 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content="AB X AB' AB'X 1/3 b 2/3 AB 6." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='075 ×104Efm (meV) AB 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content="06 2/3 AB' AB'X 1/3 6." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='065 X AB X AA X AB AB X 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content="07 XAB X AB AB'X 1/3 b 6." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='075 2/3 AB 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content="08 × 10415 2/3 AB AB' ABX 10 1/3 AB X AA X AB AB X 5 0 XAB X AB' AB'X 1/3 b 0 2/3 AB X 1 520 FM 一AFM Cr) 15 (mev 10 E 5 0 0 0." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='8 1 x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='s3 B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Atomistic spin-dynamics simulations The further investigate the dynamic stability of the different phases of bilayer CrI3, both with and without the presence of an external magnetic field, we perform atomistic spin dynamics (ASD) simulations, by numerically solving the Landau-Lifshitz-Gilbert (LLG) equation: ∂ˆSi ∂t = − γ (1 + α2) µ � ˆSi × Beff i + αˆSi × � ˆSi × Beff i �� , (1) in which γ is the gyromagnetic ratio, α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='001 the damping parameter, µ = 3µB the magnetic moment per Cr atom, and Beff i = −∂ ˆH/∂ˆSi the effective field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The latter is the resulting field due to all the magnetic interactions considered in the Heisenberg Hamiltonian (see main text).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The simulations are performed at T = 0 K on a 50 × 50 supercell for a duration of 105 iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The ASD simulations are implemented in the Spirit11 software package which is adapted to accommodate the Hamiltonian considered in this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the abscence of an external magnetic field, our LLG simulations show that a FM interlayer ordering is a (meta)stable state for each of the three stacking orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, an AFM ordering is only stable for the AB’ and AA-stackings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB-stacking, the structure always relaxes to a FM ordered state, even when the structure is initialized with an AFM one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' This spontaneous change in magnetization is attributed to the large energy difference between FM and AFM phases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In Figure 3, we report results for the magnetic field dependence of the AFM interlayer ordering, by plotting the magnetization for the top and bottom layer separately as a function of the applied field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The field is oriented in the out-of-plane direction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB-stacking [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3 (b)], the magnetization will always be fully saturated, even in the absence of a field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hence, an AFM interlayer ordering is not stable for this stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, for the AA-stacking [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3 (a)] and the AB’-stacking [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3 (c)], we see that the AFM ordering remains stable up to relatively high fields of respectively B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='6 T and B = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='9 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For higher applied fields, the magnetization in one of the layers will flip (the layer with opposite magnetization to the field), leading to a FM ordering oriented parallel to the field.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For a field with opposite orientation, we find similar results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a) AA-stacking (b) AB-stacking (c) AB’-stacking FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3: Magnetic field dependence of the interlayer ordering in the AFM ordered CrI3 bilayer for different stacking orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Magnetization of the top layer (red) and bottom layer (blue) of an AFM ordered CrI3 bilayer as a function of the applied magnetic field, for respectively the AA-stacking (a), AB-stacking (b) and the AB’-stacking (c).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The magnetization is normalized with respect to the saturation magnetization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 Mz (top) Mz (bot) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 B [T]1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 Mz (top) Mz (bot) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 /Ms 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 M 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 B [T]1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 Mz (top) Mz (bot) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='75 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='0 B [T]4 III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' MAGNETIC PARAMETERS In this section, we present the magnetic parameters – obtained through a 4SM analysis – for all the pairs necessary to characterize the magnonic behavior of monolayer and bilayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Tables I - IV, contain the magnetic parameters for respectively the monolayer [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4], the AA-stacking [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 5], the AB-stacking [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 6], and the AB’-stacking [Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For these four systems, we report parameters for the NN and NNN intra- and interlayer exchange and DMI [Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I - IV], the SIA, and the Kitaev interactions [Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The exchange- and DMI parameters are reported in Cartesian coordinates [Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' I - IV], afterwards in Table V, we consider local eigenbases that diagonalize the symmetric exchange matrices to quantify the Kitaev interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In all considered structures, the intralayer exchange is FM and anisotropic with the NN interaction delivering the dominant contribution, the NNN exchange also prefers a FM ordering but is weaker than the former.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The out-of-plane exchange anisotropy, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ∆ = ⟨Jzz ij ⟩ − ⟨Jαα ij ⟩ with α = {x, y}, is the most important contribution to the magnetic anisotropy of CrI3 and causes the spins to prefer an out-of-plane orientation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In monolayer CrI3, the NN DMI is equal to zero due to the material’s inversion symmetry.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Stacking two layers in a bilayer breaks this inversion symmetry leading to a non-zero NN DMI in all stacking orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Between intralayer NNN spins, there is no inversion symmetry resulting, for all structures, in a small non-zero DMI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The AB- and AA-stacked bilayers, both show a FM NN and NNN interlayer coupling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The strength of the interlayer coupling is significantly stronger for the AB-stacking as it has more interacting pairs that also interact more strongly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB’-stacked bilayer, there is competition between the FM NN interlayer coupling and the AFM NNN and 3NN interlayer coupling, resulting in an overall preference for an AFM interlayer ordering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For all stackings, the interlayer DMI is significantly weaker than the intralayer one, having almost no influence on the material’s properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Notice that, in this work, we only report 3NN parameters for the AB’-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In this stacking, the 3NN interlayer interaction is indispensable in order to achieve an overall AFM interlayer ordering as the ground state, in agreement with experiment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13–17 In all other cases, we can safely neglect the 3NN interactions since they deliver only minor contributions to the overall exchange.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To illustrate the latter, we calculate several 3NN parameters to demonstrate that their characteristic orders of magnitude are negligible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For example, in the AB-stacked bilayer we found 3NN intralayer exchange pairs of (Jxx 2−7, Jyy 2−7, Jzz 2−7) = (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02) meV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similar values were found for the monolayer and the other bilayers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hence, the 3NN intralayer interaction is negligible in all structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Regarding the 3NN interlayer exchange, we found values of (Jxx 7−16, Jyy 7−16, Jzz 7−16) = (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00) meV in the AB- and AA-stacking, values like this can also safely be neglected without significant influence on further results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In principle, the SIA matrix is symmetric resulting in six unique matrix elements that need to be calculated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, notice that, when there is a 3-, 4-, or 6-fold rotational symmetry around the out-of-plane axis, it suffices to calculate only one parameter Azz ii .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 For the monolayer, the AA-stacking, and the AB-stacking, in which this symmetry is upheld, we find values for ⟨Azz ii ⟩ of respectively -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 meV, -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 meV and -0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 meV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the AB’-stacked bilayer, where this symmetry is absent, we need to calculate the full SIA-matrix, however, all elements vanish except ⟨Azz ii ⟩ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 meV and ⟨Ayz ii ⟩ = ⟨Azy ii ⟩ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 meV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hence, for all structures, the SIA shows a preference for an out-of-plane orientation of the spins, delivering a small contribution to the material’s magnetic anisotropy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' To calculate the Kitaev constants, we consider for each spin pair, an eigenbasis {λ, µ, ν} that diagonalizes the corresponding symmetric exchange matrix, yielding three eigenvalues Jλ ij, Jµ ij, and Jν ij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In the case that Jλ ij ≈ Jµ ij ̸= Jν ij, the exchange constant is defined as Jij = (Jλ ij +Jµ ij)/2 and the corresponding Kitaev constant as Kij = Jν ij −Jij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The DMI in the local basis is found by projecting the Cartesian DMI vector along the {λ, µ, ν} directions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' We summarize the exchange and Kitaev constants in Table V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In all structures, the main contribution to the Kitaev interaction comes from the NN intralayer exchange.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 5 TABLE I: Exchange parameters for monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For each interacting pair of spins, the table contains the elements of both the NN and the NNN intralayer exchange matrices Jij, the out-of-plane exchange anisotropy ∆, and the components of the corresponding DMI-vectors Dij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' All parameters are given in Cartesian coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding pairs are indicated in Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' pair Jxx ij Jyy ij Jzz ij ∆ Jxy ij Jyx ij Jxz ij Jzx ij Jyz ij Jzy ij |Dx ij| |Dy ij| |Dz ij| |Dij| (i-j) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) NN (1-2) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='24 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='96 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (2-3) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='52 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='96 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (2-5) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='52 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='96 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 ⟨Jij⟩ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='79 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='79 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='96 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='17 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 NNN (1-3) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='67 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='73 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='66 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 (1-5) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='67 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='73 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='66 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 (3-5) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='76 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='66 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='66 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4: Labeled spin sites in monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Top view of a 2 × 2 × 1 supercell of monolayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chromium atoms are represented by green spheres.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Iodine atoms and bonds are not shown in the picture for the sake of simplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The supercell is marked with solid black lines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The crystal structures were drawn using VESTA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='19 Cr4 Cr8 Cr3 Cr7 b Cr2 Cr6 Cr1 Cr56 TABLE II: Exchange parameters for an AA-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For each interacting pair of spins, the table contains the elements of both the NN and the NNN intra- and interlayer exchange matrices Jij, the out-of-plane exchange anisotropy ∆, and the components of the corresponding DMI-vectors Dij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' All parameters are given in Cartesian coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding pairs are indicated in Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' pair Jxx ij Jyy ij Jzz ij ∆ Jxy ij Jyx ij Jxz ij Jzx ij Jyz ij Jzy ij |Dx ij| |Dy ij| |Dz ij| |Dij| (i-j) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) Intra NN (1-2) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='40 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='63 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='63 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 (2-3) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='52 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='60 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='41 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='22 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 (2-5) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='52 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='60 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='22 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='41 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 ⟨Jij⟩ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='88 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='88 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='17 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 NNN (1-3) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='59 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 (1-5) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='67 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 (3-5) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='59 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 Inter NN (1-9) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 NNN (7-12) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (7-14) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (7-16) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (a) Top view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (b) Side view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 5: Labeled spin sites in an AA-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Top view (a) and side view (b) of a 2 × 2 × 1 supercell of an AA-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chromium atoms in the top- and bottom layer are represented by green and blue spheres respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Iodine atoms and bonds are not shown in the picture for the sake of simplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The supercell is marked with solid black lines in panel (a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The crystal structures were drawn using VESTA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='19 Cr4 Cr8 Cr3 Cr7 Cr2 Cr6 Cr1 Cr5 Cr12 Cr16 Cr11 Cr15 a Cr10 Cr14 Cr9 Cr13Cr4 Cr8 Cr3 Cr7 b Cr2 Cr6 Cr1 Cr57 TABLE III: Exchange parameters for an AB-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For each interacting pair of spins, the table contains the elements of both the NN and the NNN intra- and interlayer exchange matrices Jij, the out-of-plane exchange anisotropy ∆, and the components of the corresponding DMI-vectors Dij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Since the sublattice symmetry is broken, parameters are given for the two sublattices (A,B) separately.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' All parameters are given in Cartesian coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding pairs are indicated in Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' pair Jxx ij Jyy ij Jzz ij ∆ Jxy ij Jyx ij Jxz ij Jzx ij Jyz ij Jzy ij |Dx ij| |Dy ij| |Dz ij| |Dij| (i-j) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) Intra NN (1-2) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='63 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 (2-3) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='43 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='93 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='86 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='43 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (2-5) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='84 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='53 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='42 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='22 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 ⟨Jij⟩ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='96 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='94 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='16 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 NNNA (1-3) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='53 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='58 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='52 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 (1-5) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='53 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='58 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='51 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (3-5) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='55 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='55 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='51 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NNNB (2-4) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='58 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 (2-6) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='67 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (4-6) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='59 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 Inter NNB (2-9) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='27 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 NNNA (7-10) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='39 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='46 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (7-11) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='32 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 (7-12) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='39 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (7-13) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='32 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='28 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 (7-14) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='43 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='45 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (7-15) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='28 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='37 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='37 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='39 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 NNNB (8-12) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='32 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 (8-14) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='32 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='28 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 (8-16) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='28 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='32 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 (a) Top view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (b) Side view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 6: Labeled spin sites in an AB-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Top view (a) and side view (b) of a 2 × 2 × 1 supercell of an AB-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chromium atoms in the top- and bottom layer are represented by green and blue spheres respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Iodine atoms and bonds are not shown in the picture for the sake of simplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The supercell is marked with solid black lines in panel (a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The crystal structures were drawn using VESTA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='19 Cr1 Gr5 Gr1 Cr12 Cr16 Cr4 Cr8 Cr7 Cr3 Cr10 Cr14 Cr2 Cr6 cH Cr5 ch1Cr1 Cr5 Cr1 Cr4 Cr8 Cr3 Cr7 Cr3 Cr2 Cr6 Cr1 Cr5 Cr1 C Cr12 Cr16 Cr11 Cr15 Cr10 Cr14 Cr9 Cr138 TABLE IV: Exchange parameters for an AB’-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' For each interacting pair of spins, the table contains the elements of both the NN and the NNN intra- and interlayer exchange matrices Jij, the out-of-plane exchange anisotropy ∆, and the components of the corresponding DMI-vectors Dij.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Since the sublattice symmetry is broken, parameters are given for the two sublattices (A,B) separately.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' All parameters are given in Cartesian coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding pairs are indicated in Figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' pair Jxx ij Jyy ij Jzz ij ∆ Jxy ij Jyx ij Jxz ij Jzx ij Jyz ij Jzy ij |Dx ij| |Dy ij| |Dz ij| |Dij| (i-j) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) (meV) Intra NN (1-2) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='63 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='66 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 (2-3) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='43 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='93 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='86 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (2-7) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='85 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='34 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='51 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='43 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='22 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 ⟨Jij⟩ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='97 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='94 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 NNNA (1-3) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='68 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='60 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (1-7) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='53 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 (3-7) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NNNB (2-4) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 (2-8) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (4-8) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='58 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='59 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='61 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='60 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 Inter NN (7-20) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (1-19) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 (2-19) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='27 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='28 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (2-20) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='26 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='29 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NNN (9-21) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 (9-26) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='18 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (10-22) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 (10-29) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='21 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='17 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='19 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 3NN (9-20) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (9-23) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 (9-25) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 (9-28) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 (10-23) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 (10-24) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 (10-26) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 (10-33) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 ⟨Jij⟩ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='07 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 9 (a) Top view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (b) Side view.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7: Labeled spin sites in an AB’-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Top view (a) and side view (b) of a 3 × 3 × 1 supercell of an AB’-stacked CrI3 bilayer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chromium atoms in the top- and bottom layer are represented by green and blue spheres respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Iodine atoms and bonds are not shown in the picture for the sake of simplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The supercell is marked with solid black lines in panel (a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The crystal structures were drawn using VESTA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='19 cr6 Cr18 cr3 Cr28 cr3 Cr9 cr2 Cr20 Cr8 Cr26 Cr32Cr6 Cr12 Cr18 Cr5 Cr11 Cr17 Cr4 Cr10 Cr16 Cr9 Cr2 Cr8 Cr1 Cr7 Cr13 Cr24 Cr30 cr36 Cr23 Cr29 Cr35 Cr22 Cr28 Cr34 Cr21 Cr33 Cr20 Cr26 Cr32 Cr25 Cr3110 TABLE V: Exchange parameters for CrI3 in the {λ, µ, ν} basis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Intra- and interlayer NN and NNN exchange, DMI and Kitaev parameters for monolayer (1L) and bilayer (2L) CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' The Kitaev interaction for the AB-stacking and the AB’-stacking is anisotropic so multiple values are given for each interaction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' pair(s) Jij Kij |Dij| (i-j) (meV) (meV) (meV) 1L Intra NN All 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='35 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 NNN All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='74 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='17 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 2L-AA Intra NN All 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='42 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='44 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 NNN All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 Inter NN All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 NNN All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 2L-AB Intra NN (1-2) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='45 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='13 NN (2-3) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='20 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='39 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NN (2-5) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 NNNA All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='59 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NNNB All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='14 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 Inter NNB All 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 NNN a 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 NNN b 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 2L-AB’ Intra NN (1-2) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 NN (2-3) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='20 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='39 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NN (2-7) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='64 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 NNN (1-3), (4-8) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='68 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='16 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 NNN (1-7), (2-4) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 NNN (3-7), (2-8) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='62 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 Inter NN (7-20), (2-19) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 NN (1-19), (2-20) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='31 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 NNN (9-26), (10-29) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='16 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 NNN (9-21), (10-22) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='11 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 3NN (9-20), (10-23) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 3NN (9-23), (10-24) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='04 3NN (9-25), (10-26) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 3NN (9-28), (10-33) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='01 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='00 a Pairs: (7-11), (7-13), (7-15), (8-12), (8-14), (8-16) b Pairs: (7-10), (7-12), (7-14) 11 IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ARTIFICIAL TUNING OF THE MAGNONIC DISPERSION As discussed in the main text, we can ’tune’ the spin-wave dispersion by artificially changing the magnetic pa- rameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' As shown in Figure 8(a,b) the size of the magnonic bandgap is tunable by changing the strength of the NNN DMI and the Kitaev constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' However, changing the Kitaev constant influences the shape of the dispersion in the whole Brillouin zone whereas the NNN DMI mainly effects the dispersion around the K-point and the K’-point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similar bandgap tuning can be observed by artificially changing parameters in bilayer CrI3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' In Figure 8(c), we show for an FM ordered AB-stacked bilayer that by decreasing the out-of-plane exchange difference between sublattices ∆Jzz = |Jzz A − Jzz B |, which is caused by the breaking of sublattice symmetry, we can also decreases the size of the magnonic bandgap and induce a topological phase transition from a topologically trivial to a topologically non-trivial phase.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Similar changes in the bandgap can be observed by changing decreasing ∆Jzz in the AB’-stacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a) (b) (c) FIG.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 8: Artificial tuning of the magnonic dispersion by manually changing the magnetic parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (a) Size of the magnonic bandgap in monolayer CrI3 plotted as a function of the NNN DMI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (b) Magnonic dispersion of monolayer CrI3 for different values of the Kitaev constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' (c) Magnonic bandgap in an AB-stacked CrI3 bilayer with a FM interlayer ordering as a function of the out-of-plane exchange difference between the two sublattices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Corresponding (composite) Chern numbers are indicated for the topological trivial phase (white) and the topological non-trivial phase (red).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='19 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='98 (meV) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='77 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='46 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='09 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='18 D2 (meV)25 K = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 meV K=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='98meV 1 20 K=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='76meV 一 ≥15 (me) E10 一 一 一 5 1 一 1 1 1 1 0 I K M0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='56 C102 = 0 C3 = 0 C4 = 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='54 K 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='52 C1θ2 = +1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='5 C = -1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='49 0= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='74 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='77 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='83 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='86 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='89 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='92 AJ22 (meV)12 ∗ Electronic address: milorad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='milosevic@uantwerpen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='be 1 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hafner, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 47, 558 (1993).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 2 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Furthm¨uller, Computational materials science 6, 15 (1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 3 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Furthm¨uller, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 54, 11169 (1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 4 P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bl¨ochl, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 50, 17953 (1994).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 5 J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Perdew, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Burke, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ernzerhof, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 77, 3865 (1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 6 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Grimme, Journal of Computational Chemistry 27, 1787 (2006).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 7 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Steiner, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Khmelevskyi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Marsmann, and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kresse, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 93, 224425 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 8 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dudarev, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Botton, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sevrasov, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Humphreys, and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sutton, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 57, 1505 (1998).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 9 P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhong, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yuan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lu and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ji, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 99, 144401 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 10 N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sivadas, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Okamoto, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fennie and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xiao, Nano Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18, 7658 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 11 G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' M¨uller, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hoffmann, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Dißelkamp, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sch¨urhoff, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Mavros, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sallermann, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Kiselev, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J´onsson, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bl¨ugel, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 99, 224414 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 12 D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' ˇSabani, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Bacaksiz, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Miloˇsevi´c, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' B 102, 014457 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 13 B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Huang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Clark, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Navarro-Moratalla, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Klein, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cheng, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Seyler, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhong, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Schmidgall, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' McGuire, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cobden, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature 546, 270 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 14 L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Thiel, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Tschudin, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rohner, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Guti´errez-Lezama, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ubrig, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gibertini, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Giannini, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Morpurgo and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Maletinsky, Science 364, 973 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 15 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Song, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fei, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yankowitz, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Lin, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Hwangbo, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Zhang, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Taniguchi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Watanabe et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18, 1298 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 16 T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Li, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jiang, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sivadas, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Weber, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Goldberger, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Watanabe, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Taniguchi, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Fennie et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=', Nature Mater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18, 1303 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 17 W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Sun, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Xu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Wu, and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Gao, Science 366, 983 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 18 S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Jeong, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Yoon, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Ryee and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Han, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Materials 3, 031001 (2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 19 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Momma, and F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Izumi, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' Cryst.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
+page_content=' 44, 1272 (2011).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/YdE0T4oBgHgl3EQfmwE9/content/2301.02502v1.pdf'}
diff --git a/Z9FQT4oBgHgl3EQffTZu/content/2301.13339v1.pdf b/Z9FQT4oBgHgl3EQffTZu/content/2301.13339v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..04bb72db5d180c03bdb1f4cec9dfb891f652d5fd
--- /dev/null
+++ b/Z9FQT4oBgHgl3EQffTZu/content/2301.13339v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:beda59dc5d162240c42166fbf03fd4ee6dd79f13c856c43376dabd1783f93646
+size 1314763
diff --git a/Z9FQT4oBgHgl3EQffTZu/vector_store/index.pkl b/Z9FQT4oBgHgl3EQffTZu/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..ce4885a51e16b67ed2a1a60017f8a6edffbce461
--- /dev/null
+++ b/Z9FQT4oBgHgl3EQffTZu/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8383a583e1d2d9d3e4b1f9c0d07b45c6e0d9d5f61ad1d4f6771991427007df20
+size 152261
diff --git a/ZdE2T4oBgHgl3EQfEgbW/content/2301.03637v1.pdf b/ZdE2T4oBgHgl3EQfEgbW/content/2301.03637v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..4347f7e35b4fe66d476a69cf0ea5ab45051cc095
--- /dev/null
+++ b/ZdE2T4oBgHgl3EQfEgbW/content/2301.03637v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4486e1d5872d64d7483a6120d0d8d2777824dda659e1481d5124644d3a4c4ce
+size 818436
diff --git a/ZdE2T4oBgHgl3EQfEgbW/vector_store/index.faiss b/ZdE2T4oBgHgl3EQfEgbW/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..2eeecfca964681e4d6d2dd5e2ebf00d3591398bf
--- /dev/null
+++ b/ZdE2T4oBgHgl3EQfEgbW/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43ea41a2a8d12441d87a15c51269acee9078320e04bb7c4560874b60b52df477
+size 3866669
diff --git a/ZdE2T4oBgHgl3EQfEgbW/vector_store/index.pkl b/ZdE2T4oBgHgl3EQfEgbW/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..08f11685c2e739e8b3f2587cb594cc141eb88387
--- /dev/null
+++ b/ZdE2T4oBgHgl3EQfEgbW/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fefccb8929d399cd9ece3cc42f50515a04ea606d688d4855c6bffa46cc307c7
+size 147522
diff --git a/b9E1T4oBgHgl3EQfKwO_/content/2301.02969v1.pdf b/b9E1T4oBgHgl3EQfKwO_/content/2301.02969v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..2dea3503b70b0dd050fdfd7235bf1a39f1635afd
--- /dev/null
+++ b/b9E1T4oBgHgl3EQfKwO_/content/2301.02969v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9ec42c9acab819d306400e6448a223bec7a89b1f128ffdab34204e83afc2d2b
+size 1133004
diff --git a/b9E1T4oBgHgl3EQfKwO_/vector_store/index.faiss b/b9E1T4oBgHgl3EQfKwO_/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..bdeca39c5b943f1d3fe197addf1ab65427912327
--- /dev/null
+++ b/b9E1T4oBgHgl3EQfKwO_/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5e7782047372f755bbfc9980b7d8d824c71632804d7c357836f77bb17cbb115
+size 4456493
diff --git a/c9FKT4oBgHgl3EQfqS4B/content/2301.11873v1.pdf b/c9FKT4oBgHgl3EQfqS4B/content/2301.11873v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..12ac56ddde8ddfbe3704470df3086b7fa5aadb72
--- /dev/null
+++ b/c9FKT4oBgHgl3EQfqS4B/content/2301.11873v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5e9ae1eab11408eb9850cab3c39e5414a453d2474f05c6536a634f9df53397d
+size 2243507
diff --git a/c9FKT4oBgHgl3EQfqS4B/vector_store/index.pkl b/c9FKT4oBgHgl3EQfqS4B/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..f2964ee2e47e2660960c5feb7ddff14da0786535
--- /dev/null
+++ b/c9FKT4oBgHgl3EQfqS4B/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:309366f88c017a3aaf6e6becc0196e43af0fb7538fe1e2771c8d4d0565699088
+size 305130
diff --git a/d9AzT4oBgHgl3EQfn_2x/vector_store/index.faiss b/d9AzT4oBgHgl3EQfn_2x/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..d109661a9d011602e9918648ab82fde0bd01fbb2
--- /dev/null
+++ b/d9AzT4oBgHgl3EQfn_2x/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e634e06fa9945e398006c1b7bb639058a0e13693954c98c44df6732a8723d43b
+size 3407917
diff --git a/edE2T4oBgHgl3EQfxQji/content/tmp_files/2301.04110v1.pdf.txt b/edE2T4oBgHgl3EQfxQji/content/tmp_files/2301.04110v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7cd59b6ef087a7b325505d946863fdace8254e72
--- /dev/null
+++ b/edE2T4oBgHgl3EQfxQji/content/tmp_files/2301.04110v1.pdf.txt
@@ -0,0 +1,1739 @@
+Structured Case-based Reasoning for
+Inference-time Adaptation of Text-to-SQL parsers
+Abhijeet Awasthi, Soumen Chakrabarti, Sunita Sarawagi
+Department of Computer Science and Engineering
+Indian Institute of Technology Bombay, Mumbai, India
+{awasthi,soumen,sunita}@cse.iitb.ac.in
+Abstract
+Inference-time adaptation methods for semantic parsing are
+useful for leveraging examples from newly-observed domains
+without repeated fine-tuning. Existing approaches typically
+bias the decoder by simply concatenating input-output exam-
+ple pairs (cases) from the new domain at the encoder’s input in
+a Seq-to-Seq model. Such methods cannot adequately leverage
+the structure of logical forms in the case examples. We pro-
+pose StructCBR, a structured case-based reasoning approach,
+which leverages subtree-level similarity between logical forms
+of cases and candidate outputs, resulting in better decoder deci-
+sions. For the task of adapting Text-to-SQL models to unseen
+schemas, we show that exploiting case examples in a struc-
+tured manner via StructCBR offers consistent performance
+improvements over prior inference-time adaptation methods
+across five different databases. To the best of our knowledge,
+we are the first to attempt inference-time adaptation of Text-
+to-SQL models, and harness trainable structured similarity
+between subqueries.
+1
+Introduction
+Natural language interfaces to databases (Zelle and Mooney
+1996; Tang and Mooney 2000; Popescu, Etzioni, and Kautz
+2003) enable access to structured information for users who
+are not familiar with languages like SQL by parsing user
+provided text-queries into executable SQLs. Text-to-SQL se-
+mantic parsing is a challenging task that not only demands
+robust natural language understanding but simultaneously
+requires reasoning over the schema structure of the databases.
+Databases containing similar information (e.g. census in vari-
+ous countries) may be designed using diverse schema struc-
+tures, thus making it hard for the model to generalize across
+schemas unseen during training. Hence, Text-to-SQL models
+often struggle to parse text queries for a new schema in a zero-
+shot manner (Suhr et al. 2020; Lee, Polozov, and Richardson
+2021; Hazoom, Malik, and Bogin 2021). In practice, a small
+number of Text-to-SQL examples in the target schema are
+often essential for successful model adaptation. However,
+finetuning a Text-to-SQL model for each new database is
+not generally practical, for the following reasons: (i) Huge
+variation in database schema makes it tedious to collect suf-
+ficiently large finetuning datasets for each schema, while
+Copyright © 2023, Association for the Advancement of Artificial
+Intelligence (www.aaai.org). All rights reserved.
+finetuning on small datasets is unavoidably fraught with over–
+fitting, catastrophic forgetting, and instability w.r.t. random
+seeds. (ii) Finetuning may take considerable time, preventing
+fast incorporation of new data into the model. (iii) Often,
+a single large-footprint model serves multiple clients with
+diverse databases at the same time. Fine-tuning a separate
+model for each database is considered too resource-intensive
+in such multi-tenant scenarios.
+Therefore, we focus on fast online adaptation of Text-
+to-SQL models without parameter updates, until the next
+cycle of finetuning is deemed feasible. Recently, case-based
+reasoning (CBR), which utilizes a memory of past labeled
+examples as cases, has emerged as a promising paradigm of
+inference-time adaptation without finetuning (Das et al. 2020,
+2021; Pasupat, Zhang, and Guu 2021; Gupta et al. 2021).
+CBR has been found effective for tasks like knowledge graph
+completion (KGC) (Das et al. 2020), question answering over
+knowledge bases (KBQA) (Das et al. 2021), task-oriented
+semantic parsing (Pasupat, Zhang, and Guu 2021; Gupta
+et al. 2021), translation (Khandelwal et al. 2021), and
+text-based games (Atzeni et al. 2022). However, many prior
+CBR approaches designed around Seq2Seq architectures
+simply concatenate input-output cases with the current input
+at the encoder (Das et al. 2021; Pasupat, Zhang, and Guu
+2021; Gupta et al. 2021). These methods do not leverage the
+structure of logical forms (query plan trees) in case examples.
+In response, we propose StructCBR, a structured CBR
+approach that directly exploits sub-tree level similarities
+between the candidate outputs and the case examples
+for adapting a Text-to-SQL model to a new schema. We
+start with SmBoP (Rubin and Berant 2021), a recent
+semi-auto-regressive architecture that decodes query trees
+bottom-up, respecting the structure of SQL grammar
+production rules, instead of left-to-right token-level decoding
+in Seq2Seq models (Guo et al. 2019; Wang et al. 2020;
+Scholak et al. 2021; Scholak, Schucher, and Bahdanau
+2021). We implement a novel structured case memory
+lookup module to boost scores of promising candidate
+trees using sub-tree level similarity with case trees under
+similar input context. This similarity is trainable. We show
+that explicitly-learned structured memory lookup leads
+to more accurate transfer from cases, compared to prior
+inference-time adaptation methods such as ConcatCBR and
+arXiv:2301.04110v1 [cs.CL] 10 Jan 2023
+
+GTM (Khandelwal et al. 2020; Das et al. 2021; Khandelwal
+et al. 2021) that we implemented both on SmBoP, and other
+Seq2Seq Text-to-SQL architectures like T5-large.
+We summarize our contributions as follows:
+1) We propose StructCBR, which, to our knowledge, is the
+first inference-time adaptation method for Text-to-SQL pars-
+ing without parameter fine-tuning.
+2) StructCBR incorporates a novel structured case memory
+and trainable query subtree similarity module that can boost
+scores of likely-correct outputs during inference. This is in
+contrast with earlier approaches like ConcatCBR and GTM.
+3) We propose a trainable compositional sub-tree similarity
+function that is both more accurate and more efficient for
+scoring large search frontiers, compared to default whole-
+tree embeddings.
+4) Through experiments with five database schemas (§ 5) of
+varying complexity, we observe that StructCBR is consis-
+tently better than prior inference-time adaptation methods on
+both SmBoP and sequence-based Text-to-SQL models.
+5) We show that StructCBR provides almost instant adapta-
+tion to a target schema. In contrast, finetuning (§ 5) can be up
+to 500 times slower.
+2
+SmBoP preliminaries
+We present a brief background on SmBoP here. Readers fa-
+miliar with SmBoP can largely skip this section. SmBoP
+converts a natural language question ¯x ∈ X (called the ‘utter-
+ance’) targeting a database schema ¯s ∈ S, to an SQL query
+ˆq ∈ Q. We describe the major modules in SmBoP.
+Utterance and schema encoding:
+Given token sequence
+¯x = [x1, x2, . . . , xn] in the text query, and database schema
+¯s = [s1, s2, . . . , sm] denoting table and column names,
+SmBoP jointly encodes them using a pre-trained Transformer
+like RoBERTa (Liu et al. 2020) followed by relation-aware-
+transformer (RAT) layers (Shaw, Uszkoreit, and Vaswani
+2018; Wang et al. 2020; Scholak et al. 2021). We denote the
+output from the last encoder layer as ¯x = [x1, x2, . . . xn]
+and ¯s = [s1, s2, . . . sm], representing the jointly encoded
+contextual embeddings of text tokens and schema elements
+respectively.
+Decoding SQL output:
+Unlike standard sequence-based
+decoding (Wang et al. 2020; Scholak et al. 2021; Scholak,
+Schucher, and Bahdanau 2021), SmBoP decodes the SQL
+tree bottom-up and in layers. SmBoP views any SQL query
+as a height-balanced relational algebra tree converted using
+a special idempotent KEEP operator κ as shown in Figure 1.
+Given a beam size K, at decoding step t, the decoder beam
+Bt comprises K candidate sub-trees of height t from the
+bottom. At step t + 1, trees from Bt are grown either via
+unary operators (e.g. COUNT), or by combining two trees in
+Bt using a binary operator (e.g. >), as per the SQL grammar.
+The candidate trees at step t + 1 form a frontier set Ft+1
+and is of size |Ft+1| = K2|B| + K|U|, where B and U
+represent the set of binary and unary operations respectively.
+SmBoP assigns each candidate tree z ∈ Ft+1 a score sθ(z)
+age 60
+>
+actors
+ 𝜿
+ 𝝈
+name
+ 𝜿
+ 𝜿
+ 𝚷
+t=1
+t=2
+t=3
+t=4
+age 60
+>
+actors
+ 𝜅
+ 𝜎
+name
+ 𝜅
+ 𝜅
+ Π
+t=1
+t=2
+t=3
+t=4
+Figure 1: SmBoP (Rubin and Berant 2021) decodes SQL
+as a balanced relational algebra tree. At each level t, trees
+in the beam combine via unary or binary operators to form
+candidates of the next beam. StructCBR leverages CBR on
+generated sub-trees.
+Text 1
+Give the code of the airport with the fewest number
+of flights
+SmBoP
+output
+SELECT sourceairport FROM flights
+GROUP BY sourceairport ORDER BY
+SUM(flightno) ASC LIMIT 1
+Correct
+SQL
+SELECT airportcode FROM airports
+JOIN flights ON airportcode
+= sourceairport GROUP BY
+sourceairport ORDER BY COUNT(*)
+ASC LIMIT 1
+Text 2
+What is the code of the airport that has the highest
+number of flights?
+Table 1: Illustration of the lack of generalization of Text-to-
+SQL to new schema.
+(described below). The top-K highest scoring trees in Ft+1
+form the next beam Bt+1. This continues up to a maximum
+height T, when the highest scoring tree in BT is output.
+Scoring a tree:
+A tree z = (zb, zℓ, zr) consists of root op-
+erator zb and subtrees zℓ, zr. SmBoP encodes a variable-size
+tree z into two fixed dimensional vectors: (i) z: an embedding
+of the tree computed recursively on the tree structure, where
+a transformer outputs z = TXθ([zb, zℓ, zr]); (ii) z′: a contex-
+tual representation of z grounded in input text ¯x computed
+via a multiheaded cross attention module z′ = XAttθ(z, ¯x).
+SmBoP computes the score of a tree z ∈ Ft+1, as follows:
+sθ(z) = wT
+zb FFθ([zℓ; z′
+ℓ; zr; z′
+r])
+(1)
+where FFθ is a feed forward network, and wzb represents a
+learned embedding of operator zb.
+The model is trained using Text-SQL pairs from a set of
+training schema to maximize the likelihood of the correct sub-
+trees at each beam. During inference, when presented with
+text utterances relating to a new database schema, the model
+often fails to discover the mapping of the text to schema
+names and relationships in the new schema. Table 1 presents
+an example where a SmBoP model trained on the Spider
+dataset (Yu et al. 2018) is deployed on a new schema about
+flights. On inspecting the predicted and correct SQL, we
+find that the model failed to reason that number of flights
+requires a count(*) instead of sum(flightno). Now
+suppose an expert provides the correct SQL as additional in-
+formation to be used during inference of subsequent queries.
+Consider a second query (shown as Text 2 in Table 1) that
+
+name
+actors
+age
+60K
+K
+>KIalso needs to reason about number of flights, and the de-
+fault SmBoP makes similar errors (not shown). Only existing
+mechanism in SmBoP is to fine-tune parameters which could
+be time-consuming and unstable. In the next section we show
+how our method can instantaneously leverage test-time user
+labels to predict the correct SQL for Text 2. More such anec-
+dotes appear in Table A4 of the Appendix.
+3
+Our proposed method: StructCBR
+We aim to learn a Text-to-SQL model M, using a dataset
+Dtrain of Text-SQL pairs such that it is capable of C1: Inde-
+pendently translating the text queries ¯x to executable SQL
+programs ˆq, and C2: Utilizing a small set Dnew of Text-SQL
+pairs from a target schema snew, to improve its own predic-
+tions during inference, without finetuning. In line with prior
+work (Das et al. 2020, 2021), we refer to the second capabil-
+ity C2 as Case-based reasoning (CBR), and the dataset Dnew
+of Text-SQL pairs in the target schema as cases.
+The StructCBR module leverages the similarity between
+gold subtrees that appear in similar contexts in the set of cases
+Dnew and the candidate subtrees in SmBoP’s frontier Ft+1, to
+boost the scores of likely-correct candidates at each decoding
+step t + 1. Consider a subtree z in the frontier Ft+1 for an
+input text ¯x, a case-example with text question as ¯xc, and
+the gold SQL tree as Zc
+gold. Let zc be a subtree of Zc
+gold. The
+key idea of StructCBR is, if z and zc are structurally similar,
+and appear in similar contexts w.r.t. ¯x and ¯xc, then there
+is a strong evidence that the subtree z should also appear
+as a part of the gold tree Zgold of ¯x. Figure 2 provides an
+illustration with z = age > 60 in the candidate frontier Ft+1,
+and a similarly structured case tree zc = age > 80 appearing
+in a similar context ¯xc (both contain the phrase who are
+past).
+Even though the key idea of matching with case sub-trees
+is simple, several important design choices had to be made
+to ensure that CBR inter-operates efficiently with SmBoP’s
+own scoring, and consistently improves its performance in
+the presence of multiple cases of varying levels of related-
+ness. First, how should we compute the contextual similarity
+of a candidate tree z with a case tree, given that memory
+would also contain unrelated cases that would match wrong
+candidate trees? Second, how can we efficiently compute
+the similarity of all candidate trees with all entries in the
+case memory? Unlike Seq2Seq models that do not perform
+beam-search during training, SmBoP generates a large search
+frontier even during training. We elaborate on how our design
+tackles these challenges next.
+Algorithm 1 presents the high-level pseudo code, with the
+text in blue font representing the StructCBR additions to the
+SmBoP model.
+3.1
+Choosing tree representations
+We need to choose a representation of a tree z using which
+we can efficiently compute similarity with case trees. Just
+the structural similarity of z with a case zc is not sufficient
+unless we also contextualize them on their respective inputs.
+Accordingly, we design an embedding function Gφ(z, ¯x) �→
+Rd that jointly encodes a candidate tree z corresponding to
+Algorithm 1: SmBoP with StructCBR.
+1 input: ¯x, ¯s, Dnew
+2 M ← CreateCaseMemory(Dnew) (§ 3.2)
+3 ¯x,¯s ← EncodeTextSchemaθ(¯x, ¯s)
+4 B0 ← top-Kschema constants and DB values
+5 for t ← 0 . . . T − 1 do
+6
+z ← CreateTreeReps(z)
+7
+z′ ← GroundTreeReps(z, x)
+8
+pθ ← SmBoPScores(z, z′) (§ 2)
+9
+Gφ(z, ¯x) ← JointReps(z, z′, x) (Eqn 2)
+10
+simφ ← TreeSim(Gφ(z, ¯x), M) (Eqn 4)
+11
+pφ ← StructCBRScores(simφ, M) (Eqn 5)
+12
+Ft+1 ← CombineScores(pθ, pφ) (Eqn 6)
+13
+Bt+1 ← top-K(Ft+1)
+14 return argmaxz(BT )
+an input ¯x as a d dimensional vector. We train a separate
+transformer model TXφ with parameters φ that takes as input
+four vectors: z that encodes the structure of the tree z, z′
+that is the contextual representation of z defined in § 2, an
+embedding wb of z’s root node b, and pool(x) a mean-pooled
+version of the input text representation ¯x:
+Gφ(z, ¯x) = TXφ([z, z′, wb, pool(x)]).
+(2)
+This embedding captures both the structure and context and
+the parameters φ are trained to co-embed similar trees in
+matching contexts, while pulling apart pairs differing either
+structurally or contextually. For example, in Figure 2 if the
+query text was Name all actors who are 60 or
+above, then the similarity of candidate age > 60 from
+the same case sub-tree should be reduced. Unlike recursive
+tree representations (Socher et al. 2013), here contextualiza-
+tion w.r.t. ¯x plays a critical role.
+3.2
+Case memory design
+We construct a case memory M over the gold SQL trees
+{Zc
+gold} for all cases in Dnew. Corresponding to each node
+b of a gold tree Zc
+gold we form a subtree rooted at b and
+including the part of Zc
+gold below b. Thus, the size of the case
+memory is the total number of nodes over all gold trees in
+cases. The encoding Gφ(zc, ¯xc) of each subtree zc for a case
+(¯xc, Zc
+gold) in Dnew is pre-computed using Equation 2 and
+stored in M.
+3.3
+Efficient tree similarity computation
+We need to compute the similarity of each tree z in the frontier
+Ft+1 with all case sub-trees zc ∈ M. One way to compute
+similarity between trees z and zc is based on ℓ2 distance1
+between their Gφ representations as follows:
+simφ(z, zc, ¯x, ¯xc) = − ∥Gφ(z, ¯x) − Gφ(zc, ¯xc)∥2
+(3)
+However, computing Gφ representations for each tree z ∈
+Ft+1 entails large memory and compute costs since the fron-
+tier size |Ft+1| = K2|B| + K|U| is quadratic in beam-size
+1Like Khandelwal et al. (2020) we observed better results with
+ℓ2 distance, in comparison to inner product.
+
+Text 𝑥̅ : Name all the actors who are past 60 | Schema (𝑠̅): || T.actor | id | name | age | dob || T.director | id | age | movies …
+Case Text ( 𝑥̅� ): Show the
+directors names who are past 80
+Case SQL ( 𝑞�� ): SELECT name
+FROM director WHERE age>80
+director
+ 𝜿
+ 𝝈
+age 80
+>
+director
+ 𝜿
+ 𝝈
+name
+ 𝜿
+ 𝜿
+ 𝚷
+age
+80
+>
+director
+ 𝜿
+Encode Text and Schema (RoBERTa)
+name
+actor
+age
+60
+𝜅
+𝜅
+≥
+>
+≤
+=
+𝜎
+name
+actor
+age
+60
+𝜅
+𝜅
+≥
+>
+≤
+=
+𝜎
+5
+7
+4
+1
+2
+1
+−1
+−3
+−5
+−∞
+−1
+−∞
+−∞
+−∞
+name
+ 𝜿
+Π
+−3
+Π
+−∞
+top-K
+name
+ 𝜿
+actor
+ 𝜿
+age 60
+≥
+age 60
+≤
+.22
+.19
+.15
+. 𝟐𝟓
+.02
+.02
+.01
+.01
+Combine Scores
+top-K
+name
+ 𝜿
+actor
+ 𝜿
+age 60
+>
+age 60
+≥
+Create Case Memory
+𝐹���
+𝐵�
+𝐵���
+𝑠�
+𝑠�
+SmBoP Scoring
+StructCBR Scoring
+Case Memory
+80
+>
+age
+Decoding step
+A
+B
+C
+Figure 2: Augmenting SmBoP with StructCBR (Structured Case-based Reasoning): In part A
+⃝, the top-K step in SmBoP
+scoring misses the correct sub-tree age > 60 due to a lower score (score=1) w.r.t. competing sub-trees in the frontier Ft+1 like
+age ≥ 60 (score=4) and age ≤ 60 (score=2). In part C
+⃝, StructCBR creates a memory of all the sub-tree representations
+available in cases as described in § 3.2. In part B
+⃝, StructCBR scores the frontier candidates based on learned tree-similarities
+w.r.t. the sub-trees in cases as described in § 3.3 and § 3.4. For example, StructCBR boosts the score of age > 60 because of
+its high similarity with the case sub-tree age > 80 and similarity of context who are past. Thus, the top-K step applied
+on the combined SmBoP and StructCBR scores recovers the correct sub-trees that otherwise may get missed based on SmBoP’s
+scoring alone. For brevity, we consider only one case-example in this figure.
+K. With the default K for SmBoP being 30, and size of
+the SmBoP grammar, this translates to around 23 thousand
+trees per frontier. Pruning the frontier Ft+1 based on SmBoP
+scores alone resulted in poor performance. This led us to
+design an alternative compositional CBR scoring method that
+can more efficiently score all candidate trees in the frontier.
+Our key idea is to compute the similarity between two trees
+compositionally as a function of similarity between their left
+and right children respectively. This requires only O(K) tree
+representations for the trees in beam Bt as against K2|B| +
+K|U| operations of the whole-tree approach. Recall that the
+trees z ∈ Ft+1 are formed by combining trees in beam Bt
+via SQL operations. A tree z ∈ Ft+1 can thus be represented
+as z = (zb, zℓ, zr) where zℓ, zr ∈ Bt denote the left and
+right child respectively, and zb is the root node combining
+zℓ and zr. After the beam Bt is created, we compute the
+embedding Gφ for each tree in Bt using Equation (2). Now,
+the similarity between a candidate tree z = (zb, zℓ, zr) for
+an input text ¯x, and a case sub-tree zc = (zc
+b, zc
+ℓ, zc
+r) on input
+text ¯xc in memory M is computed as:
+�
+simφ(z, zc, ¯x, ¯xc) = simφ(zℓ, zc
+ℓ, ¯x, ¯xc)
++ simφ(zr, zc
+r, ¯x, ¯xc).
+(4)
+In Section 5, we show that using this similarity function
+provides better results by allowing the entire frontier to be
+scored more efficiently in comparison to computing similar-
+ities based on Equation 3 only for a subset of trees in the
+frontier pruned based on SmBoP scores.
+3.4
+Boosting SmBoP frontier with tree
+similarities
+To compute an overall score of a candidate tree z ∈ Ft+1
+based on its similarity with the case sub-trees in M, we
+aggregate over all the case sub-trees zc with the same root
+node (zb = zc
+b) using a logsumexp operator, which provides
+us a soft-maxed similarity of z w.r.t. case sub-trees.
+sφ(z) = log
+�
+c∈M∧zb=zc
+b
+exp(�
+simφ(z, zc, ¯x, ¯xc))
+(5)
+Now every candidate tree z ∈ Ft+1 has two scores: sθ(z) as-
+signed by default SmBoP and sφ(z) computed by StructCBR.
+The scores sθ(z) and sφ(z) can lie in very different ranges.
+Summing them in a normalized probability space provided
+better results than summing the scores directly. Hence, we
+independently normalize sθ(z) to pθ(z) and sφ(z) to pφ(z)
+by a softmax operation applied over all trees in the frontier.
+The combined score of a frontier tree z is:
+p(z) = (pθ(z) + pφ(z))/2.
+(6)
+3.5
+Supervising StructCBR
+During training, we assume availability of training data
+Dtrain = {(¯xi, ¯si, ¯qi)}N
+i=1 consisting of utterances ¯xi on a
+schema ¯si, and the corresponding gold SQL queries ¯qi. We
+first train the SmBoP model, parameterized as θ, using Dtrain.
+The training objective of SmBoP for a single example maxi-
+mizes the likelihood of sub-trees that are part of the tree Zgold
+
+corresponding to gold SQL ¯q:
+Lθ = −
+T
+�
+t=0
+�
+zt∈Zgold
+log pθ(zt).
+(7)
+Next, we introduce the StructCBR module parameterized
+as φ on top of the (now frozen) SmBoP model. We ob-
+served training the StructCBR parameters φ while freezing
+the learned SmBoP parameters θ to provide slightly better
+results in comparison to training both θ and φ jointly. The
+parameters φ are also learned using Dtrain by maximizing the
+likelihood of the gold subtrees as per the distributions pφ and
+p through the following loss function:
+Lφ = −
+T
+�
+t=0
+�
+zt∈Zgold
+log pφ(zt) + log p(zt)
+(8)
+The − log pφ(zt) term maximizes the likelihood of gold trees
+w.r.t. the CBR distribution pφ, independent of the SmBoP
+distribution pθ. Similarly, the − log p(zt) term maximize the
+likelihood of the gold trees w.r.t. the combined distribution
+p (Eqn 6). During training, we design each training batch to
+contain C examples from same schema so that for a given
+train example, the remaining C − 1 examples serve as the
+cases from the same schema. We train with C = 32 and a
+batch-size of 64.
+4
+Related work
+We review prior work on inference-time model adaptation
+for related tasks and also describe our adaptation of some of
+these works in the context of Text-to-SQL for comparisons
+with StructCBR.
+Concatenating related examples with input:
+A common
+approach, that we call ConcatCBR, for utilizing cases dur-
+ing inference is to concatenate the input-output pair of each
+case along with the input text at the encoder of a Seq2Seq
+model. During training, the decoder is expected to learn to
+utilize the cases on the encoder side. Das et al. (2021) utilize
+ConcatCBR for question answering over knowledge bases,
+and Pasupat, Zhang, and Guu (2021); Gupta et al. (2021)
+utilize ConcatCBR for other semantic parsing tasks. Con-
+catCBR is similar to the retrieve and edit framework for
+structured outputs (Hashimoto et al. 2018) and machine trans-
+lation (Hossain, Ghazvininejad, and Zettlemoyer 2020). For
+the Text-to-SQL task, we implement a ConcatCBR baseline
+that trains an SmBoP model to use retrieved Text-SQL exam-
+ples concatenated with the input-text. During inference, the
+retrieval index is updated with the case-examples from the
+target schema.
+Generalization through Memorization (GTM):
+Khan-
+delwal et al. (2020, 2021) propose a memory look-up based
+method for adapting pre-trained language and machine trans-
+lation models to a target domain. Given a target dataset, their
+method constructs a look-up index by using contextual em-
+beddings from the pre-trained model as keys and the corre-
+sponding text tokens as values. During inference the model
+scores are interpolated with the similarity scores aggregated
+over the nearest neighbours in the loop-up index. For our Text-
+to-SQL set-up, we implement this baseline using a trained
+SmBoP model. We memorize the dataset Dnew in the target
+schema by creating a look-up index with embeddings of child
+subtrees from SmBoP as keys: [zℓ; z′
+ℓ; zr; z′
+r], and their par-
+ent nodes as values. During inference, the scores from the
+SmBoP model are interpolated with neighbour similarities in
+a way similar to Khandelwal et al. (2021). Unlike StructCBR
+and ConcatCBR, this baseline (GTM) does not explicitly
+train the SmBoP model for utilizing the cases during infer-
+ence.
+We discuss other related work in Appendix A.6.
+5
+Experiments
+We evaluate StructCBR for adapting a Text-to-SQL model to
+five different target schemas without finetuning. The target
+schemas are chosen from varying domains. We compare
+StructCBR with prior inference-time adaptation methods
+discussed in § 4, and present an ablation study. We also show
+that StructCBR enables much faster adaptation of Text-to-
+SQL models in comparison to finetuning.
+Datasets:
+We utilize Spider (Yu et al. 2018), which is a
+collection of Text-to-SQL examples covering 200 unique
+schemas. We use the train split of Spider as Dtrain, for train-
+ing all the models. Dtrain contains 7000 Text-SQL example
+pairs from 140 databases. For evaluation, we hold out the
+following five databases containing the most examples from
+Spider’s dev set 2: {world 1, car 1, cre Doc Template Mgt,
+dog kennels, flight 2}. The five evaluation databases do not
+occur in the train set, and belong to sufficiently different do-
+mains of varying difficulty. The remaining part of the dev set
+containing 576 examples is used for model selection while
+training on Dtrain. We hold out 30 randomly selected exam-
+ples from each of the five selected databases as Dnew (cases)
+for adaptation, and use the remaining examples as the test set,
+Dtest. The average size of Dtest is 60, and varies from roughly
+50 to 90 examples across the five schemas. To ensure robust
+evaluation, we report numbers averaged over three random
+Dnew/Dtest splits. We also report the numbers micro-averaged
+over all the 300 test examples across the five schemas.
+Evaluation metrics:
+Following prior work (Yu et al. 2018),
+we report Execution Accuracy (EX) and Exact-Set-Match
+Accuracy (EM) for all the methods. EX returns 1 if executing
+the gold query ¯q and the predicted query ˆq on the target
+database gives the same results. EM compares all the SQL
+clauses within ¯q and ˆq and returns 1 if all the clauses match,
+except possibly the DB-values (constants) in the SQL query.
+Most Text-to-SQL models utilize beam search, and return the
+top-K highest scoring candidates in the beam as the output.
+Hence, we also report the top-K versions of EM and EX
+metrics as BEM and BEX respectively, where K is the beam
+size. In our experiments, K = 30. BEM/BEX for a beam is
+1, if at least one of the candidates in the beam has an EM/EX
+of 1.
+2Spider’s test set is publicly inaccessible as of 08/15/2022.
+
+Methods compared:
+We compare the accuracy of Struct-
+CBR after adaptation with the following methods: (i) SmBoP:
+The base model without any adaptation to benchmark the
+gains from different inference-time adaptation methods.
+(ii) ConcatCBR: The standard method of concatenating input-
+output case examples with the input-text. (iii) GTM: Mapping
+Dnew using SmBoP into a non-parametric memory for aug-
+menting model’s predictions with inference-time memory
+look-ups similar to Khandelwal et al. (2020, 2021). We dis-
+cussed ConcatCBR and GTM baselines in Section 4. All the
+baselines are implemented using SmBoP as the base model.
+In Appendix A.2 we also present ConcatCBR implemented
+on a T5-based Seq2Seq model.
+Implementation details:
+We implemented StructCBR and
+baselines using AllenNLP (Gardner et al. 2017) and Trans-
+formers (Wolf et al. 2020) libraries. We utilize the authors’
+implementation of SmBoP (Rubin and Berant 2021). Due to
+limited computing resources, we primarily experiment with
+the ROBERTA-BASE checkpoint for initializing the text en-
+coder, followed by four RAT layers (Wang et al. 2020) to
+encode the schema structure. All other hyper-parameters are
+the set to their default values. The SmBoP model is trained
+on Dtrain for 60K steps with a batch size of 80, using the
+default learning rate (LR) of 1.86×10−4. The GTM baseline
+utilizes the output of this model for memory look-ups. For
+ConcatCBR baseline we train the SmBoP model further for
+60K steps with a LR of 5×10−5, while concatenating the
+retrieved cases in the encoder’s input. StructCBR introduces
+2.53% additional parameters (φ) over the SmBoP parame-
+ters (θ). We train the parameters φ on Dtrain using a batch
+size of 64 for 60K steps with the default LR of 1.86×10−4.
+Additional training details are provided in Appendix A.4.
+Overall Results:
+In Table 2, we compare StructCBR with
+different inference-time methods for adapting SmBoP based
+models on five different evaluation schemas. The perfor-
+mance of the unadapted SmBoP model varies from 46.1 EM
+to 84.3 EM across the five schemas indicating that the evalu-
+ation schemas are of varying difficulty. We find that Struct-
+CBR almost consistently offers substantial gains over the
+base SmBoP model w.r.t. all the metrics, and across all the
+five schemas. StructCBR gains upto 6.3 EM points and on
+average 4.6 EM points over the base SmBoP model, while
+achieving almost 4 times higher gains than best existing
+method. In contrast, the ConcatCBR method, which has been
+shown to work well for other semantic parsing tasks (Das
+et al. 2021; Pasupat, Zhang, and Guu 2021), provides posi-
+tive EM gains for only three out of five schemas, and causes
+overall drop in micro-averaged EM over all the test instances.
+We also explored a ConcatCBR implementation based on T5-
+large model with a comparable base accuracy (Appendix A.2).
+Even compared to this method, we continue to observe higher
+and more consistent gains from StructCBR on SmBoP. GTM,
+another inference-time adaptation baseline, utilizes memory
+lookups similar to Khandelwal et al. (2021), and offers al-
+most consistent but much smaller gains in comparison to
+StructCBR. The GTM baseline performs memory look-ups
+based on representations learned by the base SmBoP model
+whereas StructCBR is explicitly trained to perform sub-tree
+EM
+EX
+BEM
+BEX
+snew = world 1 , |Dtest| = 89
+SmBOP
+46.1
+36.3
+55.4
+49.0
+ConcatCBR
+-4.1
+-4.9
+-5.2
+-6.3
+GTM
++0.0
++0.0
++0.4
+-0.4
+StructCBR (Ours)
++2.6
++2.6
++3.8
++0.0
+snew = car 1 , |Dtest| = 60
+SmBOP
+43.3
+45.6
+50.6
+53.9
+ConcatCBR
++3.9
++3.9
++4.4
++3.4
+GTM
++2.2
++2.2
++3.9
++2.8
+StructCBR (Ours)
++6.1
++6.7
++8.3
++7.2
+snew = cre Doc Template Mgt , |Dtest| = 53
+SmBOP
+84.3
+89.3
+94.3
+96.2
+ConcatCBR
++5.7
++5.0
++1.9
++1.3
+GTM
++3.2
++1.9
++1.9
++1.3
+StructCBR (Ours)
++6.3
++3.8
++3.8
++2.5
+snew = dog kennels , |Dtest| = 49
+SmBOP
+66.6
+59.2
+80.3
+74.8
+ConcatCBR
++0.7
++0.0
+-2.0
+-1.3
+GTM
++1.4
++2.7
++0.7
++3.4
+StructCBR (Ours)
++3.4
++6.1
++3.4
++4.1
+snew = flight 2 , |Dtest| = 49
+SmBOP
+55.8
+67.4
+59.2
+80.3
+ConcatCBR
+-8.8
+-6.1
+-2.0
+-1.3
+GTM
++0.0
+-1.4
++2.1
++0.0
+StructCBR (Ours)
++5.5
++4.1
++12.3
++4.8
+Micro-Average , |Dtest| = 300
+SmBOP
+57.2
+56.3
+66.0
+67.6
+ConcatCBR
+-0.8
+-0.8
+-1.0
+-1.4
+GTM
++1.2
++1.0
++1.7
++1.2
+StructCBR (Ours)
++4.6
++4.4
++6.0
++3.4
+Table 2: Comparison of StructCBR with prior inference time
+adaptation methods (ConcatCBR and GTM) on 5 different
+schemas. SmBoP row provides the performance of the un-
+adapted model, while other rows report gains over SmBoP
+after adaptation. Micro-average refers to numbers averaged
+over all the test instances spanning across the five schemas.
+|Dtest| refers to size of the test-set, and snew provides the
+schema name.
+level memory look-ups. In particular, StructCBR boosts the
+top-K EM score (BEM) by up to 12.3 points. With a large-
+sized SmBoP architecture we continue to observe consistent
+gains for most of the schemas. Section A.5 and Table A3 in
+Appendix provide results for adapting the large-sized models
+in the same setting as Table 2. Table A4 in Appendix provides
+some anecdotal examples that were originally mispredicted
+by the base SmBoP model, but were fixed by StructCBR.
+Impact of number of cases:
+We run another set of ex-
+periments by reducing the number of cases from 30 to 20.
+Figure 3 shows that gains from both GTM and StructCBR are
+smaller with fewer cases, as expected. Interestingly, Struct-
+
+-1
+0
+1
+2
+3
+4
+5
+ConcatCBR
+GTM
+StructCBR
+|Cases|=20
+|Cases|=30
+Figure 3: Impact of case size on gains in EM (y-axis) of
+adapted models (x-axis). With 20 cases, StructCBR still out-
+performs GTM with 30 case examples.
+CBR with 20 cases still outperforms GTM with 30 case
+examples. For ConcatCBR we do not observe significant
+changes because it additionally augments the case memory
+with examples retrieved from Spider’s train set. Not augment-
+ing cases via retrieval resulted in even worse performance of
+ConcatCBR.
+Justification for tree similarity:
+In Section 3.3, we argued
+that directly computing tree similarities (simφ) using Equa-
+tion (3) was inefficient, and required pruning to be practical.
+Instead in StructCBR we compute tree similarity (�
+simφ as
+per Equation 4) more efficiently as a composition of simi-
+larity of its children, and does not require pruning. Table 3
+justifies our design by comparing results of scoring a pruned
+frontier containing top-5K trees using simφ, with scoring
+the entire frontier using �
+simφ. Efficiently scoring the entire
+frontier provides us better results on 4 out 5 schemas and a
+micro-averaged gain of 2.2 points in EM.
+world 1 car 1
+cre
+dog flights 2 Micro
+Doc kenn
+avg
+Whole-tree
+43.4 48.9 88.7
+76.2
+55.1
+59.6
+Ours
+48.7 49.4 90.6
+70.1
+61.2
+61.8
+Table 3: Scoring pruned frontier as per similarity of whole
+trees as in Equation (3) vs. our scoring of the entire frontier
+as per similarity composed from subtrees, as in Equation (4).
+Better performance on four out of five schemas justifies our
+design of tree similarity.
+Comparison with fine-tuning:
+Adapting transformer
+models via finetuning often provides significant accuracy
+improvements, thanks to recent advances in optimization
+of transformer models (Huang et al. 2020; Xu et al. 2021).
+However, finetuning is not viable when a model needs to
+be adapted instantaneously from just a few data points, e.g.
+Adaptation Method
+Time(s)↓
+EM%↑
+SmBOP (Unadapted)
+0.0
+48.3
+StructCBR
+0.1
+50.6
+Finetuning (1 epochs)
+5.0
+48.3
+Finetuning (2 epochs)
+10.0
+47.5
+Finetuning (5 epochs)
+25.0
+48.3
+Finetuning (10 epochs)
+50.0
+50.2
+Finetuning (20 epochs)
+100.0
+50.9
+Finetuning (100 epochs)
+500.0
+52.1
+Table 4: Comparing adaptation time and EM accuracy of
+StructCBR and finetuning for different number of epochs on
+30 cases of world 1 schema. We report wall clock times in
+seconds. All the numbers were averaged over 3 runs. Finetun-
+ing took atleast 500x more time (10 epochs) to achieve EM
+gains that are almost instantly (0.1s) achieved by StructCBR
+quickly incorporating expert feedback in the form of a few
+examples. In Table 4, we show that StructCBR serves the
+purpose of instantaneously improving model performance
+(+2.6 pts EM), while being roughly 50× faster than finetun-
+ing for a single epoch, and 5000× faster than finetuning for
+100 epochs of 30 case examples from world 1 schema. Fine-
+tuning required atleast 10 epochs to outperform StructCBR’s
+accuracy. Each epoch involved four parameter update steps of
+batch-size 8. We note that applying StructCBR over SmBoP
+incurs a small increase (∼1.2×) in inference time per exam-
+ple. Overall, we observe that on three out of five schemas
+StructCBR instantly offers more than 50% of gains achieved
+by finetuning for 100 epochs, and 43% of gains on average
+across all schemas (Table A2 in appendix). This establishes
+StructCBR as a method for instantly utilizing available case
+examples for fast model adaptation until the next cycle of
+finetuning becomes affordable.
+6
+Conclusion and Future Work
+We presented StructCBR, a method for instant adaptation
+of Text-to-SQL models without finetuning. We show that
+utilzing case examples in a more structured way via sub-tree
+level look-ups offers better performance in comparison to the
+standard method of concatenating case examples with input
+text into a Seq2Seq encoder. We find that explicitly learning
+to perform memory look-ups provides larger gains in com-
+parison to look-ups using a pre-trained model. Finally, we
+show that StructCBR enables much faster model adaptation
+in comparison to finetuning, potentially allowing instanta-
+neous adaptation to expert feedback provided in form of a
+few examples. We propose StructCBR as a faster alternative
+to model adaptation, until the next finetuning cycle is deemed
+feasible. Despite its speed, there remains an accuracy gap
+between StructCBR and sufficient finetuning, which might be
+narrowed by more sophisticated similarity networks. Our ex-
+ploration of StructCBR focused only on the Text-to-SQL task.
+In future we wish to explore the effectiveness of StructCBR
+for other semantic parsing tasks.
+
+References
+Atzeni, M.; Dhuliawala, S. Z.; Murugesan, K.; and SACHAN,
+M. 2022. Case-based Reasoning for Better Generalization
+in Text-Adventure Games. In International Conference on
+Learning Representations.
+Brown, T.; Mann, B.; Ryder, N.; Subbiah, M.; Kaplan, J. D.;
+Dhariwal, P.; Neelakantan, A.; Shyam, P.; Sastry, G.; Askell,
+A.; et al. 2020.
+Language models are few-shot learners.
+Advances in neural information processing systems, 33: 1877–
+1901.
+Chen, M.; Tworek, J.; Jun, H.; Yuan, Q.; de Oliveira Pinto,
+H. P.; Kaplan, J.; Edwards, H.; Burda, Y.; Joseph, N.; Brock-
+man, G.; Ray, A.; Puri, R.; Krueger, G.; Petrov, M.; Khlaaf,
+H.; Sastry, G.; Mishkin, P.; Chan, B.; Gray, S.; Ryder, N.;
+Pavlov, M.; Power, A.; Kaiser, L.; Bavarian, M.; Winter, C.;
+Tillet, P.; Such, F. P.; Cummings, D.; Plappert, M.; Chantzis,
+F.; Barnes, E.; Herbert-Voss, A.; Guss, W. H.; Nichol, A.;
+Paino, A.; Tezak, N.; Tang, J.; Babuschkin, I.; Balaji, S.;
+Jain, S.; Saunders, W.; Hesse, C.; Carr, A. N.; Leike, J.;
+Achiam, J.; Misra, V.; Morikawa, E.; Radford, A.; Knight,
+M.; Brundage, M.; Murati, M.; Mayer, K.; Welinder, P.; Mc-
+Grew, B.; Amodei, D.; McCandlish, S.; Sutskever, I.; and
+Zaremba, W. 2021.
+Evaluating Large Language Models
+Trained on Code.
+Das, R.; Godbole, A.; Monath, N.; Zaheer, M.; and McCal-
+lum, A. 2020. Probabilistic Case-based Reasoning for Open-
+World Knowledge Graph Completion. In Findings of the
+Association for Computational Linguistics: EMNLP 2020,
+4752–4765.
+Das, R.; Zaheer, M.; Thai, D.; Godbole, A.; Perez, E.; Lee,
+J.-Y.; Tan, L.; Polymenakos, L.; and McCallum, A. 2021.
+Case-based Reasoning for Natural Language Queries over
+Knowledge Bases. arXiv preprint arXiv:2104.08762.
+Gardner, M.; Grus, J.; Neumann, M.; Tafjord, O.; Dasigi,
+P.; Liu, N. F.; Peters, M.; Schmitz, M.; and Zettlemoyer,
+L. S. 2017. AllenNLP: A Deep Semantic Natural Language
+Processing Platform.
+Guo, J.; Zhan, Z.; Gao, Y.; Xiao, Y.; Lou, J.-G.; Liu, T.;
+and Zhang, D. 2019.
+Towards Complex Text-to-SQL in
+Cross-Domain Database with Intermediate Representation.
+In Proceedings of the 57th Annual Meeting of the Association
+for Computational Linguistics, 4524–4535. Florence, Italy:
+Association for Computational Linguistics.
+Gupta, V.; Shrivastava, A.; Sagar, A.; Aghajanyan, A.;
+and Savenkov, D. 2021.
+RETRONLU: Retrieval Aug-
+mented Task-Oriented Semantic Parsing.
+arXiv preprint
+arXiv:2109.10410.
+Hashimoto, T. B.; Guu, K.; Oren, Y.; and Liang, P. S. 2018. A
+retrieve-and-edit framework for predicting structured outputs.
+Advances in Neural Information Processing Systems, 31.
+Hazoom, M.; Malik, V.; and Bogin, B. 2021. Text-to-SQL in
+the Wild: A Naturally-Occurring Dataset Based on Stack Ex-
+change Data. In Proceedings of the 1st Workshop on Natural
+Language Processing for Programming (NLP4Prog 2021),
+77–87. Online: Association for Computational Linguistics.
+Hossain, N.; Ghazvininejad, M.; and Zettlemoyer, L. 2020.
+Simple and effective retrieve-edit-rerank text generation. In
+Proceedings of the 58th Annual Meeting of the Association
+for Computational Linguistics, 2532–2538.
+Huang, X. S.; Perez, F.; Ba, J.; and Volkovs, M. 2020. Improv-
+ing Transformer Optimization Through Better Initialization.
+In III, H. D.; and Singh, A., eds., Proceedings of the 37th
+International Conference on Machine Learning, volume 119
+of Proceedings of Machine Learning Research, 4475–4483.
+PMLR.
+Khandelwal, U.; Fan, A.; Jurafsky, D.; Zettlemoyer, L.; and
+Lewis, M. 2021. Nearest Neighbor Machine Translation. In
+International Conference on Learning Representations.
+Khandelwal, U.; Levy, O.; Jurafsky, D.; Zettlemoyer, L.; and
+Lewis, M. 2020. Generalization through Memorization: Near-
+est Neighbor Language Models. In International Conference
+on Learning Representations.
+Le Scao, T.; and Rush, A. M. 2021. How many data points is
+a prompt worth? In Proceedings of the 2021 Conference of
+the North American Chapter of the Association for Compu-
+tational Linguistics: Human Language Technologies, 2627–
+2636.
+Lee, C.-H.; Polozov, O.; and Richardson, M. 2021. Kag-
+gleDBQA: Realistic Evaluation of Text-to-SQL Parsers. In
+Proceedings of the 59th Annual Meeting of the Association
+for Computational Linguistics and the 11th International
+Joint Conference on Natural Language Processing (Volume
+1: Long Papers), 2261–2273. Online: Association for Com-
+putational Linguistics.
+Liu, Y.; Ott, M.; Goyal, N.; Du, J.; Joshi, M.; Chen, D.;
+Levy, O.; Lewis, M.; Zettlemoyer, L.; and Stoyanov, V. 2020.
+Ro{BERT}a: A Robustly Optimized {BERT} Pretraining
+Approach.
+Min, S.; Lyu, X.; Holtzman, A.; Artetxe, M.; Lewis, M.;
+Hajishirzi, H.; and Zettlemoyer, L. 2022. Rethinking the
+Role of Demonstrations: What Makes In-Context Learning
+Work? arXiv preprint arXiv:2202.12837.
+Pasupat, P.; Zhang, Y.; and Guu, K. 2021. Controllable Se-
+mantic Parsing via Retrieval Augmentation. In Proceedings
+of the 2021 Conference on Empirical Methods in Natural
+Language Processing, 7683–7698.
+Pawlik, M.; and Augsten, N. 2015. Efficient computation
+of the tree edit distance. ACM Transactions on Database
+Systems (TODS), 40(1): 1–40.
+Pawlik, M.; and Augsten, N. 2016. Tree edit distance: Robust
+and memory-efficient. Information Systems, 56: 157–173.
+Perez, E.; Kiela, D.; and Cho, K. 2021. True few-shot learn-
+ing with language models. Advances in Neural Information
+Processing Systems, 34.
+Poesia, G.; Polozov, O.; Le, V.; Tiwari, A.; Soares, G.; Meek,
+C.; and Gulwani, S. 2022.
+Synchromesh: Reliable code
+generation from pre-trained language models. arXiv preprint
+arXiv:2201.11227.
+Popescu, A.-M.; Etzioni, O.; and Kautz, H. 2003. Towards a
+Theory of Natural Language Interfaces to Databases. In Pro-
+ceedings of the 8th International Conference on Intelligent
+User Interfaces, 149–157.
+
+Raffel, C.; Shazeer, N.; Roberts, A.; Lee, K.; Narang, S.;
+Matena, M.; Zhou, Y.; Li, W.; and Liu, P. J. 2020. Explor-
+ing the Limits of Transfer Learning with a Unified Text-to-
+Text Transformer. Journal of Machine Learning Research,
+21(140): 1–67.
+Rubin, O.; and Berant, J. 2021. SmBoP: Semi-autoregressive
+Bottom-up Semantic Parsing. In Proceedings of the 2021
+Conference of the North American Chapter of the Association
+for Computational Linguistics: Human Language Technolo-
+gies, 311–324.
+Scholak, T.; Li, R.; Bahdanau, D.; de Vries, H.; and Pal, C.
+2021. DuoRAT: Towards Simpler Text-to-SQL Models. In
+Proceedings of the 2021 Conference of the North American
+Chapter of the Association for Computational Linguistics:
+Human Language Technologies, 1313–1321.
+Scholak, T.; Schucher, N.; and Bahdanau, D. 2021. PICARD
+- Parsing Incrementally for Constrained Auto-Regressive De-
+coding from Language Models. In Proceedings of the 2021
+Conference on Empirical Methods in Natural Language Pro-
+cessing. Association for Computational Linguistics.
+Shaw, P.; Uszkoreit, J.; and Vaswani, A. 2018. Self-Attention
+with Relative Position Representations. In Proceedings of the
+2018 Conference of the North American Chapter of the As-
+sociation for Computational Linguistics: Human Language
+Technologies, Volume 2 (Short Papers), 464–468. New Or-
+leans, Louisiana: Association for Computational Linguistics.
+Shin, T.; Razeghi, Y.; Logan IV, R. L.; Wallace, E.; and
+Singh, S. 2020. AutoPrompt: Eliciting Knowledge from
+Language Models with Automatically Generated Prompts. In
+Proceedings of the 2020 Conference on Empirical Methods
+in Natural Language Processing (EMNLP), 4222–4235.
+Socher, R.; Chen, D.; Manning, C. D.; and Ng, A. 2013.
+Reasoning with neural tensor networks for knowledge base
+completion. Advances in neural information processing sys-
+tems, 26.
+Suhr, A.; Chang, M.-W.; Shaw, P.; and Lee, K. 2020. Explor-
+ing unexplored generalization challenges for cross-database
+semantic parsing. In Proceedings of the 58th Annual Meeting
+of the Association for Computational Linguistics, 8372–8388.
+Tang, L. R.; and Mooney, R. J. 2000. Automated Construc-
+tion of Database Interfaces: Intergrating Statistical and Re-
+lational Learning for Semantic Parsing. In 2000 Joint SIG-
+DAT Conference on Empirical Methods in Natural Language
+Processing and Very Large Corpora, 133–141. Hong Kong,
+China: Association for Computational Linguistics.
+Wang, B.; Shin, R.; Liu, X.; Polozov, O.; and Richardson,
+M. 2020. RAT-SQL: Relation-Aware Schema Encoding and
+Linking for Text-to-SQL Parsers.
+In Proceedings of the
+58th Annual Meeting of the Association for Computational
+Linguistics, 7567–7578.
+Wolf, T.; Debut, L.; Sanh, V.; Chaumond, J.; Delangue, C.;
+Moi, A.; Cistac, P.; Rault, T.; Louf, R.; Funtowicz, M.; Davi-
+son, J.; Shleifer, S.; von Platen, P.; Ma, C.; Jernite, Y.; Plu, J.;
+Xu, C.; Scao, T. L.; Gugger, S.; Drame, M.; Lhoest, Q.; and
+Rush, A. M. 2020. Transformers: State-of-the-Art Natural
+Language Processing. In Proceedings of the 2020 Confer-
+ence on Empirical Methods in Natural Language Processing:
+System Demonstrations, 38–45. Online: Association for Com-
+putational Linguistics.
+Xie, T.; Wu, C. H.; Shi, P.; Zhong, R.; Scholak, T.; Yasunaga,
+M.; Wu, C.-S.; Zhong, M.; Yin, P.; Wang, S. I.; et al. 2022.
+UnifiedSKG: Unifying and Multi-Tasking Structured Knowl-
+edge Grounding with Text-to-Text Language Models. arXiv
+preprint arXiv:2201.05966.
+Xu, P.; Kumar, D.; Yang, W.; Zi, W.; Tang, K.; Huang, C.;
+Cheung, J. C. K.; Prince, S. J.; and Cao, Y. 2021. Optimiz-
+ing Deeper Transformers on Small Datasets. In Proceedings
+of the 59th Annual Meeting of the Association for Compu-
+tational Linguistics and the 11th International Joint Con-
+ference on Natural Language Processing (Volume 1: Long
+Papers), 2089–2102. Online: Association for Computational
+Linguistics.
+Yu, T.; Wu, C.-S.; Lin, X. V.; Tan, Y. C.; Yang, X.; Radev,
+D.; Xiong, C.; et al. 2020. GraPPa: Grammar-Augmented
+Pre-Training for Table Semantic Parsing. In International
+Conference on Learning Representations.
+Yu, T.; Zhang, R.; Yang, K.; Yasunaga, M.; Wang, D.; Li,
+Z.; Ma, J.; Li, I.; Yao, Q.; Roman, S.; et al. 2018. Spider:
+A Large-Scale Human-Labeled Dataset for Complex and
+Cross-Domain Semantic Parsing and Text-to-SQL Task. In
+Proceedings of the 2018 Conference on Empirical Methods
+in Natural Language Processing, 3911–3921.
+Zelle, J. M.; and Mooney, R. J. 1996. Learning to Parse
+Database Queries using Inductive Logic Programming. In
+AAAI/IAAI, 1050–1055. Portland, OR: AAAI Press/MIT
+Press.
+Zhao, Z.; Wallace, E.; Feng, S.; Klein, D.; and Singh, S. 2021.
+Calibrate Before Use: Improving Few-shot Performance of
+Language Models. In Meila, M.; and Zhang, T., eds., Pro-
+ceedings of the 38th International Conference on Machine
+Learning, volume 139 of Proceedings of Machine Learning
+Research, 12697–12706. PMLR.
+
+A
+Appendix
+A.1
+Details of the retriever used with ConcatCBR
+The ConcatCBR baseline discussed in Section 4 first re-
+trieves Text-SQL pairs for a given utterance ¯x, which are
+then concatenated with the utterance ¯x as an in input to
+model’s encoder. Similar to Das et al. (2021); Poesia et al.
+(2022), we train a RoBERTa-BASE based sentence embed-
+ding model E : X �→ Rd, that retrieves Text-SQL pairs
+{¯xr, ¯qr} from cases or training data having higher cosine
+similarty with the input utterance ¯x. The retriever is trained
+independent of Text-to-SQL models. For any two Text-SQL
+pairs (¯xi, ¯qi) and (¯xj, ¯qj) in the train set, we utilize normal-
+ized tree-edit-distance TED(¯qi, ¯qj) ∈ [0, 1] between queries
+(¯qi, ¯qj) to supervise the cosine-similarity scores between sen-
+tence embeddings CSim(E(¯xi), E(¯xj)), such that pairs with
+low tree-edit-distance have higher cosine-similarity. We uti-
+lize APTED library (Pawlik and Augsten 2015, 2016) to
+compute tree-edit-distance between the relational algebra
+trees of the SQL queries. We modify the cost function of
+tree-edit-distance to ignore the leaf values and constants in
+the tree so that structurally similar trees are assigned lower
+distance. We normalize tree-edit-distance by size of the larger
+of two trees, so it lies in range [0, 1].
+More concretely, for a given Text-SQL pair in a training batch
+(¯xi, ¯qi) we sample 15 more Text-SQL pairs {(¯xj, ¯qj)}15
+j=1.
+Then we compute tree-edit-distances {TED(¯qi, ¯qj)}15
+j=1, and
+cosine similarities {CSim(E(¯xi), E(¯xj))}15
+j=1 between sen-
+tences embeddings. The cosine-similarity scores are now
+supervised using tree-edit-distances as per loss in Equation 9.
+wi,j =
+exp(1 − 2 TED(¯qi, ¯qj))
+�
+j exp(1 − 2 TED(¯qi, ¯qj))
+LE = −
+�
+i,j
+wi,j log
+exp(CSim(E(¯xi), E(¯xj)))
+�
+j exp(CSim(E(¯xi), E(¯xj)))
+(9)
+During inference on a new schema, we update the retrieval
+index with available cases from the new schema. Now, given
+a text query ¯x, we retrieve the 5 most similar examples as per
+cosine-similarity.
+A.2
+ConcatCBR baseline using T5
+For a fair comparison, all the baselines in Section 5 are built
+on the top of the SmBoP architecture that utilizes non-auto-
+regressive bottom-up decoding. However, ConcatCBR ar-
+chitectures in prior works (Das et al. 2021; Pasupat, Zhang,
+and Guu 2021) utilize the standard auto-regressive Seq2Seq
+architectures. Hence, to further ensure a fair comparison,
+we explored ConcatCBR baselines built on the top of T5-
+large (Raffel et al. 2020) models. We utilize the implemen-
+tation of UnifiedSKG library (Xie et al. 2022) for T5-large
+based Text-to-SQL models 3 and modify it to concatenate
+input-output examples at T5’s encoder to get the ConcatCBR
+baseline. The T5-large models were intialized with an LM-
+Adapted 4 version. The default T5-large based Text-to-SQL
+3https://github.com/HKUNLP/UnifiedSKG/blob/main/configure/Salesforce/
+T5 large finetune spider with cell value.cfg
+4https://github.com/google-research/text-to-text-transfer-transformer/blob/main/
+released checkpoints.md#lm-adapted-t511lm100k
+world
+car
+cre
+dog
+flights
+avg
+1
+1
+Doc
+kenn
+2
+EM
+T5
+46.4
+40.5
+83
+61.2
+66
+57.3
++ConcatCBR
++2.3
++1.1
+-2.5
++10.9
+-3.4
++1.7
+EX
+T5
+43.1
+41.7
+92.5
+66.7
+77.6
+61
++ConcatCBR
++4.1
++2.8
+-8.2
++7.5
+-5.4
++0.7
+Table A1: EM and EX performance of a T5-large based
+Text-to-SQL model and gains from its ConcatCBR extension.
+Similar to Table 2, we observe that ConcatCBR does not
+provide consistent gains over the unadapted model. For two
+schemas the gains are negative, and gains micro-averaged
+(avg) across five schemas are small.
+model finetuned on Spider dataset has a comparable EM ac-
+curacy of 57.3 w.r.t. SmBoP’s 57.2 micro-averaged across
+five schemas.
+Table A1 presents the EM and EX accuracies of the default
+T5-large based Text-to-SQL model, and gains obtained by its
+ConcatCBR extension. Similar to Table 2, gains from Con-
+catCBR are not consistent. For two out of five databases, we
+observe a drop in EM and EX. The micro-averaged gains
+across all five schemas are also small. Thus, our observations
+remain largely consistent with ConcatCBR baseline based on
+SmBoP.
+A.3
+StructCBR gains relative to finetuning
+Table A2 shows gains of StructCBR over SmBoP rela-
+tive to gains of finetuning over SmBoP. Relative Gain =
++StructCBR
++Finetuning × 100. StructCBR instantly obtains 20% to 83%
+of EM gains achieved by finetuning for 100 epochs across
+different schemas. For 3 out of 5 schemas relative gains are
+more than 50%. Micro-averaged across all examples, Struct-
+CBR achieves 42.8% of gains relative to finetuning. It is
+important to note here that we do not position StructCBR as
+an alternative to finetuning, but instead as a method of uti-
+lizing available cases for instantaneously improving model’s
+performance until the next cycle of finetuning becomes af-
+fordable.
+world
+car
+cre
+dog
+flights
+avg
+1
+1
+Doc
+kenn
+2
+SmBOP
+46
+43.3
+84.3
+66.6
+55.8
+57.2
++StructCBR
++2.6
++6.1
++6.3
++3.4
++5.5
++4.6
++Finetuning
++6.0
++11.7
++7.6
++4.8
++27.2
++10.7
+Relative Gain (%)
+43.9
+52.3
+83.7
+71.5
+20.1
+42.8
+Table A2: StructCBR instantly obtains 20% to 83% EM
+gains achieved by finetuning for 100 epochs across different
+schemas. On 3 out 5 schemas gains are higher than 50% of
+gains from finetuning. Averaged (avg) across all examples,
+gains are 42.8% of finetuning.
+A.4
+Additional hyperparameter and training
+details
+The baseline SmBoP model is based on RoBERTa-base ar-
+chitecture and contains four RAT layers. The total number of
+
+trainable parameters in this more is roughly 133M. Adding
+StructCBR module on top of SmBoP introduces only 2.53%
+additional parameters. In particular, the transformer model
+TXφ used for computing Gφ representations in Equation 2
+is composed of two transformer blocks of hidden size=256,
+attention heads=8, feedforward dim=1024. All the experi-
+ments were performed on a NVIDIA RTX 3060 GPU, and
+a NVIDIA RTX A6000. Training times for various Text-to-
+SQL models usually varied from 12 hours to 16 hours.
+A.5
+Results with a larger model size
+Constrained by limited computing resources for training
+large models, we conducted all the experiments and abla-
+tions in Section 5 using SmBoP models initialized with a
+pretrained RoBERTa-BASE checkpoint (Liu et al. 2020) fol-
+lowed by 4 RAT layers (Wang et al. 2020). In this section,
+we validate our key results on a larger model size, by repeat-
+ing the experiment reported in Table 2 of Section 5. More
+specifically, we replace RoBERTa-BASE in SmBoP with a
+pretrained RoBERTa-LARGE checkpoint based on grammar
+augmented pre-training (Yu et al. 2020) as used by Rubin
+and Berant (2021). The number of RAT layers is also in-
+creased from 4 to 8. We refer to the larger SmBoP architec-
+ture as SmBoP-LARGE (367M parameters) and the base-sized
+SmBoP model used in Section 5 as SmBoP-BASE (133M
+parameters). Similar to Table 2, Table A3 compares differ-
+ent inference-time methods for adapting the SmBoP-LARGE
+model to new schemas. The performance of SmBoP-BASE
+is reported just as a reference. For each adaptation method
+(ConcatCBR, GTM, StructCBR), we report it’s gains over
+the SmBoP-LARGE model. We consistently observe positive
+gains from StructCBR across all the metrics on four out of
+five schemas. In contrast, the gains from prior inference time
+adaptation methods are not consistently positive across dif-
+ferent metrics and schemas. It is also interesting to note that
+the gains in top-K accuracy metrics (BEM and BEX) from
+StructCBR are significantly higher than prior methods and
+consistently positive across all the schemas. Better perfor-
+mance in top-K metrics indicates that boosting candidate
+subtree scores via StructCBR improves the recall of correct
+subtrees during beam decoding.
+A.6
+Additional Related Work
+Incontext Learning with LLMs:
+Learning new tasks via
+carefully designed exemplar prompts (Shin et al. 2020;
+Le Scao and Rush 2021; Perez, Kiela, and Cho 2021) for
+large language models (LLMs) like GPT-3 (Brown et al.
+2020) and Codex (Chen et al. 2021) has shown promise for
+many tasks (Zhao et al. 2021; Min et al. 2022) without requir-
+ing to finetune model parameters. For Text-to-SQL, Poesia
+et al. (2022) retrieve and encode input-specific prompts in
+a way similar to the ConcatCBR method. They differ from
+ConcatCBR as they do not explicitly train the LLM to per-
+form CBR with the extracted propmts. While their method
+achieves impressive performance by using LLMs, training
+significantly (100x) smaller Text-to-SQL models on task-
+specific datasets like Spider outperforms their LLMs by con-
+siderable margins.
+EM
+EX
+BEM
+BEX
+snew = world 1 , |Dtest| = 89
+SmBoP-BASE
+46.1
+36.3
+55.4
+49.0
+SmBoP-LARGE
+50.6
+42.3
+58.1
+49.8
+ConcatCBR
+-2.3
+-0.4
+-1.9
++0.7
+GTM
+-0.8
+-0.8
++0.7
++1.1
+StructCBR (Ours)
+-1.5
+-0.7
++3.4
++2.2
+snew = car 1 , |Dtest| = 60
+SmBoP-BASE
+43.3
+45.6
+50.6
+53.9
+SmBoP-LARGE
+53.3
+57.2
+66.7
+67.8
+ConcatCBR
++3.9
+-6.1
+-2.2
+-2.8
+GTM
++6.6
++3.3
++3.3
+-0.5
+StructCBR (Ours)
++3.3
++3.3
++1.1
++1.7
+snew = cre Doc Template Mgt , |Dtest| = 53
+SmBoP-BASE
+84.3
+89.3
+94.3
+96.2
+SmBoP-LARGE
+90.6
+95.6
+96.2
+98.1
+ConcatCBR
++3.7
++0.6
++3.2
++1.3
+GTM
+-1.9
+-1.2
++1.3
++0.6
+StructCBR (Ours)
++3.1
++1.9
++3.2
++1.9
+snew = dog kennels , |Dtest| = 49
+SmBoP-BASE
+66.6
+59.2
+80.3
+74.8
+SmBoP-LARGE
+70.1
+62.6
+81.0
+74.1
+ConcatCBR
++1.3
++2.7
++2.7
++6.2
+GTM
+-0.7
+-0.7
+-1.4
++4.8
+StructCBR (Ours)
++2.1
++3.4
++5.4
++8.2
+snew = flight 2 , |Dtest| = 49
+SmBoP-BASE
+55.8
+67.4
+59.2
+80.3
+SmBoP-LARGE
+59.9
+65.3
+66.0
+75.5
+ConcatCBR
++6.1
++10.9
++5.4
++11.6
+GTM
++1.3
++0.7
++4.7
++7.5
+StructCBR (Ours)
++4.1
++3.4
++14.9
++12.2
+Micro-Average , |Dtest| = 300
+SmBoP-BASE
+57.2
+56.3
+66.0
+67.6
+SmBoP-LARGE
+62.6
+61.8
+71.3
+70.1
+ConcatCBR
++2.3
++1.0
++1.1
++2.8
+GTM
++1.2
++0.2
++1.9
++2.3
+StructCBR (Ours)
++2.1
++1.9
++5.3
++4.7
+Table A3: Results using SmBoP-LARGE: We compare Struct-
+CBR with prior inference time adaptation methods (Con-
+catCBR and GTM) for adapting SmBoP-LARGE on 5 differ-
+ent schemas. SmBoP-BASE and SmBoP-LARGE rows pro-
+vide the performance of unadapted model SmBoP models,
+while other rows report gains over SmBoP-LARGE after adap-
+tation. SmBoP-BASE numbers are just of reference. Micro-
+average refers to numbers averaged over all the test instances
+spanning across the five schemas. |Dtest| refers to size of the
+test-set, and snew provides the schema name.
+
+A.7
+Anecdotes
+We include some anecdotal examples where StructCBR fixes the mistakes made by SmBoP by utilizing case examples.
+Text
+What is the code of airport that has the highest number of flights?
+Incorrect SQL
+SELECT flights.sourceairport FROM flights GROUP BY flights.sourceairport ORDER
+BY SUM( flights.flightno ) DESC LIMIT 1
+Corrected SQL
+SELECT airports.airportcode FROM airports JOIN flights ON airports.airportcode
+= flights.sourceairport GROUP BY flights.sourceairport ORDER BY COUNT( * ) DESC
+LIMIT 1
+Case Text
+Give the code of the airport with the least flights.
+Case SQL
+SELECT airports.airportcode FROM airports JOIN flights ON airports.airportcode
+= flights.sourceairport GROUP BY flights.sourceairport ORDER BY COUNT( * ) ASC
+LIMIT 1
+Mistake(s) Fixed
+Correct column selection, Add missing join, Replace SUM by COUNT
+Text
+Which airlines have a flight with source airport AHD?
+Incorrect SQL
+SELECT flights.airline FROM airlines JOIN flights ON airlines.uid =
+flights.airline JOIN airports ON flights.sourceairport = airports.airportcode
+WHERE airports.airportname = ’AHD’
+Corrected SQL
+SELECT airlines.airline FROM airlines JOIN flights ON airlines.uid =
+flights.airline WHERE flights.sourceairport = ’AHD’
+Case Text
+What are airlines that have flights arriving at airport ’AHD’?
+Case SQL
+SELECT airlines.airline FROM airlines JOIN flights ON airlines.uid =
+flights.airline WHERE flights.destairport = ’AHD’
+Mistake(s) Fixed
+Drop additional condition on Join
+Text
+What are the population and life expectancies in Brazil?
+Incorrect SQL
+SELECT country.population , country.lifeexpectancy FROM city JOIN country ON
+city.countrycode = country.code WHERE country.name = ’Brazil’
+Corrected SQL
+SELECT country.lifeexpectancy , country.population FROM country WHERE
+country.name = ’Brazil’
+Case Text
+What are the region and population of Angola?
+Case SQL
+SELECT Population , Region FROM country WHERE Name = ’Angola’
+Mistake(s) Fixed
+Drop the join condition
+Text
+Return the codes of countries that do not speak English and do not have Republics for governments.
+Incorrect SQL
+SELECT country.code FROM country WHERE countrylanguage.language =
+’English’ INTERSECT SELECT countrylanguage.countrycode FROM country WHERE
+country.governmentform != ’Republic’
+Corrected SQL
+SELECT country.code FROM country WHERE country.governmentform != ’Republic’
+EXCEPT SELECT countrylanguage.countrycode FROM countrylanguage WHERE
+countrylanguage.language = ’English’
+Case Text
+What are the codes of the countries that do not speak English and whose government forms are not Republic?
+Case SQL
+SELECT country.code FROM country WHERE country.governmentform != ’Republic’
+EXCEPT SELECT countrylanguage.countrycode FROM countrylanguage WHERE
+countrylanguage.language = ’English’
+Mistake(s) Fixed
+Set operation and Equality
+Table A4: Anecdotal examples where mistakes in output SQLs generated by SmBoP are fixed with help of StructCBR through
+related examples in case memory. We show only one of the related examples from cases for brevity
+
diff --git a/edE2T4oBgHgl3EQfxQji/content/tmp_files/load_file.txt b/edE2T4oBgHgl3EQfxQji/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1ef55f071e2df2ef34b7793a51d87f06406d4664
--- /dev/null
+++ b/edE2T4oBgHgl3EQfxQji/content/tmp_files/load_file.txt
@@ -0,0 +1,1544 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf,len=1543
+page_content='Structured Case-based Reasoning for Inference-time Adaptation of Text-to-SQL parsers Abhijeet Awasthi, Soumen Chakrabarti, Sunita Sarawagi Department of Computer Science and Engineering Indian Institute of Technology Bombay, Mumbai, India {awasthi,soumen,sunita}@cse.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='iitb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='in Abstract Inference-time adaptation methods for semantic parsing are useful for leveraging examples from newly-observed domains without repeated fine-tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Existing approaches typically bias the decoder by simply concatenating input-output exam- ple pairs (cases) from the new domain at the encoder’s input in a Seq-to-Seq model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Such methods cannot adequately leverage the structure of logical forms in the case examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We pro- pose StructCBR, a structured case-based reasoning approach, which leverages subtree-level similarity between logical forms of cases and candidate outputs, resulting in better decoder deci- sions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For the task of adapting Text-to-SQL models to unseen schemas, we show that exploiting case examples in a struc- tured manner via StructCBR offers consistent performance improvements over prior inference-time adaptation methods across five different databases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' To the best of our knowledge, we are the first to attempt inference-time adaptation of Text- to-SQL models, and harness trainable structured similarity between subqueries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 1 Introduction Natural language interfaces to databases (Zelle and Mooney 1996;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tang and Mooney 2000;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Popescu, Etzioni, and Kautz 2003) enable access to structured information for users who are not familiar with languages like SQL by parsing user provided text-queries into executable SQLs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Text-to-SQL se- mantic parsing is a challenging task that not only demands robust natural language understanding but simultaneously requires reasoning over the schema structure of the databases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Databases containing similar information (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' census in vari- ous countries) may be designed using diverse schema struc- tures, thus making it hard for the model to generalize across schemas unseen during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hence, Text-to-SQL models often struggle to parse text queries for a new schema in a zero- shot manner (Suhr et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lee, Polozov, and Richardson 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hazoom, Malik, and Bogin 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In practice, a small number of Text-to-SQL examples in the target schema are often essential for successful model adaptation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' However, finetuning a Text-to-SQL model for each new database is not generally practical, for the following reasons: (i) Huge variation in database schema makes it tedious to collect suf- ficiently large finetuning datasets for each schema, while Copyright © 2023, Association for the Advancement of Artificial Intelligence (www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='aaai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='org).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' All rights reserved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' finetuning on small datasets is unavoidably fraught with over– fitting, catastrophic forgetting, and instability w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' random seeds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (ii) Finetuning may take considerable time, preventing fast incorporation of new data into the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (iii) Often, a single large-footprint model serves multiple clients with diverse databases at the same time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Fine-tuning a separate model for each database is considered too resource-intensive in such multi-tenant scenarios.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Therefore, we focus on fast online adaptation of Text- to-SQL models without parameter updates, until the next cycle of finetuning is deemed feasible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Recently, case-based reasoning (CBR), which utilizes a memory of past labeled examples as cases, has emerged as a promising paradigm of inference-time adaptation without finetuning (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020, 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pasupat, Zhang, and Guu 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gupta et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' CBR has been found effective for tasks like knowledge graph completion (KGC) (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020), question answering over knowledge bases (KBQA) (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021), task-oriented semantic parsing (Pasupat, Zhang, and Guu 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gupta et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021), translation (Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021), and text-based games (Atzeni et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' However, many prior CBR approaches designed around Seq2Seq architectures simply concatenate input-output cases with the current input at the encoder (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pasupat, Zhang, and Guu 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gupta et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' These methods do not leverage the structure of logical forms (query plan trees) in case examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In response, we propose StructCBR, a structured CBR approach that directly exploits sub-tree level similarities between the candidate outputs and the case examples for adapting a Text-to-SQL model to a new schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We start with SmBoP (Rubin and Berant 2021), a recent semi-auto-regressive architecture that decodes query trees bottom-up, respecting the structure of SQL grammar production rules, instead of left-to-right token-level decoding in Seq2Seq models (Guo et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak, Schucher, and Bahdanau 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We implement a novel structured case memory lookup module to boost scores of promising candidate trees using sub-tree level similarity with case trees under similar input context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' This similarity is trainable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We show that explicitly-learned structured memory lookup leads to more accurate transfer from cases, compared to prior inference-time adaptation methods such as ConcatCBR and arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='04110v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='CL] 10 Jan 2023 GTM (Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021) that we implemented both on SmBoP, and other Seq2Seq Text-to-SQL architectures like T5-large.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We summarize our contributions as follows: 1) We propose StructCBR, which, to our knowledge, is the first inference-time adaptation method for Text-to-SQL pars- ing without parameter fine-tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2) StructCBR incorporates a novel structured case memory and trainable query subtree similarity module that can boost scores of likely-correct outputs during inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' This is in contrast with earlier approaches like ConcatCBR and GTM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 3) We propose a trainable compositional sub-tree similarity function that is both more accurate and more efficient for scoring large search frontiers, compared to default whole- tree embeddings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 4) Through experiments with five database schemas (§ 5) of varying complexity, we observe that StructCBR is consis- tently better than prior inference-time adaptation methods on both SmBoP and sequence-based Text-to-SQL models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 5) We show that StructCBR provides almost instant adapta- tion to a target schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In contrast, finetuning (§ 5) can be up to 500 times slower.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2 SmBoP preliminaries We present a brief background on SmBoP here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Readers fa- miliar with SmBoP can largely skip this section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP converts a natural language question ¯x ∈ X (called the ‘utter- ance’) targeting a database schema ¯s ∈ S, to an SQL query ˆq ∈ Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We describe the major modules in SmBoP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Utterance and schema encoding: Given token sequence ¯x = [x1, x2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' , xn] in the text query, and database schema ¯s = [s1, s2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' , sm] denoting table and column names, SmBoP jointly encodes them using a pre-trained Transformer like RoBERTa (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) followed by relation-aware- transformer (RAT) layers (Shaw, Uszkoreit, and Vaswani 2018;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We denote the output from the last encoder layer as ¯x = [x1, x2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' xn] and ¯s = [s1, s2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' sm], representing the jointly encoded contextual embeddings of text tokens and schema elements respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Decoding SQL output: Unlike standard sequence-based decoding (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak, Schucher, and Bahdanau 2021), SmBoP decodes the SQL tree bottom-up and in layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP views any SQL query as a height-balanced relational algebra tree converted using a special idempotent KEEP operator κ as shown in Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Given a beam size K, at decoding step t, the decoder beam Bt comprises K candidate sub-trees of height t from the bottom.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' At step t + 1, trees from Bt are grown either via unary operators (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' COUNT), or by combining two trees in Bt using a binary operator (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' >), as per the SQL grammar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The candidate trees at step t + 1 form a frontier set Ft+1 and is of size |Ft+1| = K2|B| + K|U|, where B and U represent the set of binary and unary operations respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP assigns each candidate tree z ∈ Ft+1 a score sθ(z) age 60 > actors 𝜿 𝝈 name 𝜿 𝜿 𝚷 t=1 t=2 t=3 t=4 age 60 > actors 𝜅 𝜎 name 𝜅 𝜅 Π t=1 t=2 t=3 t=4 Figure 1: SmBoP (Rubin and Berant 2021) decodes SQL as a balanced relational algebra tree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' At each level t, trees in the beam combine via unary or binary operators to form candidates of the next beam.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' StructCBR leverages CBR on generated sub-trees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Text 1 Give the code of the airport with the fewest number of flights SmBoP output SELECT sourceairport FROM flights GROUP BY sourceairport ORDER BY SUM(flightno) ASC LIMIT 1 Correct SQL SELECT airportcode FROM airports JOIN flights ON airportcode = sourceairport GROUP BY sourceairport ORDER BY COUNT(*) ASC LIMIT 1 Text 2 What is the code of the airport that has the highest number of flights?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Table 1: Illustration of the lack of generalization of Text-to- SQL to new schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (described below).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The top-K highest scoring trees in Ft+1 form the next beam Bt+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' This continues up to a maximum height T, when the highest scoring tree in BT is output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scoring a tree: A tree z = (zb, zℓ, zr) consists of root op- erator zb and subtrees zℓ, zr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP encodes a variable-size tree z into two fixed dimensional vectors: (i) z: an embedding of the tree computed recursively on the tree structure, where a transformer outputs z = TXθ([zb, zℓ, zr]);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (ii) z′: a contex- tual representation of z grounded in input text ¯x computed via a multiheaded cross attention module z′ = XAttθ(z, ¯x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP computes the score of a tree z ∈ Ft+1, as follows: sθ(z) = wT zb FFθ([zℓ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' z′ ℓ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' zr;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' z′ r]) (1) where FFθ is a feed forward network, and wzb represents a learned embedding of operator zb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The model is trained using Text-SQL pairs from a set of training schema to maximize the likelihood of the correct sub- trees at each beam.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' During inference, when presented with text utterances relating to a new database schema, the model often fails to discover the mapping of the text to schema names and relationships in the new schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Table 1 presents an example where a SmBoP model trained on the Spider dataset (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018) is deployed on a new schema about flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' On inspecting the predicted and correct SQL, we find that the model failed to reason that number of flights requires a count(*) instead of sum(flightno).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Now suppose an expert provides the correct SQL as additional in- formation to be used during inference of subsequent queries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Consider a second query (shown as Text 2 in Table 1) that name actors age 60K K >KIalso needs to reason about number of flights, and the de- fault SmBoP makes similar errors (not shown).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Only existing mechanism in SmBoP is to fine-tune parameters which could be time-consuming and unstable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In the next section we show how our method can instantaneously leverage test-time user labels to predict the correct SQL for Text 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' More such anec- dotes appear in Table A4 of the Appendix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 3 Our proposed method: StructCBR We aim to learn a Text-to-SQL model M, using a dataset Dtrain of Text-SQL pairs such that it is capable of C1: Inde- pendently translating the text queries ¯x to executable SQL programs ˆq, and C2: Utilizing a small set Dnew of Text-SQL pairs from a target schema snew, to improve its own predic- tions during inference, without finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In line with prior work (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020, 2021), we refer to the second capabil- ity C2 as Case-based reasoning (CBR), and the dataset Dnew of Text-SQL pairs in the target schema as cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The StructCBR module leverages the similarity between gold subtrees that appear in similar contexts in the set of cases Dnew and the candidate subtrees in SmBoP’s frontier Ft+1, to boost the scores of likely-correct candidates at each decoding step t + 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Consider a subtree z in the frontier Ft+1 for an input text ¯x, a case-example with text question as ¯xc, and the gold SQL tree as Zc gold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Let zc be a subtree of Zc gold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The key idea of StructCBR is, if z and zc are structurally similar, and appear in similar contexts w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' ¯x and ¯xc, then there is a strong evidence that the subtree z should also appear as a part of the gold tree Zgold of ¯x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Figure 2 provides an illustration with z = age > 60 in the candidate frontier Ft+1, and a similarly structured case tree zc = age > 80 appearing in a similar context ¯xc (both contain the phrase who are past).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Even though the key idea of matching with case sub-trees is simple, several important design choices had to be made to ensure that CBR inter-operates efficiently with SmBoP’s own scoring, and consistently improves its performance in the presence of multiple cases of varying levels of related- ness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' First, how should we compute the contextual similarity of a candidate tree z with a case tree, given that memory would also contain unrelated cases that would match wrong candidate trees?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Second, how can we efficiently compute the similarity of all candidate trees with all entries in the case memory?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Unlike Seq2Seq models that do not perform beam-search during training, SmBoP generates a large search frontier even during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We elaborate on how our design tackles these challenges next.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Algorithm 1 presents the high-level pseudo code, with the text in blue font representing the StructCBR additions to the SmBoP model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 Choosing tree representations We need to choose a representation of a tree z using which we can efficiently compute similarity with case trees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Just the structural similarity of z with a case zc is not sufficient unless we also contextualize them on their respective inputs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Accordingly, we design an embedding function Gφ(z, ¯x) �→ Rd that jointly encodes a candidate tree z corresponding to Algorithm 1: SmBoP with StructCBR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 1 input: ¯x, ¯s, Dnew 2 M ← CreateCaseMemory(Dnew) (§ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2) 3 ¯x,¯s ← EncodeTextSchemaθ(¯x, ¯s) 4 B0 ← top-Kschema constants and DB values 5 for t ← 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' T − 1 do 6 z ← CreateTreeReps(z) 7 z′ ← GroundTreeReps(z, x) 8 pθ ← SmBoPScores(z, z′) (§ 2) 9 Gφ(z, ¯x) ← JointReps(z, z′, x) (Eqn 2) 10 simφ ← TreeSim(Gφ(z, ¯x), M) (Eqn 4) 11 pφ ← StructCBRScores(simφ, M) (Eqn 5) 12 Ft+1 ← CombineScores(pθ, pφ) (Eqn 6) 13 Bt+1 ← top-K(Ft+1) 14 return argmaxz(BT ) an input ¯x as a d dimensional vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We train a separate transformer model TXφ with parameters φ that takes as input four vectors: z that encodes the structure of the tree z, z′ that is the contextual representation of z defined in § 2, an embedding wb of z’s root node b, and pool(x) a mean-pooled version of the input text representation ¯x: Gφ(z, ¯x) = TXφ([z, z′, wb, pool(x)]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2) This embedding captures both the structure and context and the parameters φ are trained to co-embed similar trees in matching contexts, while pulling apart pairs differing either structurally or contextually.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For example, in Figure 2 if the query text was Name all actors who are 60 or above, then the similarity of candidate age > 60 from the same case sub-tree should be reduced.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Unlike recursive tree representations (Socher et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2013), here contextualiza- tion w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' ¯x plays a critical role.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 Case memory design We construct a case memory M over the gold SQL trees {Zc gold} for all cases in Dnew.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Corresponding to each node b of a gold tree Zc gold we form a subtree rooted at b and including the part of Zc gold below b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Thus, the size of the case memory is the total number of nodes over all gold trees in cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The encoding Gφ(zc, ¯xc) of each subtree zc for a case (¯xc, Zc gold) in Dnew is pre-computed using Equation 2 and stored in M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 Efficient tree similarity computation We need to compute the similarity of each tree z in the frontier Ft+1 with all case sub-trees zc ∈ M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' One way to compute similarity between trees z and zc is based on ℓ2 distance1 between their Gφ representations as follows: simφ(z, zc, ¯x, ¯xc) = − ∥Gφ(z, ¯x) − Gφ(zc, ¯xc)∥2 (3) However, computing Gφ representations for each tree z ∈ Ft+1 entails large memory and compute costs since the fron- tier size |Ft+1| = K2|B| + K|U| is quadratic in beam-size 1Like Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2020) we observed better results with ℓ2 distance, in comparison to inner product.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Text 𝑥̅ : Name all the actors who are past 60 | Schema (𝑠̅): || T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='actor | id | name | age | dob || T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='director | id | age | movies … Case Text ( 𝑥̅� ): Show the directors names who are past 80 Case SQL ( 𝑞�� ): SELECT name FROM director WHERE age>80 director 𝜿 𝝈 age 80 > director 𝜿 𝝈 name 𝜿 𝜿 𝚷 age 80 > director 𝜿 Encode Text and Schema (RoBERTa) name actor age 60 𝜅 𝜅 ≥ > ≤ = 𝜎 name actor age 60 𝜅 𝜅 ≥ > ≤ = 𝜎 5 7 4 1 2 1 −1 −3 −5 −∞ −1 −∞ −∞ −∞ name 𝜿 Π −3 Π −∞ top-K name 𝜿 actor 𝜿 age 60 ≥ age 60 ≤ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='22 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='19 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='15 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 𝟐𝟓 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='02 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='02 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='01 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='01 Combine Scores top-K name 𝜿 actor 𝜿 age 60 > age 60 ≥ Create Case Memory 𝐹��� 𝐵� 𝐵��� 𝑠� 𝑠� SmBoP Scoring StructCBR Scoring Case Memory 80 > age Decoding step A B C Figure 2: Augmenting SmBoP with StructCBR (Structured Case-based Reasoning): In part A ⃝, the top-K step in SmBoP scoring misses the correct sub-tree age > 60 due to a lower score (score=1) w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' competing sub-trees in the frontier Ft+1 like age ≥ 60 (score=4) and age ≤ 60 (score=2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In part C ⃝, StructCBR creates a memory of all the sub-tree representations available in cases as described in § 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In part B ⃝, StructCBR scores the frontier candidates based on learned tree-similarities w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' the sub-trees in cases as described in § 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 and § 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For example, StructCBR boosts the score of age > 60 because of its high similarity with the case sub-tree age > 80 and similarity of context who are past.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Thus, the top-K step applied on the combined SmBoP and StructCBR scores recovers the correct sub-trees that otherwise may get missed based on SmBoP’s scoring alone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For brevity, we consider only one case-example in this figure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' With the default K for SmBoP being 30, and size of the SmBoP grammar, this translates to around 23 thousand trees per frontier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pruning the frontier Ft+1 based on SmBoP scores alone resulted in poor performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' This led us to design an alternative compositional CBR scoring method that can more efficiently score all candidate trees in the frontier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Our key idea is to compute the similarity between two trees compositionally as a function of similarity between their left and right children respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' This requires only O(K) tree representations for the trees in beam Bt as against K2|B| + K|U| operations of the whole-tree approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Recall that the trees z ∈ Ft+1 are formed by combining trees in beam Bt via SQL operations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A tree z ∈ Ft+1 can thus be represented as z = (zb, zℓ, zr) where zℓ, zr ∈ Bt denote the left and right child respectively, and zb is the root node combining zℓ and zr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' After the beam Bt is created, we compute the embedding Gφ for each tree in Bt using Equation (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Now, the similarity between a candidate tree z = (zb, zℓ, zr) for an input text ¯x, and a case sub-tree zc = (zc b, zc ℓ, zc r) on input text ¯xc in memory M is computed as: � simφ(z, zc, ¯x, ¯xc) = simφ(zℓ, zc ℓ, ¯x, ¯xc) + simφ(zr, zc r, ¯x, ¯xc).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (4) In Section 5, we show that using this similarity function provides better results by allowing the entire frontier to be scored more efficiently in comparison to computing similar- ities based on Equation 3 only for a subset of trees in the frontier pruned based on SmBoP scores.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 Boosting SmBoP frontier with tree similarities To compute an overall score of a candidate tree z ∈ Ft+1 based on its similarity with the case sub-trees in M, we aggregate over all the case sub-trees zc with the same root node (zb = zc b) using a logsumexp operator, which provides us a soft-maxed similarity of z w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' case sub-trees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' sφ(z) = log � c∈M∧zb=zc b exp(� simφ(z, zc, ¯x, ¯xc)) (5) Now every candidate tree z ∈ Ft+1 has two scores: sθ(z) as- signed by default SmBoP and sφ(z) computed by StructCBR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The scores sθ(z) and sφ(z) can lie in very different ranges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Summing them in a normalized probability space provided better results than summing the scores directly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hence, we independently normalize sθ(z) to pθ(z) and sφ(z) to pφ(z) by a softmax operation applied over all trees in the frontier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The combined score of a frontier tree z is: p(z) = (pθ(z) + pφ(z))/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (6) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 Supervising StructCBR During training, we assume availability of training data Dtrain = {(¯xi, ¯si, ¯qi)}N i=1 consisting of utterances ¯xi on a schema ¯si, and the corresponding gold SQL queries ¯qi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We first train the SmBoP model, parameterized as θ, using Dtrain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The training objective of SmBoP for a single example maxi- mizes the likelihood of sub-trees that are part of the tree Zgold corresponding to gold SQL ¯q: Lθ = − T � t=0 � zt∈Zgold log pθ(zt).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (7) Next, we introduce the StructCBR module parameterized as φ on top of the (now frozen) SmBoP model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We ob- served training the StructCBR parameters φ while freezing the learned SmBoP parameters θ to provide slightly better results in comparison to training both θ and φ jointly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The parameters φ are also learned using Dtrain by maximizing the likelihood of the gold subtrees as per the distributions pφ and p through the following loss function: Lφ = − T � t=0 � zt∈Zgold log pφ(zt) + log p(zt) (8) The − log pφ(zt) term maximizes the likelihood of gold trees w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' the CBR distribution pφ, independent of the SmBoP distribution pθ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Similarly, the − log p(zt) term maximize the likelihood of the gold trees w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' the combined distribution p (Eqn 6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' During training, we design each training batch to contain C examples from same schema so that for a given train example, the remaining C − 1 examples serve as the cases from the same schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We train with C = 32 and a batch-size of 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 4 Related work We review prior work on inference-time model adaptation for related tasks and also describe our adaptation of some of these works in the context of Text-to-SQL for comparisons with StructCBR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Concatenating related examples with input: A common approach, that we call ConcatCBR, for utilizing cases dur- ing inference is to concatenate the input-output pair of each case along with the input text at the encoder of a Seq2Seq model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' During training, the decoder is expected to learn to utilize the cases on the encoder side.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2021) utilize ConcatCBR for question answering over knowledge bases, and Pasupat, Zhang, and Guu (2021);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gupta et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2021) utilize ConcatCBR for other semantic parsing tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Con- catCBR is similar to the retrieve and edit framework for structured outputs (Hashimoto et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018) and machine trans- lation (Hossain, Ghazvininejad, and Zettlemoyer 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For the Text-to-SQL task, we implement a ConcatCBR baseline that trains an SmBoP model to use retrieved Text-SQL exam- ples concatenated with the input-text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' During inference, the retrieval index is updated with the case-examples from the target schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Generalization through Memorization (GTM): Khan- delwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2020, 2021) propose a memory look-up based method for adapting pre-trained language and machine trans- lation models to a target domain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Given a target dataset, their method constructs a look-up index by using contextual em- beddings from the pre-trained model as keys and the corre- sponding text tokens as values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' During inference the model scores are interpolated with the similarity scores aggregated over the nearest neighbours in the loop-up index.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For our Text- to-SQL set-up, we implement this baseline using a trained SmBoP model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We memorize the dataset Dnew in the target schema by creating a look-up index with embeddings of child subtrees from SmBoP as keys: [zℓ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' z′ ℓ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' zr;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' z′ r], and their par- ent nodes as values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' During inference, the scores from the SmBoP model are interpolated with neighbour similarities in a way similar to Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Unlike StructCBR and ConcatCBR, this baseline (GTM) does not explicitly train the SmBoP model for utilizing the cases during infer- ence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We discuss other related work in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 5 Experiments We evaluate StructCBR for adapting a Text-to-SQL model to five different target schemas without finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The target schemas are chosen from varying domains.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We compare StructCBR with prior inference-time adaptation methods discussed in § 4, and present an ablation study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We also show that StructCBR enables much faster adaptation of Text-to- SQL models in comparison to finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Datasets: We utilize Spider (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018), which is a collection of Text-to-SQL examples covering 200 unique schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We use the train split of Spider as Dtrain, for train- ing all the models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Dtrain contains 7000 Text-SQL example pairs from 140 databases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For evaluation, we hold out the following five databases containing the most examples from Spider’s dev set 2: {world 1, car 1, cre Doc Template Mgt, dog kennels, flight 2}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The five evaluation databases do not occur in the train set, and belong to sufficiently different do- mains of varying difficulty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The remaining part of the dev set containing 576 examples is used for model selection while training on Dtrain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We hold out 30 randomly selected exam- ples from each of the five selected databases as Dnew (cases) for adaptation, and use the remaining examples as the test set, Dtest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The average size of Dtest is 60, and varies from roughly 50 to 90 examples across the five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' To ensure robust evaluation, we report numbers averaged over three random Dnew/Dtest splits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We also report the numbers micro-averaged over all the 300 test examples across the five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Evaluation metrics: Following prior work (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018), we report Execution Accuracy (EX) and Exact-Set-Match Accuracy (EM) for all the methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' EX returns 1 if executing the gold query ¯q and the predicted query ˆq on the target database gives the same results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' EM compares all the SQL clauses within ¯q and ˆq and returns 1 if all the clauses match, except possibly the DB-values (constants) in the SQL query.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Most Text-to-SQL models utilize beam search, and return the top-K highest scoring candidates in the beam as the output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hence, we also report the top-K versions of EM and EX metrics as BEM and BEX respectively, where K is the beam size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In our experiments, K = 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' BEM/BEX for a beam is 1, if at least one of the candidates in the beam has an EM/EX of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2Spider’s test set is publicly inaccessible as of 08/15/2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Methods compared: We compare the accuracy of Struct- CBR after adaptation with the following methods: (i) SmBoP: The base model without any adaptation to benchmark the gains from different inference-time adaptation methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (ii) ConcatCBR: The standard method of concatenating input- output case examples with the input-text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (iii) GTM: Mapping Dnew using SmBoP into a non-parametric memory for aug- menting model’s predictions with inference-time memory look-ups similar to Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2020, 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We dis- cussed ConcatCBR and GTM baselines in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' All the baselines are implemented using SmBoP as the base model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 we also present ConcatCBR implemented on a T5-based Seq2Seq model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Implementation details: We implemented StructCBR and baselines using AllenNLP (Gardner et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2017) and Trans- formers (Wolf et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) libraries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We utilize the authors’ implementation of SmBoP (Rubin and Berant 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Due to limited computing resources, we primarily experiment with the ROBERTA-BASE checkpoint for initializing the text en- coder, followed by four RAT layers (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) to encode the schema structure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' All other hyper-parameters are the set to their default values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The SmBoP model is trained on Dtrain for 60K steps with a batch size of 80, using the default learning rate (LR) of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='86×10−4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The GTM baseline utilizes the output of this model for memory look-ups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For ConcatCBR baseline we train the SmBoP model further for 60K steps with a LR of 5×10−5, while concatenating the retrieved cases in the encoder’s input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' StructCBR introduces 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='53% additional parameters (φ) over the SmBoP parame- ters (θ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We train the parameters φ on Dtrain using a batch size of 64 for 60K steps with the default LR of 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='86×10−4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Additional training details are provided in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Overall Results: In Table 2, we compare StructCBR with different inference-time methods for adapting SmBoP based models on five different evaluation schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The perfor- mance of the unadapted SmBoP model varies from 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 EM to 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 EM across the five schemas indicating that the evalu- ation schemas are of varying difficulty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We find that Struct- CBR almost consistently offers substantial gains over the base SmBoP model w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' all the metrics, and across all the five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' StructCBR gains upto 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 EM points and on average 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 EM points over the base SmBoP model, while achieving almost 4 times higher gains than best existing method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In contrast, the ConcatCBR method, which has been shown to work well for other semantic parsing tasks (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pasupat, Zhang, and Guu 2021), provides posi- tive EM gains for only three out of five schemas, and causes overall drop in micro-averaged EM over all the test instances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We also explored a ConcatCBR implementation based on T5- large model with a comparable base accuracy (Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Even compared to this method, we continue to observe higher and more consistent gains from StructCBR on SmBoP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' GTM, another inference-time adaptation baseline, utilizes memory lookups similar to Khandelwal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2021), and offers al- most consistent but much smaller gains in comparison to StructCBR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The GTM baseline performs memory look-ups based on representations learned by the base SmBoP model whereas StructCBR is explicitly trained to perform sub-tree EM EX BEM BEX snew = world 1 , |Dtest| = 89 SmBOP 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 ConcatCBR 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 GTM +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 StructCBR (Ours) +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 snew = car 1 , |Dtest| = 60 SmBOP 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 ConcatCBR +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 GTM +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 StructCBR (Ours) +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 snew = cre Doc Template Mgt , |Dtest| = 53 SmBOP 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 ConcatCBR +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 GTM +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 StructCBR (Ours) +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 snew = dog kennels , |Dtest| = 49 SmBOP 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 ConcatCBR +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 GTM +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 StructCBR (Ours) +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 snew = flight 2 , |Dtest| = 49 SmBOP 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 ConcatCBR 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 GTM +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 StructCBR (Ours) +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 Micro-Average , |Dtest| = 300 SmBOP 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 ConcatCBR 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 GTM +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 StructCBR (Ours) +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 Table 2: Comparison of StructCBR with prior inference time adaptation methods (ConcatCBR and GTM) on 5 different schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP row provides the performance of the un- adapted model, while other rows report gains over SmBoP after adaptation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Micro-average refers to numbers averaged over all the test instances spanning across the five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' |Dtest| refers to size of the test-set, and snew provides the schema name.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' level memory look-ups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In particular, StructCBR boosts the top-K EM score (BEM) by up to 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 points.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' With a large- sized SmBoP architecture we continue to observe consistent gains for most of the schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Section A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 and Table A3 in Appendix provide results for adapting the large-sized models in the same setting as Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Table A4 in Appendix provides some anecdotal examples that were originally mispredicted by the base SmBoP model, but were fixed by StructCBR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Impact of number of cases: We run another set of ex- periments by reducing the number of cases from 30 to 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Figure 3 shows that gains from both GTM and StructCBR are smaller with fewer cases, as expected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Interestingly, Struct- 1 0 1 2 3 4 5 ConcatCBR GTM StructCBR |Cases|=20 |Cases|=30 Figure 3: Impact of case size on gains in EM (y-axis) of adapted models (x-axis).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' With 20 cases, StructCBR still out- performs GTM with 30 case examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' CBR with 20 cases still outperforms GTM with 30 case examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For ConcatCBR we do not observe significant changes because it additionally augments the case memory with examples retrieved from Spider’s train set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Not augment- ing cases via retrieval resulted in even worse performance of ConcatCBR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Justification for tree similarity: In Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3, we argued that directly computing tree similarities (simφ) using Equa- tion (3) was inefficient, and required pruning to be practical.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Instead in StructCBR we compute tree similarity (� simφ as per Equation 4) more efficiently as a composition of simi- larity of its children, and does not require pruning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Table 3 justifies our design by comparing results of scoring a pruned frontier containing top-5K trees using simφ, with scoring the entire frontier using � simφ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Efficiently scoring the entire frontier provides us better results on 4 out 5 schemas and a micro-averaged gain of 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 points in EM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' world 1 car 1 cre dog flights 2 Micro Doc kenn avg Whole-tree 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 Ours 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 Table 3: Scoring pruned frontier as per similarity of whole trees as in Equation (3) vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' our scoring of the entire frontier as per similarity composed from subtrees, as in Equation (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Better performance on four out of five schemas justifies our design of tree similarity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Comparison with fine-tuning: Adapting transformer models via finetuning often provides significant accuracy improvements, thanks to recent advances in optimization of transformer models (Huang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Xu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' However, finetuning is not viable when a model needs to be adapted instantaneously from just a few data points, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Adaptation Method Time(s)↓ EM%↑ SmBOP (Unadapted) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 StructCBR 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 Finetuning (1 epochs) 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 Finetuning (2 epochs) 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 47.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 Finetuning (5 epochs) 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 Finetuning (10 epochs) 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 Finetuning (20 epochs) 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 Finetuning (100 epochs) 500.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 Table 4: Comparing adaptation time and EM accuracy of StructCBR and finetuning for different number of epochs on 30 cases of world 1 schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We report wall clock times in seconds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' All the numbers were averaged over 3 runs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Finetun- ing took atleast 500x more time (10 epochs) to achieve EM gains that are almost instantly (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1s) achieved by StructCBR quickly incorporating expert feedback in the form of a few examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Table 4, we show that StructCBR serves the purpose of instantaneously improving model performance (+2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 pts EM), while being roughly 50× faster than finetun- ing for a single epoch, and 5000× faster than finetuning for 100 epochs of 30 case examples from world 1 schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Fine- tuning required atleast 10 epochs to outperform StructCBR’s accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Each epoch involved four parameter update steps of batch-size 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We note that applying StructCBR over SmBoP incurs a small increase (∼1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2×) in inference time per exam- ple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Overall, we observe that on three out of five schemas StructCBR instantly offers more than 50% of gains achieved by finetuning for 100 epochs, and 43% of gains on average across all schemas (Table A2 in appendix).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' This establishes StructCBR as a method for instantly utilizing available case examples for fast model adaptation until the next cycle of finetuning becomes affordable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 6 Conclusion and Future Work We presented StructCBR, a method for instant adaptation of Text-to-SQL models without finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We show that utilzing case examples in a more structured way via sub-tree level look-ups offers better performance in comparison to the standard method of concatenating case examples with input text into a Seq2Seq encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We find that explicitly learning to perform memory look-ups provides larger gains in com- parison to look-ups using a pre-trained model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Finally, we show that StructCBR enables much faster model adaptation in comparison to finetuning, potentially allowing instanta- neous adaptation to expert feedback provided in form of a few examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We propose StructCBR as a faster alternative to model adaptation, until the next finetuning cycle is deemed feasible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Despite its speed, there remains an accuracy gap between StructCBR and sufficient finetuning, which might be narrowed by more sophisticated similarity networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Our ex- ploration of StructCBR focused only on the Text-to-SQL task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In future we wish to explore the effectiveness of StructCBR for other semantic parsing tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' References Atzeni, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Dhuliawala, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Murugesan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and SACHAN, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Case-based Reasoning for Better Generalization in Text-Adventure Games.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Brown, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Mann, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ryder, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Subbiah, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Kaplan, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Dhariwal, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Neelakantan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shyam, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Sastry, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Askell, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Language models are few-shot learners.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Advances in neural information processing systems, 33: 1877– 1901.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tworek, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Jun, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yuan, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' de Oliveira Pinto, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Kaplan, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Edwards, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Burda, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Joseph, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Brock- man, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ray, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Puri, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Krueger, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Petrov, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Khlaaf, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Sastry, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Mishkin, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gray, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ryder, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pavlov, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Power, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Kaiser, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Bavarian, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Winter, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tillet, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Such, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Cummings, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Plappert, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chantzis, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Barnes, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Herbert-Voss, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Guss, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Nichol, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Paino, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tezak, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Babuschkin, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Balaji, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Jain, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Saunders, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hesse, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Carr, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Leike, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Achiam, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Misra, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Morikawa, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Radford, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Knight, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Brundage, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Murati, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Mayer, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Welinder, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Mc- Grew, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Amodei, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' McCandlish, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Sutskever, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Zaremba, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Evaluating Large Language Models Trained on Code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Das, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Godbole, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Monath, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zaheer, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and McCal- lum, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Probabilistic Case-based Reasoning for Open- World Knowledge Graph Completion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Findings of the Association for Computational Linguistics: EMNLP 2020, 4752–4765.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Das, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zaheer, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Thai, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Godbole, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Perez, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lee, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tan, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Polymenakos, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and McCallum, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Case-based Reasoning for Natural Language Queries over Knowledge Bases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' arXiv preprint arXiv:2104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='08762.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gardner, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Grus, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Neumann, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tafjord, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Dasigi, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Liu, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Peters, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Schmitz, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' AllenNLP: A Deep Semantic Natural Language Processing Platform.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Guo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Xiao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Liu, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Zhang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Towards Complex Text-to-SQL in Cross-Domain Database with Intermediate Representation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 4524–4535.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Florence, Italy: Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gupta, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shrivastava, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Sagar, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Aghajanyan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Savenkov, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' RETRONLU: Retrieval Aug- mented Task-Oriented Semantic Parsing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' arXiv preprint arXiv:2109.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='10410.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hashimoto, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Guu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Oren, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Liang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A retrieve-and-edit framework for predicting structured outputs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hazoom, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Malik, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Bogin, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Text-to-SQL in the Wild: A Naturally-Occurring Dataset Based on Stack Ex- change Data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 1st Workshop on Natural Language Processing for Programming (NLP4Prog 2021), 77–87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Online: Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hossain, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ghazvininejad, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Simple and effective retrieve-edit-rerank text generation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 2532–2538.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Huang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Perez, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ba, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Volkovs, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Improv- ing Transformer Optimization Through Better Initialization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In III, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=', eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=', Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, 4475–4483.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Khandelwal, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Fan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Jurafsky, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Lewis, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Nearest Neighbor Machine Translation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Khandelwal, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Levy, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Jurafsky, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Lewis, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Generalization through Memorization: Near- est Neighbor Language Models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Le Scao, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Rush, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' How many data points is a prompt worth?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2021 Conference of the North American Chapter of the Association for Compu- tational Linguistics: Human Language Technologies, 2627– 2636.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lee, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Polozov, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Richardson, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Kag- gleDBQA: Realistic Evaluation of Text-to-SQL Parsers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), 2261–2273.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Online: Association for Com- putational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Liu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ott, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Goyal, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Du, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Joshi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Levy, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lewis, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Stoyanov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ro{BERT}a: A Robustly Optimized {BERT} Pretraining Approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Min, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lyu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Holtzman, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Artetxe, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lewis, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hajishirzi, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' arXiv preprint arXiv:2202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='12837.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pasupat, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Guu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Controllable Se- mantic Parsing via Retrieval Augmentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 7683–7698.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pawlik, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Augsten, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Efficient computation of the tree edit distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' ACM Transactions on Database Systems (TODS), 40(1): 1–40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pawlik, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Augsten, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tree edit distance: Robust and memory-efficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Information Systems, 56: 157–173.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Perez, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Kiela, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Cho, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' True few-shot learn- ing with language models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Poesia, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Polozov, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Le, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tiwari, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Soares, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Meek, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Gulwani, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Synchromesh: Reliable code generation from pre-trained language models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' arXiv preprint arXiv:2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='11227.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Popescu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Etzioni, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Kautz, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Towards a Theory of Natural Language Interfaces to Databases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Pro- ceedings of the 8th International Conference on Intelligent User Interfaces, 149–157.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Raffel, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shazeer, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Roberts, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Narang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Matena, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhou, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Li, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Liu, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Explor- ing the Limits of Transfer Learning with a Unified Text-to- Text Transformer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Journal of Machine Learning Research, 21(140): 1–67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Rubin, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Berant, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP: Semi-autoregressive Bottom-up Semantic Parsing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technolo- gies, 311–324.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Li, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Bahdanau, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' de Vries, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Pal, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' DuoRAT: Towards Simpler Text-to-SQL Models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 1313–1321.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Schucher, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Bahdanau, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' PICARD Parsing Incrementally for Constrained Auto-Regressive De- coding from Language Models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Pro- cessing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shaw, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Uszkoreit, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Vaswani, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Self-Attention with Relative Position Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2018 Conference of the North American Chapter of the As- sociation for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), 464–468.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' New Or- leans, Louisiana: Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shin, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Razeghi, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Logan IV, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wallace, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Singh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated Prompts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 4222–4235.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Socher, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Manning, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Ng, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Reasoning with neural tensor networks for knowledge base completion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Advances in neural information processing sys- tems, 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Suhr, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shaw, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Explor- ing unexplored generalization challenges for cross-database semantic parsing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 8372–8388.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Mooney, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Automated Construc- tion of Database Interfaces: Intergrating Statistical and Re- lational Learning for Semantic Parsing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In 2000 Joint SIG- DAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora, 133–141.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hong Kong, China: Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wang, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shin, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Polozov, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Richardson, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' RAT-SQL: Relation-Aware Schema Encoding and Linking for Text-to-SQL Parsers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 7567–7578.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wolf, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Debut, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Sanh, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Chaumond, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Delangue, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Moi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Cistac, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Rault, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Louf, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Funtowicz, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Davi- son, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shleifer, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' von Platen, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ma, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Jernite, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Plu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Xu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scao, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Gugger, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Drame, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lhoest, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Rush, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Transformers: State-of-the-Art Natural Language Processing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2020 Confer- ence on Empirical Methods in Natural Language Processing: System Demonstrations, 38–45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Online: Association for Com- putational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Xie, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Shi, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhong, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Scholak, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yasunaga, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhong, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yin, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' UnifiedSKG: Unifying and Multi-Tasking Structured Knowl- edge Grounding with Text-to-Text Language Models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' arXiv preprint arXiv:2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='05966.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Xu, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Kumar, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zi, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Huang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Cheung, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Prince, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Cao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Optimiz- ing Deeper Transformers on Small Datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 59th Annual Meeting of the Association for Compu- tational Linguistics and the 11th International Joint Con- ference on Natural Language Processing (Volume 1: Long Papers), 2089–2102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Online: Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yu, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Lin, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Tan, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Radev, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Xiong, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' GraPPa: Grammar-Augmented Pre-Training for Table Semantic Parsing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yu, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhang, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yang, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yasunaga, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Li, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Ma, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Li, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Yao, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Roman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Spider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-SQL Task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, 3911–3921.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zelle, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Mooney, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 1996.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Learning to Parse Database Queries using Inductive Logic Programming.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In AAAI/IAAI, 1050–1055.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Portland, OR: AAAI Press/MIT Press.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Zhao, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Wallace, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Feng, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Klein, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Singh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Calibrate Before Use: Improving Few-shot Performance of Language Models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In Meila, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' and Zhang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=', eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=', Pro- ceedings of the 38th International Conference on Machine Learning, volume 139 of Proceedings of Machine Learning Research, 12697–12706.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 Details of the retriever used with ConcatCBR The ConcatCBR baseline discussed in Section 4 first re- trieves Text-SQL pairs for a given utterance ¯x, which are then concatenated with the utterance ¯x as an in input to model’s encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Similar to Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2021);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Poesia et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2022), we train a RoBERTa-BASE based sentence embed- ding model E : X �→ Rd, that retrieves Text-SQL pairs {¯xr, ¯qr} from cases or training data having higher cosine similarty with the input utterance ¯x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The retriever is trained independent of Text-to-SQL models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For any two Text-SQL pairs (¯xi, ¯qi) and (¯xj, ¯qj) in the train set, we utilize normal- ized tree-edit-distance TED(¯qi, ¯qj) ∈ [0, 1] between queries (¯qi, ¯qj) to supervise the cosine-similarity scores between sen- tence embeddings CSim(E(¯xi), E(¯xj)), such that pairs with low tree-edit-distance have higher cosine-similarity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We uti- lize APTED library (Pawlik and Augsten 2015, 2016) to compute tree-edit-distance between the relational algebra trees of the SQL queries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We modify the cost function of tree-edit-distance to ignore the leaf values and constants in the tree so that structurally similar trees are assigned lower distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We normalize tree-edit-distance by size of the larger of two trees, so it lies in range [0, 1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' More concretely, for a given Text-SQL pair in a training batch (¯xi, ¯qi) we sample 15 more Text-SQL pairs {(¯xj, ¯qj)}15 j=1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Then we compute tree-edit-distances {TED(¯qi, ¯qj)}15 j=1, and cosine similarities {CSim(E(¯xi), E(¯xj))}15 j=1 between sen- tences embeddings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The cosine-similarity scores are now supervised using tree-edit-distances as per loss in Equation 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' wi,j = exp(1 − 2 TED(¯qi, ¯qj)) � j exp(1 − 2 TED(¯qi, ¯qj)) LE = − � i,j wi,j log exp(CSim(E(¯xi), E(¯xj))) � j exp(CSim(E(¯xi), E(¯xj))) (9) During inference on a new schema, we update the retrieval index with available cases from the new schema.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Now, given a text query ¯x, we retrieve the 5 most similar examples as per cosine-similarity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 ConcatCBR baseline using T5 For a fair comparison, all the baselines in Section 5 are built on the top of the SmBoP architecture that utilizes non-auto- regressive bottom-up decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' However, ConcatCBR ar- chitectures in prior works (Das et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Pasupat, Zhang, and Guu 2021) utilize the standard auto-regressive Seq2Seq architectures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Hence, to further ensure a fair comparison, we explored ConcatCBR baselines built on the top of T5- large (Raffel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We utilize the implemen- tation of UnifiedSKG library (Xie et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022) for T5-large based Text-to-SQL models 3 and modify it to concatenate input-output examples at T5’s encoder to get the ConcatCBR baseline.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The T5-large models were intialized with an LM- Adapted 4 version.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The default T5-large based Text-to-SQL 3https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='com/HKUNLP/UnifiedSKG/blob/main/configure/Salesforce/ T5 large finetune spider with cell value.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='cfg 4https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='com/google-research/text-to-text-transfer-transformer/blob/main/ released checkpoints.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='md#lm-adapted-t511lm100k world car cre dog flights avg 1 1 Doc kenn 2 EM T5 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 83 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 66 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +ConcatCBR +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 +10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 EX T5 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 61 +ConcatCBR +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 Table A1: EM and EX performance of a T5-large based Text-to-SQL model and gains from its ConcatCBR extension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Similar to Table 2, we observe that ConcatCBR does not provide consistent gains over the unadapted model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For two schemas the gains are negative, and gains micro-averaged (avg) across five schemas are small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' model finetuned on Spider dataset has a comparable EM ac- curacy of 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP’s 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 micro-averaged across five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Table A1 presents the EM and EX accuracies of the default T5-large based Text-to-SQL model, and gains obtained by its ConcatCBR extension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Similar to Table 2, gains from Con- catCBR are not consistent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For two out of five databases, we observe a drop in EM and EX.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The micro-averaged gains across all five schemas are also small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Thus, our observations remain largely consistent with ConcatCBR baseline based on SmBoP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 StructCBR gains relative to finetuning Table A2 shows gains of StructCBR over SmBoP rela- tive to gains of finetuning over SmBoP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Relative Gain = +StructCBR +Finetuning × 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' StructCBR instantly obtains 20% to 83% of EM gains achieved by finetuning for 100 epochs across different schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For 3 out of 5 schemas relative gains are more than 50%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Micro-averaged across all examples, Struct- CBR achieves 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8% of gains relative to finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' It is important to note here that we do not position StructCBR as an alternative to finetuning, but instead as a method of uti- lizing available cases for instantaneously improving model’s performance until the next cycle of finetuning becomes af- fordable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' world car cre dog flights avg 1 1 Doc kenn 2 SmBOP 46 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +StructCBR +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +Finetuning +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 +27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 Relative Gain (%) 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 Table A2: StructCBR instantly obtains 20% to 83% EM gains achieved by finetuning for 100 epochs across different schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' On 3 out 5 schemas gains are higher than 50% of gains from finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Averaged (avg) across all examples, gains are 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8% of finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 Additional hyperparameter and training details The baseline SmBoP model is based on RoBERTa-base ar- chitecture and contains four RAT layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The total number of trainable parameters in this more is roughly 133M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Adding StructCBR module on top of SmBoP introduces only 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='53% additional parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In particular, the transformer model TXφ used for computing Gφ representations in Equation 2 is composed of two transformer blocks of hidden size=256, attention heads=8, feedforward dim=1024.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' All the experi- ments were performed on a NVIDIA RTX 3060 GPU, and a NVIDIA RTX A6000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Training times for various Text-to- SQL models usually varied from 12 hours to 16 hours.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 Results with a larger model size Constrained by limited computing resources for training large models, we conducted all the experiments and abla- tions in Section 5 using SmBoP models initialized with a pretrained RoBERTa-BASE checkpoint (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) fol- lowed by 4 RAT layers (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In this section, we validate our key results on a larger model size, by repeat- ing the experiment reported in Table 2 of Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' More specifically, we replace RoBERTa-BASE in SmBoP with a pretrained RoBERTa-LARGE checkpoint based on grammar augmented pre-training (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) as used by Rubin and Berant (2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The number of RAT layers is also in- creased from 4 to 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We refer to the larger SmBoP architec- ture as SmBoP-LARGE (367M parameters) and the base-sized SmBoP model used in Section 5 as SmBoP-BASE (133M parameters).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Similar to Table 2, Table A3 compares differ- ent inference-time methods for adapting the SmBoP-LARGE model to new schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' The performance of SmBoP-BASE is reported just as a reference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For each adaptation method (ConcatCBR, GTM, StructCBR), we report it’s gains over the SmBoP-LARGE model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We consistently observe positive gains from StructCBR across all the metrics on four out of five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' In contrast, the gains from prior inference time adaptation methods are not consistently positive across dif- ferent metrics and schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' It is also interesting to note that the gains in top-K accuracy metrics (BEM and BEX) from StructCBR are significantly higher than prior methods and consistently positive across all the schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Better perfor- mance in top-K metrics indicates that boosting candidate subtree scores via StructCBR improves the recall of correct subtrees during beam decoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 Additional Related Work Incontext Learning with LLMs: Learning new tasks via carefully designed exemplar prompts (Shin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Le Scao and Rush 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Perez, Kiela, and Cho 2021) for large language models (LLMs) like GPT-3 (Brown et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2020) and Codex (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021) has shown promise for many tasks (Zhao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Min et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' 2022) without requir- ing to finetune model parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' For Text-to-SQL, Poesia et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' (2022) retrieve and encode input-specific prompts in a way similar to the ConcatCBR method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' They differ from ConcatCBR as they do not explicitly train the LLM to per- form CBR with the extracted propmts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' While their method achieves impressive performance by using LLMs, training significantly (100x) smaller Text-to-SQL models on task- specific datasets like Spider outperforms their LLMs by con- siderable margins.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' EM EX BEM BEX snew = world 1 , |Dtest| = 89 SmBoP-BASE 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 SmBoP-LARGE 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 ConcatCBR 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 GTM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 StructCBR (Ours) 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 snew = car 1 , |Dtest| = 60 SmBoP-BASE 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 SmBoP-LARGE 53.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 ConcatCBR +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 GTM +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 StructCBR (Ours) +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 snew = cre Doc Template Mgt , |Dtest| = 53 SmBoP-BASE 84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 SmBoP-LARGE 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 ConcatCBR +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 GTM 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 StructCBR (Ours) +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 snew = dog kennels , |Dtest| = 49 SmBoP-BASE 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 SmBoP-LARGE 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 ConcatCBR +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 GTM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 StructCBR (Ours) +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 snew = flight 2 , |Dtest| = 49 SmBoP-BASE 55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 SmBoP-LARGE 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 ConcatCBR +6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 GTM +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 +7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='5 StructCBR (Ours) +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='4 +14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 Micro-Average , |Dtest| = 300 SmBoP-BASE 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 SmBoP-LARGE 62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='6 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 ConcatCBR +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='0 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='8 GTM +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='2 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 StructCBR (Ours) +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='1 +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='9 +5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='3 +4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 Table A3: Results using SmBoP-LARGE: We compare Struct- CBR with prior inference time adaptation methods (Con- catCBR and GTM) for adapting SmBoP-LARGE on 5 differ- ent schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP-BASE and SmBoP-LARGE rows pro- vide the performance of unadapted model SmBoP models, while other rows report gains over SmBoP-LARGE after adap- tation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' SmBoP-BASE numbers are just of reference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Micro- average refers to numbers averaged over all the test instances spanning across the five schemas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' |Dtest| refers to size of the test-set, and snew provides the schema name.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='7 Anecdotes We include some anecdotal examples where StructCBR fixes the mistakes made by SmBoP by utilizing case examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Text What is the code of airport that has the highest number of flights?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Incorrect SQL SELECT flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport FROM flights GROUP BY flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport ORDER BY SUM( flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='flightno ) DESC LIMIT 1 Corrected SQL SELECT airports.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airportcode FROM airports JOIN flights ON airports.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airportcode = flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport GROUP BY flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport ORDER BY COUNT( * ) DESC LIMIT 1 Case Text Give the code of the airport with the least flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Case SQL SELECT airports.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airportcode FROM airports JOIN flights ON airports.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airportcode = flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport GROUP BY flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport ORDER BY COUNT( * ) ASC LIMIT 1 Mistake(s) Fixed Correct column selection, Add missing join, Replace SUM by COUNT Text Which airlines have a flight with source airport AHD?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Incorrect SQL SELECT flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airline FROM airlines JOIN flights ON airlines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='uid = flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airline JOIN airports ON flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport = airports.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airportcode WHERE airports.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airportname = ’AHD’ Corrected SQL SELECT airlines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airline FROM airlines JOIN flights ON airlines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='uid = flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airline WHERE flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='sourceairport = ’AHD’ Case Text What are airlines that have flights arriving at airport ’AHD’?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Case SQL SELECT airlines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airline FROM airlines JOIN flights ON airlines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='uid = flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='airline WHERE flights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='destairport = ’AHD’ Mistake(s) Fixed Drop additional condition on Join Text What are the population and life expectancies in Brazil?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Incorrect SQL SELECT country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='population , country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='lifeexpectancy FROM city JOIN country ON city.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='countrycode = country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='code WHERE country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='name = ’Brazil’ Corrected SQL SELECT country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='lifeexpectancy , country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='population FROM country WHERE country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='name = ’Brazil’ Case Text What are the region and population of Angola?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Case SQL SELECT Population , Region FROM country WHERE Name = ’Angola’ Mistake(s) Fixed Drop the join condition Text Return the codes of countries that do not speak English and do not have Republics for governments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Incorrect SQL SELECT country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='code FROM country WHERE countrylanguage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='language = ’English’ INTERSECT SELECT countrylanguage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='countrycode FROM country WHERE country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='governmentform !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='= ’Republic’ Corrected SQL SELECT country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='code FROM country WHERE country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='governmentform !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='= ’Republic’ EXCEPT SELECT countrylanguage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='countrycode FROM countrylanguage WHERE countrylanguage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='language = ’English’ Case Text What are the codes of the countries that do not speak English and whose government forms are not Republic?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' Case SQL SELECT country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='code FROM country WHERE country.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='governmentform !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='= ’Republic’ EXCEPT SELECT countrylanguage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='countrycode FROM countrylanguage WHERE countrylanguage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content='language = ’English’ Mistake(s) Fixed Set operation and Equality Table A4: Anecdotal examples where mistakes in output SQLs generated by SmBoP are fixed with help of StructCBR through related examples in case memory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
+page_content=' We show only one of the related examples from cases for brevity' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/edE2T4oBgHgl3EQfxQji/content/2301.04110v1.pdf'}
diff --git a/f9FMT4oBgHgl3EQf2TEy/content/tmp_files/2301.12443v1.pdf.txt b/f9FMT4oBgHgl3EQf2TEy/content/tmp_files/2301.12443v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a783aa59162a76be778db48cf2fd67ee8f52469a
--- /dev/null
+++ b/f9FMT4oBgHgl3EQf2TEy/content/tmp_files/2301.12443v1.pdf.txt
@@ -0,0 +1,1276 @@
+Pipe-BD: Pipelined Parallel Blockwise Distillation
+Hongsun Jang†, Jaewon Jung§, Jaeyong Song‡, Joonsang Yu¶, Youngsok Kim‡, and Jinho Lee†∗
+†Department of Electrical and Computer Engineering, Seoul National University
+§Department of Artificial Intelligence, Yonsei University
+‡Department of Computer Science, Yonsei University
+¶CLOVA ImageVision, CLOVA AI Lab, NAVER
+hongsun.jang@snu.ac.kr, {jungjaewon, jaeyong.song}@yonsei.ac.kr, joonsang.yu@navercorp.com,
+youngsok@yonsei.ac.kr, leejinho@snu.ac.kr
+Abstract—Training large deep neural network models is highly
+challenging due to their tremendous computational and mem-
+ory requirements. Blockwise distillation provides one promising
+method towards faster convergence by splitting a large model into
+multiple smaller models. In state-of-the-art blockwise distillation
+methods, training is performed block-by-block in a data-parallel
+manner using multiple GPUs. To produce inputs for the student
+blocks, the teacher model is executed from the beginning until
+the current block under training. However, this results in a high
+overhead of redundant teacher execution, low GPU utilization,
+and extra data loading. To address these problems, we propose
+Pipe-BD, a novel parallelization method for blockwise distillation.
+Pipe-BD aggressively utilizes pipeline parallelism for blockwise
+distillation, eliminating redundant teacher block execution and
+increasing per-device batch size for better resource utilization. We
+also extend to hybrid parallelism for efficient workload balancing.
+As a result, Pipe-BD achieves significant acceleration without
+modifying the mathematical formulation of blockwise distillation.
+We implement Pipe-BD on PyTorch, and experiments reveal that
+Pipe-BD is effective on multiple scenarios, models, and datasets.
+Index Terms—Distributed Training, Knowledge Distillation,
+Neural Architecture Search, Model Compression
+I. INTRODUCTION
+Modern deep neural network models are known to incur huge
+computational and memory requirements, especially with large-
+scale datasets [1]. With the continuing growth in model size,
+it takes tens, if not hundreds, of GPU days to train them [2],
+and the model size often exceeds the GPU memory capacity.
+Especially for methods that explore large solution spaces such
+as the neural architecture search (NAS) [3, 4], the problem
+becomes even more significant. This problem mandates the
+use of model parallelism [5, 6], which creates substantial
+throughput loss with inevitable pipeline bubbles.
+Blockwise distillation [7, 8, 9] is one promising approach
+to mitigate such problems. As illustrated in Fig. 1, blockwise
+distillation splits the model into multiple smaller blocks. As
+opposed to traditional knowledge distillation methods that rely
+on input data and output labels from both ends, blockwise
+distillation uses the intermediate activation values of pretrained
+blocks of a ‘teacher’ to train each ‘student’ block. As a result,
+each block converges faster (i.e., fewer epochs) due to the
+smaller solution space.
+* Corresponding author
+Input
+T0
+S0
+T1
+S1
+T2
+S2
+T3
+S3
+𝑳(∆𝒐𝒖𝒕𝒑𝒖𝒕)
+Teachers
+Block 0
+Block 1
+Students
+Block 2
+Block 3
+Fig. 1: Conceptual diagram of blockwise distillation.
+Contrary to the earlier belief that teachers must be larger
+than students, recent studies have revealed that smaller teachers
+can be used to train larger students [10]. With such findings,
+blockwise distillation is used in various fields such as model
+compression [7, 11] and NAS [9, 12]. Since training a small
+teacher for a new task is quick and easy, blockwise distillation
+can be applied in most cases where traditional training is used.
+However, the existing state-of-the-art methods for blockwise
+distillation [7, 9] exhibit several inefficiencies. Relying on the
+traditional data-parallel training scheme, they train each student
+block one by one independently. While this fully exploits the
+independent nature of the blocks, it is not the best choice for
+training throughput. First, to train a single intermediate student
+block, the teacher blocks must be executed from the beginning
+to the designated block. As a result, the teacher blocks exhibit
+substantial redundant execution, especially with blocks closer
+to the output. Second, with data parallelism, a batch of data is
+split among multiple GPUs, which leads to a smaller batch size
+per GPU, often resulting in resource under-utilization. Some
+approaches use a larger batch size to mitigate this [2], but it is
+known to be difficult to ensure model convergence [13]. Last,
+the data must be redundantly loaded for each student block.
+Unless the entire dataset fits into the GPU memory, the data
+are loaded from the CPU memory or disks. As the memory and
+disks are shared system-wide, the extra data loading becomes
+another significant overhead in training.
+To address the issues, we propose Pipe-BD, a novel par-
+allel training method for blockwise distillation. We assign
+individual student blocks to different devices and compute
+a teacher network in a relayed manner, which can reduce
+teacher redundancy. Inspired by approaches with pipeline paral-
+lelism [5, 14, 15, 16], we restructure the training schedule of the
+student blocks such that the training time is greatly improved.
+arXiv:2301.12443v1 [cs.LG] 29 Jan 2023
+
+Pipe-BD comprises three components: First, we propose
+teacher relaying. Instead of relying on data parallelism, we
+spread the student model to multiple training devices (i.e.,
+GPUs) in a block granularity. Then, blocks of the teacher
+model are executed by relaying the intermediate activation
+values between the devices. This approach has the advantages
+of eliminating extra data loading and increasing resource uti-
+lization from larger batch size per device. Second, we propose
+decoupled parameter update to remove the scheduling bubbles
+and enhance the overall utilization. With teacher relaying,
+devices have to wait for the intermediate activation values
+from previous devices, creating scheduling bubbles. Decoupled
+parameter update performs model parameter updates in a mis-
+aligned manner and starts the next step right ahead, so those
+bubbles can be removed. Third, we suggest automatic hybrid
+distribution. Achieving a balance between devices is difficult
+with blockwise distillation because of the limited number of
+blocks available in typical neural network structures. Automatic
+hybrid distribution enables fine-grained balancing with further
+splitting blocks along the batch size dimensions.
+Pipe-BD is implemented on PyTorch and can automatically
+make all scheduling decisions to improve the throughput. Our
+extensive set of experiments shows Pipe-BD achieves a signifi-
+cant speedup over the state-of-the-art methods on multiple use
+cases and environments ranging from 2.37× to 7.38×.
+II. BACKGROUND AND RELATED WORK
+A. Blockwise Distillation
+Blockwise distillation [7, 8, 9] is a promising direction for
+training a neural network. In traditional knowledge distillation,
+a student model is trained against a pre-trained teacher model.
+Because the solution space size is identical to that of con-
+ventional supervised training, it faces convergence and training
+time problems. Blockwise distillation splits the larger teacher
+model into smaller ones and trains them blockwise as depicted
+in Fig. 1. Each teacher block (Ti) and student block (Si)
+pair obtains activation values from the previous teacher block
+(Ti−1). This pair performs forward pass using the activation as
+input and creates teacher output activation and student output
+activation. Blockwise distillation minimizes a loss function
+(L(∆output)) which measures the difference between these
+two activations, to distill knowledge from a teacher block to
+the dedicated student block. This blockwise distillation process
+makes target problem spaces smaller and is known to converge
+faster. Many applications such as NAS [9, 12] and model
+compression [7, 11] use blockwise distillation because of these
+characteristics.
+B. Parallelization Baseline of Blockwise Distillation
+State-of-the-art methods of blockwise distillation [9] use
+the traditional data-parallel scheme to further accelerate the
+training as illustrated in Fig. 3a. This scheme trains a student
+block (Si) with all devices in a data-parallel manner for fixed
+n epochs, then moves on to train the next student block
+(Si+1). It redundantly loads data multiple times because of this
+iterative training. Each student block (Si) requires the activation
+0
+10
+20
+30
+Baseline
+Ideal
+Time/epoch (sec)
+Data loading
+T exec.
+S exec.
+Idle
+Pipe-BD (Rank 0-3)
+Fig. 2: Motivational experiment. The breakdown demonstrates
+three major inefficiencies of baseline; redundant teacher execu-
+tion, extra data loading, and low resource utilization.
+values from the previous teacher (Ti−1), so it also entails
+redundant teacher executions. Furthermore, it uses a smaller
+batch size per device which leads to under-utilization. Due
+to these inefficiencies, the data-parallel blockwise distillation
+suffers from poor scalability. An alternative scheme [7] regards
+the training of each layer as a single task and adopts bin
+packing algorithm to balance the workload. However, it still
+has redundant teacher executions and suffers from workload
+imbalance when there are insufficient layers in the model.
+III. MOTIVATION
+In this section, we provide a motivational study highlighting
+the inefficiency of the existing parallel blockwise distillation
+training scheme and the need for a new approach. Fig. 2
+depicts the breakdown of time spent in parallel blockwise
+distillation with four RTX A6000 GPUs (NAS with Cifar-10;
+see Section VI-B for the detailed setup). ‘Baseline’ refers to the
+state-of-the-art parallel blockwise distillation method [9], where
+each block is trained sequentially using four devices with data
+parallelism. As displayed in the chart, the training time is spent
+on data loading, teacher execution (forward pass), and student
+execution (forward/backward pass).
+However, all three parts exhibit significant inefficiency, slow-
+ing down the training. To demonstrate the inefficiencies, we
+plot the ‘ideal’ bar in Fig. 2 by measuring the training time
+of each part separately with a single GPU and dividing each
+time by four. This represents an imaginary system with perfect
+parallelization and infinite device memory.
+The large gaps in teacher execution and data loading time
+occur because the baseline has many redundant teacher exe-
+cutions and extra data loading. Because each student block to
+train requires executing the teacher model from the beginning,
+the earlier teacher blocks are redundantly executed multiple
+times (see Fig. 3a). Similarly, block-by-block training forces
+loading data as many as the number of blocks. In addition, data-
+parallelism leads to smaller batch size per device, resulting in
+lower resource utilization. As demonstrated in several empirical
+studies [17, 18], a sufficient per-device batch size is critical
+for training throughput, which is the cause of the gap on
+student execution time. Pipe-BD targets these inefficiencies. As
+presented in Fig. 2, Pipe-BD reduces the training time close to
+the ideal case, with only a small overhead (idle).
+
+UP
+T0
+T0
+T0
+T0
+T2
+T2
+T2
+T2
+T1
+T1
+T1
+T1
+S2
+S2
+S2
+S2
+DP
+DL
+Batch = 64 * 4 = 256
+...
+UP
+T0
+T0
+T0
+T0
+S0
+S0
+S0
+S0
+DP
+DL
+...
+n epoch
+n epoch
+n epoch
+UP
+T0
+T0
+T0
+T0
+T1
+T1
+T1
+T1
+S1
+S1
+S1
+S1
+DP
+DL
+...
+... n epoch
+: i-th teacher exec.
+: Parameter update
+Ti
+Si
+: i-th student exec.
+DP
+: Data-parllel gradient sharing
+UP | U
+DL | D
+: Data loading
+: Step boundary
+Batch = 256
+Redundant
+T0
+T0
+T0
+T0
+T2
+T2
+T2
+T2
+T1
+T1
+T1
+T1
+DL
+T3
+T3
+T3
+T3
+S3
+S3
+S3
+S3
+UP
+DP
+(a) Baseline
+UP
+T0
+T1
+T2
+T3
+S0
+S1
+S2
+S3
+UP
+T0
+T1
+T2
+T3
+S0
+S1
+S2
+S3
+UP
+T0
+T1
+T2
+T3
+S0
+S1
+S2
+S3
+D
+Batch = 256
+...
+Redundant
+D
+D
+(b) w/ Teacher Relaying
+T0
+T1
+T2
+S0
+S1
+S2
+U
+U
+U
+U
+T0
+T1
+T2
+T3
+S0
+S1
+S2
+S3
+U
+U
+U
+U
+T0
+T1
+T2
+T3
+S0
+S1
+S2
+S3
+Batch = 256
+...
+U
+U
+U
+U
+T0
+T1
+T2
+T3
+S0
+S1
+S2
+S3
+D
+D
+D
+D
+(c) w/ Decoupled Parameter Update
+DL
+DL
+DL
+T0
+T0
+U
+UP
+S0
+T1
+T2
+T3
+S2
+S3
+S1
+S0
+DP
+T0
+T0
+U
+U
+UP
+S0
+T1
+T2
+T3
+S2
+S3
+S1
+S0
+DP
+T0
+T0
+U
+U
+UP
+S0
+T1
+T2
+T3
+S2
+S3
+S1
+S0
+DP
+U
+U
+UP
+T0
+S0
+T0
+T1
+T2
+T3
+S2
+S3
+S1
+S0
+DP
+...
+Batch = 128 * 2 = 256
+DL
+(d) w/ Automatic Hybrid Distribution
+Fig. 3: Illustration of the techniques in Pipe-BD.
+IV. PIPE-BD METHOD
+A. Teacher Relaying
+Pipe-BD starts by restructuring the training pipeline of
+blockwise distillation with teacher relaying. As opposed to the
+baseline (Fig. 3a) where a single block is fully trained in a
+data-parallel manner before moving on to the next, teacher
+relaying exclusively distributes the teacher and student blocks
+to all training devices. Then, each device relays the intermediate
+teacher activation values to the next device as depicted in
+Fig. 3b. The received activation is the input for both the teacher
+and the student block. The teacher block is executed first,
+whose output activation is sent to the next device such that
+the execution of the next block can start. Overlapped with the
+transmission, the forward pass execution of the student starts,
+taking the same input as the teacher block. After calculating the
+loss by comparing the output activations of the teacher and the
+student, the backward pass of the student follows. After all the
+backward passes are finished, parameter updates are performed
+on each block, completing the training step.
+The teacher relaying scheme has two advantages over the
+existing approach. First, each device executes the stages with
+larger batches and enjoys better resource utilization. For exam-
+ple, in the baseline using four devices with an effective batch
+size of 256, each device executes with a batch size of 64, which
+is often too small to fully utilize the hardware resources. In
+contrast, with teacher relaying, each device would run with a
+full batch size of 256, increasing resource utilization. Second,
+the overhead of data loading is reduced. When the dataset is
+large, the data must come from the main memory or the disk,
+where both are system-wide shared resources. Because teacher
+relaying does not go through multiple training passes, the
+number of data loading decreases, leading to higher throughput.
+One minor trade-off is communication overhead. In the
+baseline, gradient sharing must occur after every backward
+pass. With teacher relaying, there is some communication
+delay from relaying the intermediate activation values from
+one device to another. However, the communication time is
+almost negligible in our target settings of single-node multi-
+device training. Furthermore, in both cases, most of the com-
+munications overlapped with computations.
+B. Decoupled Parameter Update
+Although teacher relaying removes the redundant teacher
+executions, the removed redundancy is not directly translated to
+speedup. At the beginning of each step, each device has to wait
+until the previous device delivers the intermediate activation.
+Fig. 3c illustrates how decoupled parameter update addresses
+this problem. As soon as the backward pass of each block is
+complete, the parameter updates are performed without waiting
+for the other devices. Then, the teacher execution of the next
+step can start earlier, increasing the training throughput. This
+does not harm the training accuracy by any means because the
+student blocks have no dependency on the weight parameters of
+the other blocks, which is a special characteristic of blockwise
+knowledge distillation training.
+Decoupled parameter update successfully hides the teacher
+waiting time except for the beginning of each epoch, where
+full synchronization is needed for validating the whole model.
+Because there are usually tens to hundreds of steps per epoch,
+such overhead is amortized to a negligible amount.
+C. Automatic Hybrid Distribution
+With the teacher relaying and decoupled parameter update,
+the system throughput is determined by the throughput of the
+slowest device. Because of this, load balancing between devices
+plays a critical role in performance. One straightforward and
+intuitive load-balancing method is distributing the workload in
+contiguous blocks. The distribution is simple because there
+are only
+B−1CN−1 choices for B blocks and N devices.
+Unfortunately, the naive distribution scheme often fails to
+provide a good balance. In blockwise distillation, the number
+of blocks B is determined by the neural network architecture.
+Usually, B is around ten [3, 19] and N is four to eight within
+a single server. Because there are not enough number of blocks
+to distribute to the devices, the naive distribution is likely to
+end up in a severe workload imbalance.
+With automatic hybrid distribution, we provide another de-
+gree of freedom for workload distribution as presented in
+Fig. 3d. Instead of relying on the block granularity, we allow
+further splitting of each block along the batch dimension. Thus,
+when a block is too long, it can be split into two or more smaller
+effective blocks. Because a batch is split, the total workload
+can become larger because of GPU under-utilization. However,
+sometimes a slight increase in the total workload is dwarfed by
+the gain from workload balancing.
+
+Automatic hybrid distribution introduces a larger design
+space to workload distribution, which is difficult to tune manu-
+ally. To estimate throughputs of possible schedules, we measure
+consumed time of a few test execution for each block under
+feasible batch sizes. Then, considering the practical problem
+size of both B and N at around ten, the optimal solution can
+be found using an exhaustive search. Because the decision is
+made only once at the beginning, its overhead is amortized over
+the entire training and is negligible in our experiments.
+V. PIPE-BD FRAMEWORK
+A. Overall Procedure
+Algorithm 1 Pipe-BD procedure
+Input:
+1:
+G: # of devices, Di: i-th device
+2:
+Ti: Teacher blocks assigned to Di
+3:
+Si: Student blocks assigned to Di
+4: Initialization: Decide Ti and Si of each device // AHD
+5: for each epoch do
+6:
+for parallel i = 0, 1, ... , G − 1 do
+7:
+for each step do
+8:
+if Di.prev == ∅ then actini = load data()
+9:
+else actini = receive(from=Di.prev) // TR
+10:
+actt outi = Ti.forward(actini)
+11:
+if Di.next != ∅ then send(actt outi, to=Di.next) // TR
+12:
+acts outi = Si.forward(actini)
+13:
+Si.backward(L(acts outi, actt outi))
+14:
+if AHD enabled then Si.share gradient() // AHD
+15:
+if ∼DPU enabled then wait all devices() // DPU
+16:
+Si.update weight()
+17:
+end for
+18:
+end for
+19: end for
+Algorithm 1 displays the overall procedure for Pipe-BD.
+At initialization, the optimal schedule is decided from the
+profiled results, and the blocks are assigned to the devices
+(line 4). At the beginning of each step, each device receives
+the intermediate activation from the previous device (line 9).
+If Ti and Si contain the first block, the device instead starts
+with loading the data (line 8). It mostly overlaps with the
+computation except for the first step in each epoch. After the
+teacher forward pass is completed (line 10), the result is sent out
+such that the next device can execute Ti+1. (line 11). Then Si is
+executed (lines 12-13). If automatic hybrid distribution made a
+decision to share the block with other devices, gradient sharing
+is performed (line 14). Finally, decoupled parameter update
+(line 15) removes barrier operation, enabling each device to
+update its student weight without waiting for the other devices.
+B. Implementation
+We used a native PyTorch distributed package for
+point-to-point communications. All communications are imple-
+mented to overlap with computations as much as possible. We
+used Pytorch DistributedDataParallel class for data-
+parallel communications. For automatic hybrid distribution, the
+profiling function is called before training, which runs 100 steps
+of each block with feasible batch sizes to obtain execution
+times under the current environment. Based on these profiled
+execution times, Pipe-BD determines the best scheduling and
+starts training. The implementation of Pipe-BD is available at
+https://github.com/hongsunjang/Pipe-BD.
+VI. EXPERIMENTAL SETUP
+A. Workload
+To demonstrate the advantage of Pipe-BD, we applied it to
+two popular blockwise distillation applications.
+Neural Architecture Search. NAS is the current de facto
+standard for building a new neural network architecture. To
+search for a final architecture, multiple candidate operations
+in each layer are associated with a trainable architecture pa-
+rameter, representing the probability of selecting the operation
+every step. After the entire network is trained, the operation
+with the highest probability within each layer is selected as the
+final architecture. For an efficient search, blockwise distillation
+is a popular method [9, 12] for a smaller solution space. One
+notable aspect of NAS is that each step periodically requires
+two rounds of forward/backward passes for students: one for the
+architecture parameters and another for the weight parameters.
+However, this does not cause any difference to blockwise
+distillation or Pipe-BD because each round can be regarded
+as a single training step. We used ProxylessNAS [3] as the
+search backbone. For the teacher model, we used pre-trained
+MobileNetV2 [20]. For other settings, we followed the values
+suggested from the official implementations of DNA [9].
+Model Compression. Model compression is also another
+popular application of blockwise knowledge distillation [7, 11].
+A small student neural network model is trained from a larger
+pretrained teacher model. We follow the tradition and use layers
+of VGG-16 [21] as the teacher model and depth-wise separable
+convolution (DS-Conv) [22] layers as replacements. We follow
+the settings from [7] for the training.
+B. Experimental Environment
+For the experiments, we use two types of environments.
+By default, four RTX A6000 GPUs (Ampere) are attached
+to an AMD EPYC 7302 CPU. For additional experiments
+on a slightly low-cost configuration, four RTX 2080Ti GPUs
+(Turing) are attached to two Intel Xeon Silver 4214 CPUs.
+We used two datasets, CIFAR-10 [23] and ImageNet [1]. For
+the model compression, we used stochastic gradient descent
+(SGD) optimizer with a learning rate of 0.1 for compressing and
+0.0001 for finetuning. For the NAS, we used SGD optimizer
+with a learning rate of 0.005 for neural network architecture
+searching and 0.1 or 0.05 for retraining the final architecture.
+C. Baselines
+Based on the prior work mentioned in Section II-B, we
+used two baselines for our experiments. The first baseline (DP)
+is the traditional data-parallel blockwise distillation used in
+[9] official implementation. The second baseline (LS) is the
+layerwise scheduling introduced in [7]. Each baseline targets
+either one of neural architecture search or model compression,
+so we implemented these baselines to both of our target
+workloads in PyTorch.
+
+TABLE I: Experimental Environment
+HW
+Default
+(w/ A6000)
+GPU
+4× NVIDIA RTX A6000
+CPU
+1× EPYC 7302, 16 cores
+Memory
+256 GB DDR4 ECC
+Interconnect
+PCIe 4.0
+Alternative
+(w/ 2080Ti)
+GPU
+4× NVIDIA RTX 2080Ti
+CPU
+2× Xeon 4214 Silver, 12 cores
+Memory
+256 GB DDR4 ECC
+Interconnect
+PCIe 3.0
+SW
+Common
+Python
+3.10
+CUDA
+11.6
+PyTorch
+1.13
+NAS
+Teacher Model
+MobileNetV2
+Kernel Size
+3,5,7
+Expansion Ratio
+3,6
+Model
+Compression
+Teacher
+VGG-16
+Replacement
+DS-Conv
+0
+1
+2
+3
+4
+5
+Cifar-10
+ImageNet
+Time/step (ms)
+(a) NAS
+DP
+LS
+TR
+TR+DPU
+TR+IR
+TR+DPU+AHD
+0
+1
+2
+3
+4
+5
+6
+7
+8
+Cifar-10
+ImageNet
+(b) Model Compression
+0
+1
+2
+3
+4
+5
+Cifar-10
+ImageNet
+Speedup
+(a) NAS
+Fig. 4: Speedup and ablation of baselines and Pipe-BD.
+VII. EXPERIMENTAL RESULTS
+A. Speedup and Ablation
+Fig. 4 shows the speedup of Pipe-BD over the baselines
+with an ablation study of the proposed techniques using four
+RTX A6000 GPUs. Each colored bar shows the speedup of
+Pipe-BD where 1) only teacher relaying is applied (TR), 2)
+decoupled parameter update is further applied (TR+DPU), and
+3) all three schemes are applied, including automatic hybrid
+distribution (TR+DPU+AHD). In addition, we tested an alter-
+native method named Internal Relaying (TR+IR). With internal
+relaying, each device trains all existing blocks in every step,
+and parallelization is obtained via data parallelism. Instead of
+re-executing the teacher blocks or relaying activations between
+devices, the teacher activations are internally stored in memory
+and are retrieved for the next block. This approach allows for
+removing the redundancies of teacher and data loading as well
+as the load imbalance. However, it has the disadvantage of
+using a small batch size per device. In fact, internal relaying
+is a special case of Pipe-BD with TR+DPU+AHD when all
+blocks are only split along the batch dimension.
+Among the baselines, LS performs better than DP on Cifar-
+10 but worse on ImageNet. Because the composition of the
+neural networks for ImageNet typically has a few heavy blocks,
+LS suffers from severe load imbalance. Nevertheless, they
+both perform inferior to Pipe-BD. TR provides speedup for
+all cases due to eliminating extra data loading, redundant
+teacher execution, and enhancing resource utilization. Further,
+DPU provides additional speedup by removing synchronization
+barriers, which improves the overlapping of the teacher waiting
+time with student executions. Additionally, AHD removes the
+0
+1
+2
+3
+4
+2080Ti
+A6000
+Speedup
+(1) Cifar-10
+DP
+LS
+TR
+TR+DPU
+TR+DPU+AHD
+0
+1
+2
+3
+4
+2080Ti
+A6000
+Speedup
+(1) Cifar-10
+0
+1
+2
+3
+4
+5
+2080Ti
+A6000
+(2) ImageNet
+(a) Speedup.
+
+U
+U
+UP
+0
+0
+0
+1
+2
+3
+4
+5
+3
+4
+5
+2
+1
+0
+DP
+(b) 2080Ti schedule.
+U
+UP
+5
+4
+3
+0
+1
+2
+0
+1
+2
+0
+1
+2
+0
+1
+2
+0
+1
+2
+0
+1
+2
+3
+4
+5
+DP
+(c) A6000 schedule.
+Fig. 5: GPU type sensitivity of Pipe-BD on NAS.
+pipeline bubbles by balancing workloads, which drives an
+additional speedup over TR+DPU.
+With the ImageNet dataset, its larger spatial dimension of
+the images (224×224 vs 32×32) leads to heavy workloads in
+the first block. As a result, with TR only, the execution time
+of block 0 dominates all the others. Because of this, DPU has
+little room for improvement, whereas splitting the workload
+of the first block with AHD has a large impact on reducing
+the bubbles. In contrast, in the Cifar-10 case, the workload is
+already well-balanced only with TR+DPU version, and the gain
+from more balancing is offset by the loss from lower resource
+utilization caused by AHD.
+B. Sensitivity and Scheduling
+Fig. 5b and Fig. 5c show how Pipe-BD automatically de-
+termines the appropriate schedule according to two different
+environmental settings with the same NAS on ImageNet work-
+load. While the speedup trends are similar in Fig. 5a, they are
+from different schedules. The execution time of block 0 is the
+longest among the six blocks in both settings. However, the gap
+is more extensive on A6000 than on 2080Ti. To mitigate the
+imbalance, Pipe-BD settles at a schedule where the first three
+blocks (0-2) are shared on three devices (0-2) for A6000, while
+block 0 on 2080Ti is shared among two devices (0-1) and two
+blocks (1-2) are assigned to device 2.
+In Fig. 6, we demonstrate the sensitivity to the batch size
+on the NAS workload, normalized against DP of each batch
+size. In general, the advantage of Pipe-BD is not very sensitive
+to the batch size. One common trend is that the speedup is
+better in smaller batch sizes because the resource utilization
+difference becomes more significant with smaller batch sizes.
+One exception is AHD for ImageNet, where the speedup is
+better on larger batch sizes. The reason is found in the schedule
+depicted in Fig. 5c which uses three-way data parallelism to
+balance workloads. Because the training time for the student is
+shorter in both the baseline and Pipe-BD with larger batch sizes,
+reduction in the teacher redundancy and extra data loading
+account more for the overall speedup.
+C. Memory Overhead
+Fig. 7 depicts the memory overhead of Pipe-BD on the NAS
+task for each rank (GPU). Due to the characteristics of CNN-
+based models, lower-indexed teacher blocks generally have
+
+-
+-
+--
+-
+.
+.-
+-
+.
+-
+-
+.
+-
+-
+--
+-
+-
+-
+-
+-
+-
+-
+-
+--
+加
+-
+-
+.
+-
+--
+-
+-
+-
+-
+-
+-
+国.
+-
+-
+-
+-
+--
+-
+-
+-
+-
+-
+-
+--M.
+-
+-
+-
+--
+.
+.
+-
+--
+-
+-
+量-
+-
+-
+-
+-
+-
+-加
+-
+-
+-
+-
+-
+--
+-
+-
+-
+-
+.
+-
+-
+--
+-
+-
+.
+限
+-
+-限
+咖-
+-
+-
+-
+:
+-
+-
+-国TABLE II: Parallel Blockwise Distillation Training Results
+Task
+Dataset
+Teacher
+Student
+Elapsed Time (1 epoch)
+Model
+#Params
+FLOPs
+Acc. (%)
+Backbone
+#Params
+FLOPs
+Acc. (%)
+DP
+LS
+Pipe-BD
+NAS
+Cifar-10
+MobileNetV2
+2.24 M
+87.98 M
+95.42%
+ProxylessNAS [3]
+1.40 M
+76.10 M
+95.48%
+31.52s.
+16.33s.
+10.23s.
+ImageNet
+MobileNetV2
+3.50 M
+300.77 M
+72.00%
+ProxylessNAS
+4.22 M
+420.20 M
+74.54%
+62m 21s.
+125m 26s.
+14m 15s.
+Compression
+Cifar-10
+VGG-16
+14.72 M
+0.63 B
+91.85%
+DS-Conv [22]
+7.25 M
+0.39 B
+91.51%
+13m 18s.
+6m 37s.
+1m 49s.
+ImageNet
+VGG-16
+138.36 M
+30.98 B
+71.59%
+DS-Conv
+138.09 M
+26.15 B
+71.32%
+229m 23s.
+566m 49s.
+60m 39s.
+0
+1
+2
+3
+4
+128
+256
+384
+512
+Speedup
+(a) Cifar-10
+DP
+LS
+TR
+TR+DPU
+TR+DPU+AHD
+0
+1
+2
+3
+4
+128
+256
+384
+512
+Speedup
+(a) Cifar-10
+0
+1
+2
+3
+4
+5
+6
+128
+256
+384
+512
+(b) ImageNet
+Fig. 6: Batch size sensitivity of Pipe-BD on NAS.
+0
+2
+4
+6
+0
+1
+2
+3
+Max.
+Memory
+Allocation (GB)
+(a) Cifar-10
+DP
+LS
+TR/TR+DPU
+TR+DPU+AHD
+0
+1
+2
+3
+4
+5
+0
+1
+2
+3
+Max.
+Memory
+Allocation (GB)
+(a) Cifar-10
+0
+5
+10
+15
+20
+0
+1
+2
+3
+Max.
+(b) ImageNet
+Rank
+Rank
+Fig. 7: Memory overhead of Pipe-BD on NAS.
+larger feature map sizes. TR and DPU consume more memory
+than DP because of this characteristic, especially on rank 0.
+This outcome is also demonstrated in Fig. 7b because models
+for ImageNet contain even larger feature map sizes in lower-
+indexed blocks. However, AHD successfully addresses this
+issue using data parallelism in a hybrid manner, which lessens
+the memory overhead of earlier ranks as depicted in Fig. 5c. As
+a result, Pipe-BD provides superior multi-fold speedups with a
+minor 8.7% and 21.3% additional memory overheads over DP
+on average for Cifar-10 and ImageNet, respectively.
+D. Training Quality
+Pipe-BD has no component that can hurt the accuracy
+because it only alters the scheduling strategy. Nonetheless, we
+report the accuracy in Table II to demonstrate that the Pipe-BD
+framework faithfully reproduces the end-to-end training results
+in the prior art, with much shorter training time. For all use
+cases under evaluation, Pipe-BD achieves significant speedup
+with the same accuracy.
+VIII. CONCLUSION
+We propose Pipe-BD, a novel parallelization method for
+blockwise distillation. By restructuring the existing paralleliza-
+tion scheme, we achieve a multi-fold speedup on various use
+cases. In this study, we focused on a single-node, multi-
+GPU setting since it is the most common setup. However,
+if the method has to be scaled for a multi-node setting, the
+communication overhead needs to be addressed. Along with the
+heterogeneous GPU/servers, this will be our future direction.
+ACKNOWLEDGEMENT
+This work was partly supported by the National Research Foundation of
+Korea (NRF) grants (2022R1C1C1011307, 2022R1C1C1008131) and Samsung
+Electronics Co., Ltd (IO221213-04119-01) and Institute of Information &
+communications Technology Planning & Evaluation (IITP) grants (2020-0-
+01361) funded by the Korean government (MSIT).
+REFERENCES
+[1] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “ImageNet: A large-
+scale hierarchical image database,” in CVPR, 2009.
+[2] P. Goyal, P. Doll´ar, R. Girshick, P. Noordhuis, L. Wesolowski, A. Kyrola, A. Tulloch,
+Y. Jia, and K. He, “Accurate, Large Minibatch SGD: Training Imagenet in 1 Hour,”
+arXiv preprint arXiv:1706.02677, 2017.
+[3] H. Cai, L. Zhu, and S. Han, “ProxylessNAS: Direct Neural Architecture Search on
+Target Task and Hardware,” in ICLR, 2019.
+[4] H. Liu, K. Simonyan, and Y. Yang, “DARTS: Differentiable Architecture Search,”
+in ICLR, 2019.
+[5] Y. Huang, Y. Cheng, A. Bapna, O. Firat, D. Chen, M. Chen, H. Lee, J. Ngiam, Q. V.
+Le, Y. Wu, and z. Chen, “GPipe: Efficient Training of Giant Neural Networks Using
+Pipeline Parallelism,” in NeurIPS, 2019.
+[6] S. Zhao, F. Li, X. Chen, T. Shen, L. Chen, S. Wang, N. Zhang, C. Li, and H. Cui,
+“Naspipe: high performance and reproducible pipeline parallel supernet training via
+causal synchronous parallelism,” in ASPLOS, 2022.
+[7] C. Blakeney, X. Li, Y. Yan, and Z. Zong, “Parallel Blockwise Knowledge Distillation
+for Deep Neural Network Compression,” IEEE TPDS, vol. 32, no. 07, pp. 1765–
+1776, 2021.
+[8] H. Wang, H. Zhao, X. Li, and X. Tan, “Progressive Blockwise Knowledge Distilla-
+tion for Neural Network Acceleration,” in IJCAI, 2018.
+[9] C. Li, J. Peng, L. Yuan, G. Wang, X. Liang, L. Lin, and X. Chang, “Block-Wisely
+Supervised Neural Architecture Search With Knowledge Distillation,” in CVPR,
+2020.
+[10] L. Yuan, F. E. Tay, G. Li, T. Wang, and J. Feng, “Revisiting knowledge distillation
+via label smoothing regularization,” in CVPR, 2020.
+[11] J. Yu, S. Kang, and K. Choi, “Network recasting: a universal method for network
+architecture transformation,” in AAAI, 2019.
+[12] B. Moons, P. Noorzad, A. Skliar, G. Mariani, D. Mehta, C. Lott, and T. Blankevoort,
+“Distilling Optimal Neural Networks: Rapid Search in Diverse Spaces,” in ICCV,
+2021.
+[13] N. S. Keskar, D. Mudigere, J. Nocedal, M. Smelyanskiy, and P. T. P. Tang, “On
+large-batch training for deep learning: Generalization gap and sharp minima,” in
+ICLR, 2017.
+[14] D. Narayanan, A. Harlap, A. Phanishayee, V. Seshadri, N. R. Devanur, G. R. Ganger,
+P. B. Gibbons, and M. Zaharia, “PipeDream: Generalized Pipeline Parallelism for
+DNN Training,” in SOSP, 2019.
+[15] M. Shoeybi, M. Patwary, R. Puri, P. LeGresley, J. Casper, and B. Catanzaro,
+“Megatron-LM: Training Multi-billion Parameter Language Models Using Model
+Parallelism,” arXiv preprint arXiv:1909.08053, 2019.
+[16] J. Song, J. Yim, J. Jung, H. Jang, H.-J. Kim, Y. Kim, and J. Lee, “Optimus-CC:
+Efficient Large NLP Model Training with 3D Parallelism Aware Communication
+Compression,” in ASPLOS, 2023.
+[17] D. Kinghorn, “Gpu memory size and deep learning performance (batch size),”
+2018. [Online]. Available: https://www.pugetsystems.com/labs/hpc/GPU-Memory-
+Size-and-Deep-Learning-Performance-batch-size-12GB-vs-32GB----1080Ti-vs-
+Titan-V-vs-GV100-1146/
+[18] “Deep
+learning
+frameworks
+speed
+comparison,”
+2017.
+[Online].
+Available:
+https://wrosinski.github.io/deep-learning-frameworks/
+[19] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image
+Recognition,” in CVPR, 2016.
+[20] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2:
+Inverted Residuals and Linear Bottlenecks,” in CVPR, 2018.
+[21] K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-
+Scale Image Recognition,” in ICLR, 2015.
+[22] A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. An-
+dreetto, and H. Adam, “MobileNets: Efficient Convolutional Neural Networks for
+Mobile Vision Applications,” arXiv preprint arXiv:1704.04861, 2017.
+[23] A. Krizhevsky et al., “Learning multiple layers of features from tiny images,” 2009.
+
+-
+-
+-
+M
+--
+北
+二
+--
+-.-
+--
+---
+-
+-
+-
+-
+--
+-
+-
+-
+-
+--
+-
+-
+-
+-
+-
+-
+--
+-
+-
+-
+-
+-
+-
+-
+--
\ No newline at end of file
diff --git a/f9FMT4oBgHgl3EQf2TEy/content/tmp_files/load_file.txt b/f9FMT4oBgHgl3EQf2TEy/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..1eba6879b3318a2a7a71ba4acddf3615dcc69847
--- /dev/null
+++ b/f9FMT4oBgHgl3EQf2TEy/content/tmp_files/load_file.txt
@@ -0,0 +1,558 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf,len=557
+page_content='Pipe-BD: Pipelined Parallel Blockwise Distillation Hongsun Jang†, Jaewon Jung§, Jaeyong Song‡, Joonsang Yu¶, Youngsok Kim‡, and Jinho Lee†∗ †Department of Electrical and Computer Engineering, Seoul National University §Department of Artificial Intelligence, Yonsei University ‡Department of Computer Science, Yonsei University ¶CLOVA ImageVision, CLOVA AI Lab, NAVER hongsun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='jang@snu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='kr, {jungjaewon, jaeyong.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='song}@yonsei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='kr, joonsang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='yu@navercorp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='com, youngsok@yonsei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='kr, leejinho@snu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='kr Abstract—Training large deep neural network models is highly challenging due to their tremendous computational and mem- ory requirements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blockwise distillation provides one promising method towards faster convergence by splitting a large model into multiple smaller models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In state-of-the-art blockwise distillation methods, training is performed block-by-block in a data-parallel manner using multiple GPUs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To produce inputs for the student blocks, the teacher model is executed from the beginning until the current block under training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, this results in a high overhead of redundant teacher execution, low GPU utilization, and extra data loading.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To address these problems, we propose Pipe-BD, a novel parallelization method for blockwise distillation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Pipe-BD aggressively utilizes pipeline parallelism for blockwise distillation, eliminating redundant teacher block execution and increasing per-device batch size for better resource utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We also extend to hybrid parallelism for efficient workload balancing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As a result, Pipe-BD achieves significant acceleration without modifying the mathematical formulation of blockwise distillation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We implement Pipe-BD on PyTorch, and experiments reveal that Pipe-BD is effective on multiple scenarios, models, and datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Index Terms—Distributed Training, Knowledge Distillation, Neural Architecture Search, Model Compression I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' INTRODUCTION Modern deep neural network models are known to incur huge computational and memory requirements, especially with large- scale datasets [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With the continuing growth in model size, it takes tens, if not hundreds, of GPU days to train them [2], and the model size often exceeds the GPU memory capacity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Especially for methods that explore large solution spaces such as the neural architecture search (NAS) [3, 4], the problem becomes even more significant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This problem mandates the use of model parallelism [5, 6], which creates substantial throughput loss with inevitable pipeline bubbles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blockwise distillation [7, 8, 9] is one promising approach to mitigate such problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As illustrated in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 1, blockwise distillation splits the model into multiple smaller blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As opposed to traditional knowledge distillation methods that rely on input data and output labels from both ends, blockwise distillation uses the intermediate activation values of pretrained blocks of a ‘teacher’ to train each ‘student’ block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As a result, each block converges faster (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=', fewer epochs) due to the smaller solution space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Corresponding author Input T0 S0 T1 S1 T2 S2 T3 S3 𝑳(∆𝒐𝒖𝒕𝒑𝒖𝒕) Teachers Block 0 Block 1 Students Block 2 Block 3 Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 1: Conceptual diagram of blockwise distillation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Contrary to the earlier belief that teachers must be larger than students, recent studies have revealed that smaller teachers can be used to train larger students [10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With such findings, blockwise distillation is used in various fields such as model compression [7, 11] and NAS [9, 12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Since training a small teacher for a new task is quick and easy, blockwise distillation can be applied in most cases where traditional training is used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, the existing state-of-the-art methods for blockwise distillation [7, 9] exhibit several inefficiencies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Relying on the traditional data-parallel training scheme, they train each student block one by one independently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' While this fully exploits the independent nature of the blocks, it is not the best choice for training throughput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' First, to train a single intermediate student block, the teacher blocks must be executed from the beginning to the designated block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As a result, the teacher blocks exhibit substantial redundant execution, especially with blocks closer to the output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Second, with data parallelism, a batch of data is split among multiple GPUs, which leads to a smaller batch size per GPU, often resulting in resource under-utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Some approaches use a larger batch size to mitigate this [2], but it is known to be difficult to ensure model convergence [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Last, the data must be redundantly loaded for each student block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Unless the entire dataset fits into the GPU memory, the data are loaded from the CPU memory or disks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As the memory and disks are shared system-wide, the extra data loading becomes another significant overhead in training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To address the issues, we propose Pipe-BD, a novel par- allel training method for blockwise distillation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We assign individual student blocks to different devices and compute a teacher network in a relayed manner, which can reduce teacher redundancy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Inspired by approaches with pipeline paral- lelism [5, 14, 15, 16], we restructure the training schedule of the student blocks such that the training time is greatly improved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='12443v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='LG] 29 Jan 2023 Pipe-BD comprises three components: First, we propose teacher relaying.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Instead of relying on data parallelism, we spread the student model to multiple training devices (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=', GPUs) in a block granularity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Then, blocks of the teacher model are executed by relaying the intermediate activation values between the devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This approach has the advantages of eliminating extra data loading and increasing resource uti- lization from larger batch size per device.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Second, we propose decoupled parameter update to remove the scheduling bubbles and enhance the overall utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With teacher relaying, devices have to wait for the intermediate activation values from previous devices, creating scheduling bubbles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Decoupled parameter update performs model parameter updates in a mis- aligned manner and starts the next step right ahead, so those bubbles can be removed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Third, we suggest automatic hybrid distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Achieving a balance between devices is difficult with blockwise distillation because of the limited number of blocks available in typical neural network structures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Automatic hybrid distribution enables fine-grained balancing with further splitting blocks along the batch size dimensions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Pipe-BD is implemented on PyTorch and can automatically make all scheduling decisions to improve the throughput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Our extensive set of experiments shows Pipe-BD achieves a signifi- cant speedup over the state-of-the-art methods on multiple use cases and environments ranging from 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='37× to 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='38×.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' II.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' BACKGROUND AND RELATED WORK A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blockwise Distillation Blockwise distillation [7, 8, 9] is a promising direction for training a neural network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In traditional knowledge distillation, a student model is trained against a pre-trained teacher model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because the solution space size is identical to that of con- ventional supervised training, it faces convergence and training time problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blockwise distillation splits the larger teacher model into smaller ones and trains them blockwise as depicted in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Each teacher block (Ti) and student block (Si) pair obtains activation values from the previous teacher block (Ti−1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This pair performs forward pass using the activation as input and creates teacher output activation and student output activation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blockwise distillation minimizes a loss function (L(∆output)) which measures the difference between these two activations, to distill knowledge from a teacher block to the dedicated student block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This blockwise distillation process makes target problem spaces smaller and is known to converge faster.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Many applications such as NAS [9, 12] and model compression [7, 11] use blockwise distillation because of these characteristics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Parallelization Baseline of Blockwise Distillation State-of-the-art methods of blockwise distillation [9] use the traditional data-parallel scheme to further accelerate the training as illustrated in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This scheme trains a student block (Si) with all devices in a data-parallel manner for fixed n epochs, then moves on to train the next student block (Si+1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' It redundantly loads data multiple times because of this iterative training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Each student block (Si) requires the activation 0 10 20 30 Baseline Ideal Time/epoch (sec) Data loading T exec.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' S exec.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Idle Pipe-BD (Rank 0-3) Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 2: Motivational experiment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The breakdown demonstrates three major inefficiencies of baseline;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' redundant teacher execu- tion, extra data loading, and low resource utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' values from the previous teacher (Ti−1), so it also entails redundant teacher executions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Furthermore, it uses a smaller batch size per device which leads to under-utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Due to these inefficiencies, the data-parallel blockwise distillation suffers from poor scalability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' An alternative scheme [7] regards the training of each layer as a single task and adopts bin packing algorithm to balance the workload.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, it still has redundant teacher executions and suffers from workload imbalance when there are insufficient layers in the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' III.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' MOTIVATION In this section, we provide a motivational study highlighting the inefficiency of the existing parallel blockwise distillation training scheme and the need for a new approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 2 depicts the breakdown of time spent in parallel blockwise distillation with four RTX A6000 GPUs (NAS with Cifar-10;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' see Section VI-B for the detailed setup).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' ‘Baseline’ refers to the state-of-the-art parallel blockwise distillation method [9], where each block is trained sequentially using four devices with data parallelism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As displayed in the chart, the training time is spent on data loading, teacher execution (forward pass), and student execution (forward/backward pass).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, all three parts exhibit significant inefficiency, slow- ing down the training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To demonstrate the inefficiencies, we plot the ‘ideal’ bar in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 2 by measuring the training time of each part separately with a single GPU and dividing each time by four.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This represents an imaginary system with perfect parallelization and infinite device memory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The large gaps in teacher execution and data loading time occur because the baseline has many redundant teacher exe- cutions and extra data loading.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because each student block to train requires executing the teacher model from the beginning, the earlier teacher blocks are redundantly executed multiple times (see Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Similarly, block-by-block training forces loading data as many as the number of blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In addition, data- parallelism leads to smaller batch size per device, resulting in lower resource utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As demonstrated in several empirical studies [17, 18], a sufficient per-device batch size is critical for training throughput, which is the cause of the gap on student execution time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Pipe-BD targets these inefficiencies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As presented in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 2, Pipe-BD reduces the training time close to the ideal case, with only a small overhead (idle).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' UP T0 T0 T0 T0 T2 T2 T2 T2 T1 T1 T1 T1 S2 S2 S2 S2 DP DL Batch = 64 * 4 = 256 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' UP T0 T0 T0 T0 S0 S0 S0 S0 DP DL .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' n epoch n epoch n epoch UP T0 T0 T0 T0 T1 T1 T1 T1 S1 S1 S1 S1 DP DL .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' n epoch : i-th teacher exec.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' : Parameter update Ti Si : i-th student exec.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' DP : Data-parllel gradient sharing UP | U DL | D : Data loading : Step boundary Batch = 256 Redundant T0 T0 T0 T0 T2 T2 T2 T2 T1 T1 T1 T1 DL T3 T3 T3 T3 S3 S3 S3 S3 UP DP (a) Baseline UP T0 T1 T2 T3 S0 S1 S2 S3 UP T0 T1 T2 T3 S0 S1 S2 S3 UP T0 T1 T2 T3 S0 S1 S2 S3 D Batch = 256 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Redundant D D (b) w/ Teacher Relaying T0 T1 T2 S0 S1 S2 U U U U T0 T1 T2 T3 S0 S1 S2 S3 U U U U T0 T1 T2 T3 S0 S1 S2 S3 Batch = 256 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' U U U U T0 T1 T2 T3 S0 S1 S2 S3 D D D D (c) w/ Decoupled Parameter Update DL DL DL T0 T0 U UP S0 T1 T2 T3 S2 S3 S1 S0 DP T0 T0 U U UP S0 T1 T2 T3 S2 S3 S1 S0 DP T0 T0 U U UP S0 T1 T2 T3 S2 S3 S1 S0 DP U U UP T0 S0 T0 T1 T2 T3 S2 S3 S1 S0 DP .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Batch = 128 * 2 = 256 DL (d) w/ Automatic Hybrid Distribution Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3: Illustration of the techniques in Pipe-BD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' IV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' PIPE-BD METHOD A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Teacher Relaying Pipe-BD starts by restructuring the training pipeline of blockwise distillation with teacher relaying.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As opposed to the baseline (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3a) where a single block is fully trained in a data-parallel manner before moving on to the next, teacher relaying exclusively distributes the teacher and student blocks to all training devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Then, each device relays the intermediate teacher activation values to the next device as depicted in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The received activation is the input for both the teacher and the student block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The teacher block is executed first, whose output activation is sent to the next device such that the execution of the next block can start.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Overlapped with the transmission, the forward pass execution of the student starts, taking the same input as the teacher block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' After calculating the loss by comparing the output activations of the teacher and the student, the backward pass of the student follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' After all the backward passes are finished, parameter updates are performed on each block, completing the training step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The teacher relaying scheme has two advantages over the existing approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' First, each device executes the stages with larger batches and enjoys better resource utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For exam- ple, in the baseline using four devices with an effective batch size of 256, each device executes with a batch size of 64, which is often too small to fully utilize the hardware resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In contrast, with teacher relaying, each device would run with a full batch size of 256, increasing resource utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Second, the overhead of data loading is reduced.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' When the dataset is large, the data must come from the main memory or the disk, where both are system-wide shared resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because teacher relaying does not go through multiple training passes, the number of data loading decreases, leading to higher throughput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' One minor trade-off is communication overhead.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In the baseline, gradient sharing must occur after every backward pass.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With teacher relaying, there is some communication delay from relaying the intermediate activation values from one device to another.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, the communication time is almost negligible in our target settings of single-node multi- device training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Furthermore, in both cases, most of the com- munications overlapped with computations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Decoupled Parameter Update Although teacher relaying removes the redundant teacher executions, the removed redundancy is not directly translated to speedup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' At the beginning of each step, each device has to wait until the previous device delivers the intermediate activation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3c illustrates how decoupled parameter update addresses this problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As soon as the backward pass of each block is complete, the parameter updates are performed without waiting for the other devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Then, the teacher execution of the next step can start earlier, increasing the training throughput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This does not harm the training accuracy by any means because the student blocks have no dependency on the weight parameters of the other blocks, which is a special characteristic of blockwise knowledge distillation training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Decoupled parameter update successfully hides the teacher waiting time except for the beginning of each epoch, where full synchronization is needed for validating the whole model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because there are usually tens to hundreds of steps per epoch, such overhead is amortized to a negligible amount.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Automatic Hybrid Distribution With the teacher relaying and decoupled parameter update, the system throughput is determined by the throughput of the slowest device.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because of this, load balancing between devices plays a critical role in performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' One straightforward and intuitive load-balancing method is distributing the workload in contiguous blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The distribution is simple because there are only B−1CN−1 choices for B blocks and N devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Unfortunately, the naive distribution scheme often fails to provide a good balance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In blockwise distillation, the number of blocks B is determined by the neural network architecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Usually, B is around ten [3, 19] and N is four to eight within a single server.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because there are not enough number of blocks to distribute to the devices, the naive distribution is likely to end up in a severe workload imbalance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With automatic hybrid distribution, we provide another de- gree of freedom for workload distribution as presented in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 3d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Instead of relying on the block granularity, we allow further splitting of each block along the batch dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Thus, when a block is too long, it can be split into two or more smaller effective blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because a batch is split, the total workload can become larger because of GPU under-utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, sometimes a slight increase in the total workload is dwarfed by the gain from workload balancing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Automatic hybrid distribution introduces a larger design space to workload distribution, which is difficult to tune manu- ally.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To estimate throughputs of possible schedules, we measure consumed time of a few test execution for each block under feasible batch sizes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Then, considering the practical problem size of both B and N at around ten, the optimal solution can be found using an exhaustive search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because the decision is made only once at the beginning, its overhead is amortized over the entire training and is negligible in our experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' PIPE-BD FRAMEWORK A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Overall Procedure Algorithm 1 Pipe-BD procedure Input: 1: G: # of devices, Di: i-th device 2: Ti: Teacher blocks assigned to Di 3: Si: Student blocks assigned to Di 4: Initialization: Decide Ti and Si of each device // AHD 5: for each epoch do 6: for parallel i = 0, 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' , G − 1 do 7: for each step do 8: if Di.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='prev == ∅ then actini = load data() 9: else actini = receive(from=Di.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='prev) // TR 10: actt outi = Ti.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='forward(actini) 11: if Di.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='next !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='= ∅ then send(actt outi, to=Di.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='next) // TR 12: acts outi = Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='forward(actini) 13: Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='backward(L(acts outi, actt outi)) 14: if AHD enabled then Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='share gradient() // AHD 15: if ∼DPU enabled then wait all devices() // DPU 16: Si.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='update weight() 17: end for 18: end for 19: end for Algorithm 1 displays the overall procedure for Pipe-BD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' At initialization, the optimal schedule is decided from the profiled results, and the blocks are assigned to the devices (line 4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' At the beginning of each step, each device receives the intermediate activation from the previous device (line 9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' If Ti and Si contain the first block, the device instead starts with loading the data (line 8).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' It mostly overlaps with the computation except for the first step in each epoch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' After the teacher forward pass is completed (line 10), the result is sent out such that the next device can execute Ti+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' (line 11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Then Si is executed (lines 12-13).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' If automatic hybrid distribution made a decision to share the block with other devices, gradient sharing is performed (line 14).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Finally, decoupled parameter update (line 15) removes barrier operation, enabling each device to update its student weight without waiting for the other devices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Implementation We used a native PyTorch distributed package for point-to-point communications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' All communications are imple- mented to overlap with computations as much as possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We used Pytorch DistributedDataParallel class for data- parallel communications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For automatic hybrid distribution, the profiling function is called before training, which runs 100 steps of each block with feasible batch sizes to obtain execution times under the current environment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Based on these profiled execution times, Pipe-BD determines the best scheduling and starts training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The implementation of Pipe-BD is available at https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='com/hongsunjang/Pipe-BD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' VI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' EXPERIMENTAL SETUP A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Workload To demonstrate the advantage of Pipe-BD, we applied it to two popular blockwise distillation applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Neural Architecture Search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' NAS is the current de facto standard for building a new neural network architecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To search for a final architecture, multiple candidate operations in each layer are associated with a trainable architecture pa- rameter, representing the probability of selecting the operation every step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' After the entire network is trained, the operation with the highest probability within each layer is selected as the final architecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For an efficient search, blockwise distillation is a popular method [9, 12] for a smaller solution space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' One notable aspect of NAS is that each step periodically requires two rounds of forward/backward passes for students: one for the architecture parameters and another for the weight parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, this does not cause any difference to blockwise distillation or Pipe-BD because each round can be regarded as a single training step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We used ProxylessNAS [3] as the search backbone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For the teacher model, we used pre-trained MobileNetV2 [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For other settings, we followed the values suggested from the official implementations of DNA [9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Model Compression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Model compression is also another popular application of blockwise knowledge distillation [7, 11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' A small student neural network model is trained from a larger pretrained teacher model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We follow the tradition and use layers of VGG-16 [21] as the teacher model and depth-wise separable convolution (DS-Conv) [22] layers as replacements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We follow the settings from [7] for the training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Experimental Environment For the experiments, we use two types of environments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' By default, four RTX A6000 GPUs (Ampere) are attached to an AMD EPYC 7302 CPU.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For additional experiments on a slightly low-cost configuration, four RTX 2080Ti GPUs (Turing) are attached to two Intel Xeon Silver 4214 CPUs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' We used two datasets, CIFAR-10 [23] and ImageNet [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For the model compression, we used stochastic gradient descent (SGD) optimizer with a learning rate of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='1 for compressing and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='0001 for finetuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For the NAS, we used SGD optimizer with a learning rate of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='005 for neural network architecture searching and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='1 or 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='05 for retraining the final architecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Baselines Based on the prior work mentioned in Section II-B, we used two baselines for our experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The first baseline (DP) is the traditional data-parallel blockwise distillation used in [9] official implementation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The second baseline (LS) is the layerwise scheduling introduced in [7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Each baseline targets either one of neural architecture search or model compression, so we implemented these baselines to both of our target workloads in PyTorch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' TABLE I: Experimental Environment HW Default (w/ A6000) GPU 4× NVIDIA RTX A6000 CPU 1× EPYC 7302, 16 cores Memory 256 GB DDR4 ECC Interconnect PCIe 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='0 Alternative (w/ 2080Ti) GPU 4× NVIDIA RTX 2080Ti CPU 2× Xeon 4214 Silver, 12 cores Memory 256 GB DDR4 ECC Interconnect PCIe 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='0 SW Common Python 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='10 CUDA 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='6 PyTorch 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='13 NAS Teacher Model MobileNetV2 Kernel Size 3,5,7 Expansion Ratio 3,6 Model Compression Teacher VGG-16 Replacement DS-Conv 0 1 2 3 4 5 Cifar-10 ImageNet Time/step (ms) (a) NAS DP LS TR TR+DPU TR+IR TR+DPU+AHD 0 1 2 3 4 5 6 7 8 Cifar-10 ImageNet (b) Model Compression 0 1 2 3 4 5 Cifar-10 ImageNet Speedup (a) NAS Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 4: Speedup and ablation of baselines and Pipe-BD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' VII.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' EXPERIMENTAL RESULTS A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Speedup and Ablation Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 4 shows the speedup of Pipe-BD over the baselines with an ablation study of the proposed techniques using four RTX A6000 GPUs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Each colored bar shows the speedup of Pipe-BD where 1) only teacher relaying is applied (TR), 2) decoupled parameter update is further applied (TR+DPU), and 3) all three schemes are applied, including automatic hybrid distribution (TR+DPU+AHD).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In addition, we tested an alter- native method named Internal Relaying (TR+IR).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With internal relaying, each device trains all existing blocks in every step, and parallelization is obtained via data parallelism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Instead of re-executing the teacher blocks or relaying activations between devices, the teacher activations are internally stored in memory and are retrieved for the next block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This approach allows for removing the redundancies of teacher and data loading as well as the load imbalance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, it has the disadvantage of using a small batch size per device.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In fact, internal relaying is a special case of Pipe-BD with TR+DPU+AHD when all blocks are only split along the batch dimension.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Among the baselines, LS performs better than DP on Cifar- 10 but worse on ImageNet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because the composition of the neural networks for ImageNet typically has a few heavy blocks, LS suffers from severe load imbalance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Nevertheless, they both perform inferior to Pipe-BD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' TR provides speedup for all cases due to eliminating extra data loading, redundant teacher execution, and enhancing resource utilization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Further, DPU provides additional speedup by removing synchronization barriers, which improves the overlapping of the teacher waiting time with student executions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Additionally, AHD removes the 0 1 2 3 4 2080Ti A6000 Speedup (1) Cifar-10 DP LS TR TR+DPU TR+DPU+AHD 0 1 2 3 4 2080Ti A6000 Speedup (1) Cifar-10 0 1 2 3 4 5 2080Ti A6000 (2) ImageNet (a) Speedup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' U U UP 0 0 0 1 2 3 4 5 3 4 5 2 1 0 DP (b) 2080Ti schedule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' U UP 5 4 3 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 3 4 5 DP (c) A6000 schedule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 5: GPU type sensitivity of Pipe-BD on NAS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' pipeline bubbles by balancing workloads, which drives an additional speedup over TR+DPU.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' With the ImageNet dataset, its larger spatial dimension of the images (224×224 vs 32×32) leads to heavy workloads in the first block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As a result, with TR only, the execution time of block 0 dominates all the others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because of this, DPU has little room for improvement, whereas splitting the workload of the first block with AHD has a large impact on reducing the bubbles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In contrast, in the Cifar-10 case, the workload is already well-balanced only with TR+DPU version, and the gain from more balancing is offset by the loss from lower resource utilization caused by AHD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Sensitivity and Scheduling Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 5b and Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 5c show how Pipe-BD automatically de- termines the appropriate schedule according to two different environmental settings with the same NAS on ImageNet work- load.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' While the speedup trends are similar in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 5a, they are from different schedules.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The execution time of block 0 is the longest among the six blocks in both settings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, the gap is more extensive on A6000 than on 2080Ti.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' To mitigate the imbalance, Pipe-BD settles at a schedule where the first three blocks (0-2) are shared on three devices (0-2) for A6000, while block 0 on 2080Ti is shared among two devices (0-1) and two blocks (1-2) are assigned to device 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 6, we demonstrate the sensitivity to the batch size on the NAS workload, normalized against DP of each batch size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In general, the advantage of Pipe-BD is not very sensitive to the batch size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' One common trend is that the speedup is better in smaller batch sizes because the resource utilization difference becomes more significant with smaller batch sizes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' One exception is AHD for ImageNet, where the speedup is better on larger batch sizes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' The reason is found in the schedule depicted in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 5c which uses three-way data parallelism to balance workloads.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Because the training time for the student is shorter in both the baseline and Pipe-BD with larger batch sizes, reduction in the teacher redundancy and extra data loading account more for the overall speedup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Memory Overhead Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 7 depicts the memory overhead of Pipe-BD on the NAS task for each rank (GPU).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Due to the characteristics of CNN- based models, lower-indexed teacher blocks generally have -- .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='- .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' -- -- 加 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' -- 国.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' -- --M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' -- .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' -- 量- 加 -- .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' -- .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 限 限 咖- : 国TABLE II: Parallel Blockwise Distillation Training Results Task Dataset Teacher Student Elapsed Time (1 epoch) Model #Params FLOPs Acc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' (%) Backbone #Params FLOPs Acc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' (%) DP LS Pipe-BD NAS Cifar-10 MobileNetV2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='24 M 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='98 M 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='42% ProxylessNAS [3] 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='40 M 76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='10 M 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='48% 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='52s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='33s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='23s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' ImageNet MobileNetV2 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='50 M 300.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='77 M 72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='00% ProxylessNAS 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='22 M 420.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='20 M 74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='54% 62m 21s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 125m 26s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 14m 15s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Compression Cifar-10 VGG-16 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='72 M 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='63 B 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='85% DS-Conv [22] 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='25 M 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='39 B 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='51% 13m 18s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 6m 37s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 1m 49s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' ImageNet VGG-16 138.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='36 M 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='98 B 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='59% DS-Conv 138.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='09 M 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='15 B 71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='32% 229m 23s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 566m 49s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 60m 39s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 0 1 2 3 4 128 256 384 512 Speedup (a) Cifar-10 DP LS TR TR+DPU TR+DPU+AHD 0 1 2 3 4 128 256 384 512 Speedup (a) Cifar-10 0 1 2 3 4 5 6 128 256 384 512 (b) ImageNet Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 6: Batch size sensitivity of Pipe-BD on NAS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 0 2 4 6 0 1 2 3 Max.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Memory Allocation (GB) (a) Cifar-10 DP LS TR/TR+DPU TR+DPU+AHD 0 1 2 3 4 5 0 1 2 3 Max.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Memory Allocation (GB) (a) Cifar-10 0 5 10 15 20 0 1 2 3 Max.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' (b) ImageNet Rank Rank Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 7: Memory overhead of Pipe-BD on NAS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' larger feature map sizes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' TR and DPU consume more memory than DP because of this characteristic, especially on rank 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' This outcome is also demonstrated in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 7b because models for ImageNet contain even larger feature map sizes in lower- indexed blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, AHD successfully addresses this issue using data parallelism in a hybrid manner, which lessens the memory overhead of earlier ranks as depicted in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 5c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' As a result, Pipe-BD provides superior multi-fold speedups with a minor 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='7% and 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='3% additional memory overheads over DP on average for Cifar-10 and ImageNet, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Training Quality Pipe-BD has no component that can hurt the accuracy because it only alters the scheduling strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Nonetheless, we report the accuracy in Table II to demonstrate that the Pipe-BD framework faithfully reproduces the end-to-end training results in the prior art, with much shorter training time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' For all use cases under evaluation, Pipe-BD achieves significant speedup with the same accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' VIII.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' CONCLUSION We propose Pipe-BD, a novel parallelization method for blockwise distillation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' By restructuring the existing paralleliza- tion scheme, we achieve a multi-fold speedup on various use cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' In this study, we focused on a single-node, multi- GPU setting since it is the most common setup.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' However, if the method has to be scaled for a multi-node setting, the communication overhead needs to be addressed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Along with the heterogeneous GPU/servers, this will be our future direction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' ACKNOWLEDGEMENT This work was partly supported by the National Research Foundation of Korea (NRF) grants (2022R1C1C1011307, 2022R1C1C1008131) and Samsung Electronics Co.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=', Ltd (IO221213-04119-01) and Institute of Information & communications Technology Planning & Evaluation (IITP) grants (2020-0- 01361) funded by the Korean government (MSIT).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' REFERENCES [1] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Deng, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Dong, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Socher, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Fei-Fei, “ImageNet: A large- scale hierarchical image database,” in CVPR, 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [2] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Goyal, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Doll´ar, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Girshick, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Noordhuis, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wesolowski, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Kyrola, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Tulloch, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Jia, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' He, “Accurate, Large Minibatch SGD: Training Imagenet in 1 Hour,” arXiv preprint arXiv:1706.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='02677, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [3] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Cai, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhu, and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Han, “ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware,” in ICLR, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [4] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Liu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Simonyan, and Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Yang, “DARTS: Differentiable Architecture Search,” in ICLR, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [5] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Huang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Cheng, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Bapna, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Firat, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Lee, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Ngiam, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Le, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wu, and z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, “GPipe: Efficient Training of Giant Neural Networks Using Pipeline Parallelism,” in NeurIPS, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [6] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhao, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Shen, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wang, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Cui, “Naspipe: high performance and reproducible pipeline parallel supernet training via causal synchronous parallelism,” in ASPLOS, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [7] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blakeney, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Yan, and Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zong, “Parallel Blockwise Knowledge Distillation for Deep Neural Network Compression,” IEEE TPDS, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 32, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 07, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' 1765– 1776, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [8] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhao, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Tan, “Progressive Blockwise Knowledge Distilla- tion for Neural Network Acceleration,” in IJCAI, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [9] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Peng, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Yuan, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Liang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Lin, and X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chang, “Block-Wisely Supervised Neural Architecture Search With Knowledge Distillation,” in CVPR, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [10] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Yuan, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Tay, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Li, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wang, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Feng, “Revisiting knowledge distillation via label smoothing regularization,” in CVPR, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [11] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Yu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Kang, and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Choi, “Network recasting: a universal method for network architecture transformation,” in AAAI, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [12] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Moons, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Noorzad, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Skliar, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Mariani, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Mehta, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Lott, and T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Blankevoort, “Distilling Optimal Neural Networks: Rapid Search in Diverse Spaces,” in ICCV, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [13] N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Keskar, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Mudigere, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Nocedal, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Smelyanskiy, and P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Tang, “On large-batch training for deep learning: Generalization gap and sharp minima,” in ICLR, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [14] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Narayanan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Harlap, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Phanishayee, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Seshadri, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Devanur, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Ganger, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Gibbons, and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zaharia, “PipeDream: Generalized Pipeline Parallelism for DNN Training,” in SOSP, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [15] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Shoeybi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Patwary, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Puri, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' LeGresley, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Casper, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Catanzaro, “Megatron-LM: Training Multi-billion Parameter Language Models Using Model Parallelism,” arXiv preprint arXiv:1909.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='08053, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [16] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Song, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Yim, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Jung, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Jang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Kim, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Kim, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Lee, “Optimus-CC: Efficient Large NLP Model Training with 3D Parallelism Aware Communication Compression,” in ASPLOS, 2023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [17] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Kinghorn, “Gpu memory size and deep learning performance (batch size),” 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [Online].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Available: https://www.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='pugetsystems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='com/labs/hpc/GPU-Memory- Size-and-Deep-Learning-Performance-batch-size-12GB-vs-32GB----1080Ti-vs- Titan-V-vs-GV100-1146/ [18] “Deep learning frameworks speed comparison,” 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [Online].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Available: https://wrosinski.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='io/deep-learning-frameworks/ [19] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' He, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Ren, and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Sun, “Deep Residual Learning for Image Recognition,” in CVPR, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [20] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Sandler, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Howard, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhmoginov, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” in CVPR, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [21] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Simonyan and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zisserman, “Very Deep Convolutional Networks for Large- Scale Image Recognition,” in ICLR, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [22] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Howard, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Zhu, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Kalenichenko, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Wang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Weyand, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' An- dreetto, and H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Adam, “MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” arXiv preprint arXiv:1704.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='04861, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' [23] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' Krizhevsky et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=', “Learning multiple layers of features from tiny images,” 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content=' M -- 北 二 -- .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
+page_content='- -- --- -- -- -- --' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/f9FMT4oBgHgl3EQf2TEy/content/2301.12443v1.pdf'}
diff --git a/fNE1T4oBgHgl3EQfygV0/content/tmp_files/2301.03434v1.pdf.txt b/fNE1T4oBgHgl3EQfygV0/content/tmp_files/2301.03434v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..6fea4dae817265da3f063f98c516e9ecb2c9b927
--- /dev/null
+++ b/fNE1T4oBgHgl3EQfygV0/content/tmp_files/2301.03434v1.pdf.txt
@@ -0,0 +1,4423 @@
+arXiv:2301.03434v1 [math.RT] 9 Jan 2023
+Comet-shaped quiver varieties, Weyl group
+actions, and modified Kostka polynomials
+Mathieu Ballandras
+Université de Paris
+Scuola Internazionale Superiore di Studi Avanzati
+Instituto de Ciencias Matemáticas
+mballandras@imj-prg.fr
+January 10, 2023
+Abstract
+We study an algebra spanned by modified Kostka polynomials. Particular structure
+coefficients of this algebra are interpreted as traces of some Weyl group actions on
+the intersection cohomology of comet-shaped quiver varieties.
+Contents
+1
+Introduction
+2
+2
+Symmetric functions
+5
+2.1
+Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+5
+2.2
+Characters of the symmetric group and symmetric functions . . . . .
+7
+2.3
+Orthogonality and Macdonald polynomials . . . . . . . . . . . . . . .
+8
+2.3.1
+Generalities about scalar products on Sym [X] . . . . . . . . .
+8
+2.3.2
+Hall pairing and (q, t)-deformations . . . . . . . . . . . . . . .
+9
+2.4
+A result of Garsia–Haiman . . . . . . . . . . . . . . . . . . . . . . . .
+10
+3
+Geometric background
+12
+3.1
+Notations and generalities on the bounded derived category of con-
+structible sheaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+12
+3.2
+Intersection cohomology
+. . . . . . . . . . . . . . . . . . . . . . . . .
+14
+4
+Main objects and notations
+15
+4.1
+Adjoint orbits in gln
+. . . . . . . . . . . . . . . . . . . . . . . . . . .
+15
+4.1.1
+Notations for adjoint orbits
+. . . . . . . . . . . . . . . . . . .
+15
+4.1.2
+Resolutions of Zariski closures of adjoint orbits . . . . . . . . .
+16
+4.2
+The varieties QOµ,σ and their resolutions . . . . . . . . . . . . . . . .
+18
+4.3
+Intersection cohomology of the varieties QOµ,σ . . . . . . . . . . . . .
+21
+1
+
+5
+Construction in terms of Nakajima’s quiver varieties
+22
+5.1
+Generalities about Nakajima’s quiver varieties . . . . . . . . . . . . .
+22
+5.2
+Resolutions of Zariski closures of adjoint orbits as Nakajima’s framed
+quiver varieties
+. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+24
+5.3
+Comet-shaped quiver varieties . . . . . . . . . . . . . . . . . . . . . .
+26
+5.4
+Family of comet-shaped quiver varieties . . . . . . . . . . . . . . . . .
+28
+6
+Monodromic Weyl group action
+29
+6.1
+Family of resolutions of closures of adjoint orbits . . . . . . . . . . . .
+29
+6.2
+Decomposition of the family QL,P . . . . . . . . . . . . . . . . . . . .
+32
+6.3
+W-equivariant structure on the cohomology of the fibers of the family
+�QL,P . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+33
+6.4
+Monodromic Weyl group action on the cohomology of �QL,P ,σ . . . . .
+36
+6.5
+Frobenius morphism and monodromic action . . . . . . . . . . . . . .
+37
+7
+Geometric interpretations in the algebra spanned by Kostka poly-
+nomials
+41
+7.1
+Description of the algebra
+. . . . . . . . . . . . . . . . . . . . . . . .
+41
+7.2
+Interpretation of certain coefficients as traces of Weyl group actions
+on the intersection cohomology of quiver varieties . . . . . . . . . . .
+46
+7.3
+Cohomological interpretation in the multiplicative case . . . . . . . .
+47
+1
+Introduction
+The modified Kostka polynomials �Kλ,ρ(q, t) form a family of two-variable polyno-
+mials indexed by pairs of partitions of some integer n. They are a two-parameter
+deformation of the Kostka numbers and appear in the theory of symmetric func-
+tions. They were introduced by Garsia–Haiman [GH96] in the expression of the
+modified Macdonald polynomials in terms of Schur functions. The fact that they
+are polynomials with non-negative integer coefficients is an important result. This
+is known as the Macdonald conjecture [Mac88] which is a consequence of the n!-
+conjecture of Garsia–Haiman [GH93], proved by Haiman [Hai01]. In this article we
+study an aglebra spanned by those polynomials. Its structure coefficients cλ
+µ,ν(q, t)
+were introduced by Rodriguez-Villegas in unpublished notes, they are defined by
+�Kµ,ρ �Kν,ρ =
+�
+λ∈Pn
+cλ
+µ,ν �Kλ,ρ.
+with Pn the set of partitions of the integer n. We focus in particular on the co-
+efficients c1n
+µ,ν(q, t).
+They generalize the q, t-Catalan sequence of Garsia–Haiman
+[GH96].
+We give an interpretation of the coefficients c1n
+µ,ν(0, t) in terms of Weyl
+group action on the cohomology of a comet-shaped quiver variety. We also give a
+conjectural interpretation of the coefficients c1n
+µ,ν(q, t) in terms of cohomology of a
+partial resolution of a character variety. We prove the q = 1 specialization of the
+conjecture.
+The geometric framework is the following. The base field K is either C or Fq
+an algebraic closure of a field Fq with q elements. We consider cohomology with
+coefficients in κ which is either C or Ql with l and q coprime. Fix some integers
+n > 0, g ≥ 0 and k > 0. Let O = (O1, . . . , Ok) be a k-tuple of adjoint orbits in
+2
+
+gln, the Lie algebra of GLn. Denote by Oj the Zariski closure of the class Oj. Some
+genericity conditions are imposed on the eigenvalues of the k-tuple O (see Definition
+4.9). The main object in this article is the following variety:
+QO :=
+�
+(A1, B1, . . . , Ag, Bg, X1, . . . , Xk) ∈ gl2g
+n ×O1 × · · · × Ok
+��
+g
+�
+i=1
+[Ai, Bi] +
+k
+�
+j=1
+Xj = 0
+�
+// GLn,
+with [Ai, Bi] := AiBi − BiAi the Lie bracket. The quotient is a Geometric Invariant
+Theory (GIT) quotient with respect to the overall adjoint action of GLn. Those
+varieties were studied by Crawley-Boevey [CB03b, CB06] in genus g = 0. For any
+genus and semisimple adjoint orbits, they were studied by Letellier, Hausel and
+Rodriguez-Villegas [HLRV11]. Letellier [Let11] generalized to any type of adjoint
+orbits. Those varieties are called comet-shaped quiver varieties due to their interpre-
+tation as Nakajima’s quiver varieties. They are additive analogues of the following
+character varieties:
+MC :=
+�
+(A1, B1, . . . , Ag, Bg, X1, . . . , Xk) ∈ GL2g
+n ×C1 × · · · × Ck
+��
+A1B1A−1
+1 B−1
+1
+. . . AgBgA−1
+g B−1
+g X1 . . . Xk = Id
+�
+// GLn
+with C = (C1, . . . , Ck) a k-tuple of conjugacy classes in GLn. The action of GLn is
+by overall conjugation. Those varieties classify representations of the fundamental
+group of a genus g Riemann surface with k punctures and prescribed monodromies
+around those punctures. They were extensively studied by Hausel, Letellier and
+Rodriguez-Villegas [HLRV11] for semisimple classes and by Letellier [Let13] for any
+Jordan type. Similarly to the additive case there is a genericity condition imposed
+on the eigenvalues of C in order for the quotient to be well behaved (see Definition
+4.10).
+In general the varieties QO and MC are singular, therefore it is interesting to
+study their intersection cohomology. Letellier [Let11, Let13] studied intersection
+cohomology of those varieties by constructing resolutions of singularities �QL,P ,σ →
+QO (see Definition 4.12). Those resolutions of singularities originate in Springer
+theory [Spr76, BM83] and Lusztig’s theory of parabolic induction [Lus84, Lus85,
+Lus86]. Thanks to this origin, the resolutions of singularities (in both the additive
+and multiplicative cases) come with a Weyl group action on their cohomology. This
+action is called the Springer action. In this article we focus on the additive version
+(the comet-shaped quiver varieties). We come back to character varieties, at the
+end, in order to give a conjectural interpretation of the coefficient c1n
+µ,ν(q, t).
+In the additive case, thanks to the quiver variety point of view, Weyl group ac-
+tions other than the Springer action can be constructed. The Weyl group actions on
+the cohomology of Nakajima’s quiver varieties were studied extensively by Nakajima
+[Nak94, Nak00], Lusztig [Lus00] and Maffei [Maf02]. They were used to prove Kac
+conjecture by Letellier, Hausel, Rodriguez-Villegas [HLRV13] and to study unipo-
+tent characters of GLn(Fq) by Letellier [Let12]. Nakajima’s construction [Nak94] of
+the Weyl group action relies on the hyperkähler structure of the quiver varieties.
+We use this construction, together with technics from Lusztig (see Letellier [Let05,
+Proof of proposition 5.5.3]) and ideas from Mellit [Mel19, Section 8]. This allows to
+describe uniformly a monodromic action on the cohomology of quiver varieties with
+3
+
+semisimple adjoint orbits and the Springer action on the cohomology of resolutions
+of quiver varieties.
+Thanks to this construction we can obtain an interpretation of the coefficients
+c1n
+µ,ν(0, t) in terms of the Weyl group action on the compactly supported intersection
+cohomology of some comet-shaped quiver varieties.
+Theorem 1.1. Consider a generic 4-tuple of adjoint orbits of the following type:
+• O1 has one eigenvalue with Jordan type µ′ ∈ Pn,
+• O2 has one eigenvalue with Jordan type ν′ ∈ Pn,
+• O3 is semisimple regular, it has n distinct eigenvalues,
+• O4 is semisimple with one eigenvalue of multiplicity n − 1 and the other of
+multiplicity 1.
+Then the Weyl group with respect to O3 is the symmetric group Sn and it acts on
+the compactly supported intersection cohomology of QO. Let w be a n-cycle in this
+Weyl group, then
+c1n
+µ,ν (0, t) = t
+−dO
+2
+�
+r
+tr
+�
+w, IH2r
+c (QO, κ)
+�
+tr.
+In general, in the multiplicative case, only the Springer action exists. The mon-
+odromic action was constructed by Mellit [Mel19, Section 8] only with respect to
+one puncture for a regular monodromy. The Springer action is defined by Letellier
+[Let13] on the cohomology of partial resolutions of character varieties. The partial
+resolution relevant to describe the coefficient c1n
+µ,ν is the variety Mµ,ν introduced in
+7.3. The group Sn acts on the compactly supported intersection cohomology of
+Mµ,ν. For w ∈ Sn, the w-twisted mixed Hodge polynomial of Mµ,ν is
+IHw
+c (Mµ,ν; u, v) :=
+�
+i,r
+urvi tr
+�
+w, IH2r,i
+c
+(Mµ,ν, κ)
+�
+.
+with IH2r,i
+c
+(Mµ,ν, κ) the weight 2r graded part of the degree i compactly supported
+intersection cohomology of Mµ,ν.
+Conjecture 1.2. Let w be a n-cycle in Sn. The coefficient c1n
+µ,ν(q, t) is related to
+the w-twisted mixed Hodge polynomial of Mµ,ν by
+c1n
+µ,ν(q, t) = t
+− dim Mµ,ν
+2
+IHw
+c
+�
+Mµ,ν, 1
+q, √qt
+�
+.
+In [Bal22] we compute Poincaré polynomial for compactly supported intersec-
+tion cohomology of the character varieties MC. This allows to prove the following
+theorem which is the Poincaré polynomial specialization of Conjecture 1.2.
+Theorem 1.3. Let w be a n-cycle in Sn, the coefficient c1n
+µ,ν is related to the w-
+twisted Poincaré polynomial of Mµ,ν by
+c1n
+µ,ν(1, t) = t
+− dim Mµ,ν
+2
+�
+i
+t
+i
+2 tr
+�
+w, IHi
+c (Mµ,ν, κ)
+�
+.
+4
+
+Aknowledgement
+This work is part of my PhD thesis under the supervision of Emmanuel Letellier and
+Fernando Rodriguez-Villegas. I am very grateful to both of them for introducing
+me to the interesting topic of Weyl group actions on the cohomology of character
+varieties and quiver varieties. I am aslo thankful to François Bergeron for interesting
+comments and suggestions.
+2
+Symmetric functions
+2.1
+Generalities
+Notations 2.1 (Partitions). A partition of an integer n ∈ N is a decreasing sequence
+of non-negative integers
+λ = (λ1, λ2, . . . , λl(λ)) with |λ| := λ1 + λ2 + · · · + λl(λ) = n.
+The length of λ is the number l(λ) of non-zero terms. The set of partitions of n is
+denoted by Pn and
+P :=
+�
+n∈N
+Pn.
+The dominance ordering on P is defined by λ ⪯ µ if and only if |λ| = |µ| and
+k
+�
+i=1
+λi ≤
+k
+�
+i=1
+µi for all k ∈ N.
+For λ = (λ1, . . . , λl) a partition, we introduce the following notation
+Pλ := Pλ1 × · · · × Pλl.
+Notations 2.2 (Young diagrams). To a partition λ, we associate the following set
+{(i, j) |1 ≤ i ≤ l(λ) and 1 ≤ j ≤ λi} .
+This set is called the Young diagram of λ, it gives a graphical way to think about
+partitions. The transpose of a Young diagram is obtained by permuting i and j. The
+transpose λ′ of a partition λ is the partition with Young diagram the transpose of
+the Young diagram of λ. The Young diagram of the partition λ = (6, 4, 2) has the
+following graphical representation
+x
+.
+The box x has coordinates (i, j) = (1, 3). The arm length of x is number of box to
+the right of x, in this case a(x) = 3. The leg length is the number of box under x,
+we have l(x) = 1.
+5
+
+Notations 2.3 (Symmetric functions). Let X = (x1, x2, . . . ) be an infinite set of
+variable and let Sym[X] be the ring of symmetric functions in (x1, x2, . . . ). We use
+the usual notations from Macdonald’s book [Mac15]. In particular, the usual basis
+of symmetric functions indexed by partitions: mλ, eλ, hλ, pλ and sλ are respectively
+the monomial, elementary, complete, power sum and Schur symmetric functions.
+The Hall pairing is denoted by ⟨. . . , . . . ⟩ and is defined by
+⟨pλ, pµ⟩ = δλ,µzλ,
+(1)
+the symbol δλ,µ is 1 if λ and µ are the same partition and 0 otherwise. The order of
+the stabilizer of a permutation of cycle type λ is denoted by zλ, namely
+zλ =
+k
+�
+l=1
+iml
+l ml!
+for a partition λ = (i1, . . . i1
+�
+��
+�
+m1
+, . . . , ik, . . . ik
+�
+��
+�
+mk
+) .
+Definition 2.4 (Adams operator). For n ∈ Z>0, the Adams operator pn is a ring
+endomorphism of Sym[X]. It can be defined by its values on the generating family
+of power sums,
+pm [pn[X]] := pmn[X] for m ∈ N>0 and n ∈ N.
+The following notation is commonly used for Adams operators
+F [Xn] := pn [F[X]] .
+Remark 2.5. More generally, Adams operator are defined in any lambda ring. In
+this article, the only lambda rings appearing are rings of symmetric functions and
+polynomial rings such as K[u]. On such polynomial rings, the Adams operator pn is
+defined by pn[u] := un.
+Let k be a positive integer, we consider the space of multivariate symmetric
+functions in k infinite sets of variables over Q(q, t)
+Sym [X1, . . . , Xk] := Q(q, t) ⊗ Sym[X1] ⊗ · · · ⊗ Sym[Xk].
+A series with coefficients in this ring of multivariate symmetric functions will conve-
+niently encode cohomological information about comet shaped quiver varieties. The
+ring of such series is denoted by Sym [X1, . . . , Xk] [[s]], it is a lambda ring, and the
+Adams operators extend to ring endomorphisms of Sym [X1, . . . , Xk] [[s]] defined by
+pn
+�
+f(q, t)F1 [X1] ⊗ · · · ⊗ Fk [Xk] sl�
+= f(qn, tn)F1 [Xn
+1 ] ⊗ · · · ⊗ Fk [Xn
+k ] snl.
+Definition 2.6 (Plethystic substitution). Let F be a symmetric function. The ring
+of symmetric functions Sym [X] is freely generated by power sums, so that F can be
+uniquely obtained as a polynomial expression in the power sums (pn)n∈N. Interpreting
+pn as the Adams operator, the same polynomial expression defines an operator acting
+on any lambda ring Λ, this operator is denoted by F[. . . ]. For G ∈ Λ, the expression
+F[G] is called a plethystic substitution.
+6
+
+Remark 2.7. Similarly to Adams operators, the operator F[. . . ] naturally extends
+to Sym [X1, . . . , Xk] [[s]].
+Definition 2.8 (Plethystic exponential and logarithm). The plethystic exponential
+Exp : s Sym[X1, . . . , Xk][[s]] → 1 + Sym[X1, . . . , Xk][[s]] is defined by
+Exp[G] := exp
+��
+n≥1
+pn[G]
+n
+�
+,
+its inverse, the plethystic logarithm
+Log : 1 + s Sym[X1, . . . , Xk][[s]] → Sym[X1, . . . , Xk][[s]],
+is defined by
+Log[1 + G] :=
+�
+n≥1
+µ(n)
+n pn [log(1 + G)] ,
+where µ is the usual Mobius function. Contrarily to the ordinary ones, the plethystic
+exponential and logarithm are written with an uppercase character.
+Remark 2.9. Plethystics operations satisfy the relations
+Exp[F + G]
+=
+Exp[F] Exp[G],
+Log[(1 + F)(1 + G)]
+=
+Log[1 + F] + Log[1 + G],
+Log[Exp[G]]
+=
+G.
+2.2
+Characters of the symmetric group and symmetric func-
+tions
+Let Rn be the vector space spanned by characters of the symmetric group Sn and
+let R := �
+n≤0 Rn. There is a natural ring structure on R and a pairing such that R
+is naturally isomorphic to Sym[X]. Let us recall this well-known fact, see [Mac15]
+for more details.
+Let χ and η be two characters of Sn, and let Vχ, respectively Vη be the associated
+representations. The pairing is defined by
+⟨χ, η⟩ = dim HomSn (Vχ, Vη) .
+The spaces Rm and Rn are orthogonal if m ̸= n. The product of two characters
+χ ∈ Rm and η ∈ Rn is the character of the representation IndSm+n
+Sm×Sn Vχ ⊗ Vη, it is
+denoted by χ.η.
+The irreducible characters of the symmetric group Sn are indexed by partitions
+of n, they are denoted by (χλ)λ∈Pn. The irreducible representation associated to the
+character χλ is denoted by Vλ. The indexing is the same as in Macdonald’s book
+[Mac15], so that V(n) is the trivial representation and V(1n) the sign representation.
+Proposition 2.10. Define the characteristic map ch : R → Sym[X] by ch(χλ) := sλ.
+It is an isomorphism between R and Sym[X] compatible with the products and the
+pairings, Sym[X] being endowed with the Hall pairing.
+Proof. See Macdonald [Mac15, I-7].
+7
+
+Remark 2.11. The last proposition gives a representation theoretic meaning to
+symmetric functions. Consider V a representation of Sn with character χV , then
+• The pairing ⟨sλ, ch(χV )⟩ gives the multiplicity of the irreducible representation
+Vλ in the representation V .
+• The pairing ⟨pµ, ch(χV )⟩ gives the trace of an element in Sn with cycle type µ
+on the representation V .
+Definition 2.12 (Frobenius characteristic). We extend the characteristic map ch
+to bigraded representations of Sn by adding variables q and t to keep track of the
+degree. To a bigraded representation of the symmetric group V = �
+(i,j)∈N2 Vi,j is
+associated a symmetric function over Z(q, t). This symmetric function is given by
+ch(V ) =
+�
+λ∈Pn
+�
+(i,j)∈N2
+⟨Vi,j, χλ⟩ qitjsλ,
+(2)
+where we have identified the representation Vi,j with its character so that ⟨Vi,j, χλ⟩
+is the multiplicity of the irreducible representation of type λ in Vi,j. The symmet-
+ric function ch(V ) is called the q, t-graded Frobenius characteristic of the bigraded
+representation V .
+2.3
+Orthogonality and Macdonald polynomials
+In this section we recall the characterization of modified Macdonald polynomials
+following Mellit [Mel20, Mel18].
+2.3.1
+Generalities about scalar products on Sym [X]
+A scalar product on Sym [X] is a non-degenerate Q(q, t)-bilinear form
+(. . . , . . . )S
+:
+Sym [X] × Sym [X]
+→
+Q(q, t)
+F, G
+�→
+(F[X], G[X])S.
+It can be extended to multivariate symmetric functions by specifying the variable
+acted upon with a lower index
+(. . . , . . . )S
+X
+:
+Sym[X, Y1,· · · , Yk] × Sym[X, Z1,· · · , Zl]
+→
+Sym[Y1,· · · , Yk, Z1,· · · , Zl],
+on pure tensors it reads
+(F[X]⊗F ′[Y1,· · · , Yk], G[X]⊗G′[Z1,· · ·, Zl])S
+X := (F[X], G[X])SG′[Z1,· · · , Zl]F ′[Y1,· · · , Yk].
+Assumption 2.13 (Homogeneity). When considering families of symmetric func-
+tions indexed by partitions such as (uλ)λ∈P, the symmetric function uλ is always
+assumed to be homogeneous of degree |λ|.
+Definition 2.14 (Reproducing kernel). Let (uλ)λ∈P and (vµ)µ∈P be two basis of
+Sym[X] dual with respect to a scalar product (. . . , . . . )S. Then the element KS[X, Y ] ∈
+Sym[X, Y ] defined by
+KS[X, Y ] :=
+�
+λ∈P
+uλ[X]vλ[Y ],
+it is called the reproducing kernel of the scalar product (. . . , . . . )S. It depends only
+on the scalar product but not on the choice of dual basis, it satisfies
+(KS[X, Y ], F[X])S
+X = F[Y ].
+8
+
+2.3.2
+Hall pairing and (q, t)-deformations
+Remark 2.15. Recall that the Hall pairing satisfies
+⟨pλ, pµ⟩ = δλ,µzλ,
+hence (pλ)λ∈P and
+�
+z−1
+µ pµ
+�
+µ∈P are dual basis with respect to this pairing. This gives
+the reproducing kernel of the Hall pairing:
+Exp[XY ] =
+�
+λ∈P
+pλ[X]pλ[Y ]
+zλ
+=
+�
+n
+hn[XY ].
+Definition 2.16 ((q, t)-Hall pairing). The (q, t)-deformation of the Hall pairing is
+defined by
+(F[X], G[X])q,t
+:=
+⟨F[X], G[(q − 1)(1 − t)X]⟩ .
+Remark 2.17. The reproducing kernel of the (q, t)-Hall pairing is
+Exp
+�
+XY
+(q − 1)(1 − t)
+�
+.
+Proposition 2.18. Let M⪯λ be the subspace of Sym [X] spanned by monomial sym-
+metric functions mµ[X] with µ ⪯ λ. The Macdonald polynomials
+�
+˜Hλ[X; q, t]
+�
+λ∈P
+are uniquely determined by:
+• Orthogonality ( ˜Hλ[X; q, t], ˜Hµ[X; q, t])q,t = 0 if λ ̸= µ.
+• One of the triangularity condition ˜Hλ[X(t−1)] ∈ M⪯λ or ˜Hλ[X(q−1)] ∈ M⪯λ′.
+• Normalization ˜H[1; q, t] = 1.
+Moreover
+aλ(q, t) :=
+�
+˜Hλ[X; q, t], ˜Hλ[X; q, t]
+�q,t
+=
+�
+x∈λ
+(qa(x)+1 − tl(x))(qa(x) − tl(x)+1),
+(3)
+where the product is over the Young diagram of λ and a(x) is the arm length and
+l(x) the leg length (see Notations 2.1).
+Proof. [Mel20] corollary 2.8.
+The modified Macdonald polynomials ˜Hλ [X; q, t] were first introduced by Garsia–
+Haiman [GH96] as a deformation of other polynomials defined by Macdonald [Mac15].
+The definition recalled here comes from [Mel20].
+Definition 2.19 (Modified Kostka polynomials). The modified Kostka polynomials
+�
+�Kλ,ρ(q, t)
+�
+λ,ρ∈Pn are defined as the coefficients of the transition matrix between the
+basis of Schur functions and the basis of modified Macdonald polynomials
+˜Hρ[X; q, t] =
+�
+λ∈Pn
+�Kλ,ρ(q, t)sλ.
+Notations 2.20. The variables (q, t) will often be omitted and the modified Kostka
+polynomial denoted by �Kλ,ρ and the modified Macdonald polynomial by ˜Hλ[X].
+9
+
+2.4
+A result of Garsia–Haiman
+The remaining of this section is devoted to the presentation of a result of Garsia–
+Haiman [GH96, Theorem 3.4]. This result is important to study the coefficients c1n
+µ,ν
+in the next sections. Even though there are no new results, some proofs are included
+for convenience of the reader.
+Proposition 2.21. The operator ∆1 is defined by
+∆1F [X] := F[X] − F
+�
+X + (1 − q)(1 − t)
+z
+�
+Exp [−zX] |z0 ,
+where |z0 means taking the coefficient in front of z0. This operator acts on modified
+Macdonal polynomials by
+∆1 ˜Hλ [X; q, t] = (1 − t)(1 − q)
+�
+(i,j)∈λ
+qj−1ti−1 ˜Hλ [X; q, t] .
+Moreover the following relation holds
+˜Hλ [1 − u; q, t] =
+�
+(i,j)∈λ
+�
+1 − uqj−1ti−1�
+.
+(4)
+Proof. [GH96, Corollary 3.1 and Theorem 3.2]
+Lemma 2.22. At first order in u
+˜Hλ [1 + u; q, t] = 1 + u
+�
+(i,j)∈λ
+qj−1ti−1 + O(u2).
+(5)
+Proof. One should be careful with plethystic substitutions, to compute the left hand
+side of (5) one cannot just substitute −u for u in (4). Indeed pn[1 −u] = 1 −un and
+pn[1 + u] = 1 + un so that substituting −u for u in the latter gives back the former
+only when n is odd. We denote by dλ,µ the coefficient of pµ in the expansion of ˜Hλ
+in the basis of power sums (pκ)κ∈Pn, then
+˜Hλ [1 − u; q, t]
+=
+�
+|µ|=|λ|
+dλ,µ
+�
+i
+(1 − uµi),
+˜Hλ [1 + u; q, t]
+=
+�
+|µ|=|λ|
+dλ,µ
+�
+i
+(1 + uµi).
+We conclude by comparing the coefficient in front of u and using (4).
+Lemma 2.23. Let F ∈ Symn [X] be a symmetric function of degree n ≥ 2. Then
+the coefficient in front of u in F[1 + u] is given by the Hall pairing with a complete
+symmetric function
+F[1 + u]|u =
+�
+h(n−1,1)[X], F[X]
+�
+.
+Proof. The coefficient of mλ in the monomial expansion of F is denoted by cλ.
+The plethystic substitution F[1 + u] corresponds to the evaluation of the symmetric
+function F on the set of variables (1, u, 0, . . .), moreover
+F[1 + u] =
+�
+|λ|=n
+cλmλ[1 + u].
+10
+
+Therefore the only mλ contributing are the one with λ of length at most two and
+the coefficient in front of u is c(n−1,1). The conclusion follows as complete symmet-
+ric functions and monomial symmetric functions are dual with respect to the Hall
+pairing.
+Lemma 2.24. Let F ∈ Symn [X] be a symmetric function of degree n then
+F[1 − u]
+1 − u
+����
+u=1
+= ⟨F[X], pn[X]⟩ ,
+where |u=1 means setting u = 1.
+Proof. Let dλ be the coefficient in front of pλ in the power sum expansion of F,
+F[1 − u]
+=
+�
+|λ|=n
+dλpλ[1 − u]
+=
+�
+|λ|=n
+dλ
+�
+i
+(1 − uλi).
+When dividing by (1 − u) and setting u = 1 all terms coming from partitions of
+length at least two will vanish as (1 − u)2 divides them. Therefore we have
+F[1 − u]
+1 − u
+����
+u=1
+= d(n)
+1 − un
+1 − u
+����
+u=1
+= nd(n).
+The size of the centralizer of a n-cycle in Sn is z(n) = n, the conclusion follows by
+orthogonality of power sums (1).
+Now we can state and recall the proof of an important theorem of Garsia–
+Haiman.
+Theorem 2.25 (Garsia–Haiman [GH96] Theorem 3.4). We denote by �′
+(i,j)∈λ a
+product over the young diagram of a partition λ omitting the top left corner with
+(i, j) = (1, 1). The following identity holds
+(−1)n−1s(1n)[X] = (q − 1)(1 − t)
+�
+|λ|=n
+�
+(i,j)∈λ qj−1ti−1 �′
+(i,j)∈λ(1 − qj−1ti−1) ˜Hλ[X]
+aλ(q, t)
+.
+(6)
+Proof. The reproducing kernel of the (q, t)-Hall pairing is given in Remark 2.17. The
+degree n term of Exp[Z] is hn[Z]. The basis
+�
+˜Hλ[X]
+�
+λ∈P and
+� ˜Hλ[X]
+aλ
+�
+λ∈P are dual
+with respect to this scalar product. Following Definition 2.14 and Remarks 2.15,
+2.17, the degree n term of the reproducing kernel of the (q, t)-Hall pairing is
+hn
+�
+XY
+(q − 1)(1 − t)
+�
+=
+�
+|λ|=n
+˜Hλ[X] ˜Hλ[Y ]
+aλ
+.
+Now expand hn in the basis of power sums, proceed to the substitution Y = 1 − u
+and apply (4)
+�
+|µ|=n
+z−1
+µ pµ
+�
+X(1 − u)
+(q − 1)(1 − t)
+�
+=
+�
+|λ|=n
+˜Hλ[X] �
+(i,j)∈λ (1 − uqj−1ti−1)
+aλ
+.
+11
+
+Divide by (1 − u), set u = 1, apply Lemma 2.24 to the left hand side and compute
+explicitly the right hand side:
+�
+|µ|=n
+z−1
+µ
+�
+pµ
+�
+XY
+(q − 1)(1 − t)
+�
+, p(n)[Y ]
+�
+Y
+=
+�
+|λ|=n
+˜Hλ[X] �′
+(i,j)∈λ (1 − qj−1ti−1)
+aλ
+.
+As Adams operator are ring morphisms, we have
+pµ
+�
+XY
+(q − 1)(1 − t)
+�
+= pµ
+�
+X
+(q − 1)(1 − t)
+�
+pµ[Y ],
+then by orthogonality of power sums (1)
+p(n)
+�
+X
+(q − 1)(1 − t)
+�
+=
+�
+|λ|=n
+˜Hλ[X] �′
+(i,j)∈λ (1 − qj−1ti−1)
+aλ
+.
+(7)
+Apply the operator ∆1 to (7). According to Proposition 2.21, the operator ∆1 is
+diagonal in the basis of Macdonald polynomials and we obtain, up to a sign, the
+right hand side of (6). Let us compute the left hand side
+∆1p(n)
+�
+X
+(q−1)(1−t)
+�
+= p(n)
+�
+X
+(q−1)(1−t)
+�
+− p(n)
+�
+X
+(q−1)(1−t) − 1
+z
+�
+Exp[−zX] |z0
+= p(n)
+�
+X
+(q−1)(1−t)
+�
+− p(n)
+�
+X
+(q−1)(1−t)
+�
+Exp[−zX] |z0 + p(n)
+� 1
+z
+�
+Exp[−zX] |z0
+=
+1
+zn Exp[−zX] |z0 .
+In the second line we used that the Adams operator pn is a ring morphism and in the
+last line that it acts on z by raising to the power n. Now Exp[−zX] is the inverse
+of Exp[zX] so that if X is the infinite set of variables (x1, x2, . . . ), then
+Exp[−zX] =
+�
+i
+(1 − zxi).
+The coefficient in front of zn is (−1)nen[X] so that
+(−1)nen[X] = −(q − 1)(1 − t)
+�
+|λ|=n
+�
+(i,j)∈λ qj−1ti−1 �′
+(i,j)∈λ(1 − qj−1ti−1) ˜Hλ[X]
+aλ
+.
+To conclude, notice that en = s(1n).
+3
+Geometric background
+In this section we recall classical results about intersection cohomology. The main
+reference is Beilinson–Bernstein–Deligne–Gabber [BBDG18].
+3.1
+Notations and generalities on the bounded derived cate-
+gory of constructible sheaves
+The field K is either C or an algebraic closure Fq of a finite field Fq with q elements.
+Let X be an algebraic variety over K and let l be a prime different from the char-
+acteristic of K. We denote by κX the constant l-adic sheaf on X with coefficients in
+12
+
+Ql. When there are no risk of confusion we just write κ instead of κX. For K = C
+we can also consider the constant sheaf with complex coefficients, in the analytic
+topology.
+Notations 3.1. The bounded derived category of κ-constructible sheaves on X is
+denoted by Db
+c (X). Its objects are represented by complexes of sheaves K such that
+the cohomology sheaves HiK are κ-constructible sheaves on X and finitely many of
+them are non-zero. Let Y be a variety over K and let f : X → Y be a morphism,
+we have the usual four functors
+f ∗, f ! : Db
+c (Y ) → Db
+c (X) ,
+f∗, f! : Db
+c (X) → Db
+c (Y ) .
+Theorem 3.2 (Base change). Let K ∈ Db
+c (Y ′) and consider a cartesian square
+X′
+Y ′
+X
+Y,
+g
+b
+a
+f
+(8)
+then the natural morphism f ∗a!K → b!g∗K is an isomorphism.
+Remark 3.3. Let α ֒→ X be a geometric point of X and let β be its image by f.
+The variety Xα := X′ ×X α is the fiber of b over α and the variety Yβ := Y ′ ×Y β
+is the fiber of a over β. The cartesian square (8) induces the following cartesian
+square where h is an isomorphism
+Xα
+Yβ
+α
+β.
+h
+The base change isomorphism for this diagram identifies with the stalk at α of the
+base change isomorphism of Diagram (8),
+f ∗a!Kα → b!g∗Kα.
+This morphism is nothing but the morphism obtained by functoriality of the com-
+pactly supported cohomology
+H
+•
+c(Yβ, K)
+h∗
+−→ H
+•
+c(Xα, h∗K).
+Definition 3.4. Let W be a finite group acting from the left on a variety X. For
+all w ∈ W there is a morphism w : X → X.
+An action of W on an element
+K ∈ Db
+c (X) is the data of isomorphisms φw : w∗K ∼= K such that for all w, w′ ∈ W,
+φw′w = φww∗(φw′),
+(9)
+and such that φ1 = Id. We say that the complex K is W-equivariant.
+Remark 3.5. When the action of W on X is trivial, an action of W on K ∈ Db
+c (X)
+is just a group morphism from the opposite group W op to the group of automorphism
+Aut(K).
+13
+
+Proposition 3.6. Let f : X → Y be a W-equivariant morphism between varieties
+with left W-action. Let W act on K by morphisms φw : w∗K ∼= K, then W acts on
+f!K.
+Proof. Base change formulas allow to define the action, for all w ∈ W they provide
+an isomorphism w∗f!K → f!w∗K. Compose this isomorphism with f!φw to obtain
+an isomorphism �φw : w∗f!K → f!K. The compatibility (9) follows from functoriality
+of base change.
+3.2
+Intersection cohomology
+Definition 3.7 (Intersection complex). Let Y ֒→ X be a closed embedding and let
+j : U ֒→ Y be an open embedding. Assume that U is smooth, irreducible and that
+U = Y . Let ξ be a local system on U. The intersection complex IC•
+Y,ξ is the unique
+(up to isomorphism) element K in Db
+c (Y ) characterized by
+HiK
+=
+0 if i < − dim Y,
+H− dim Y K|U
+=
+ξ,
+dim
+�
+Supp HiK
+�
+<
+−i if i > − dim Y,
+dim
+�
+Supp HiDY K
+�
+<
+−i if i > − dim Y.
+We also denote by IC•
+Y,ξ its extension j∗IC•
+Y,ξ.
+Remark 3.8 (Continuation principle). The intersection complex of ξ can also be
+defined as the intermediate extension IC•
+Y,ξ = j!∗ξ. Moreover the functor j!∗ is fully
+faithful (see Kiehl-Weissauer [KW01, III - Corollary 5.11]).
+Remark 3.9. The intersection complex does not depend on the choice of smooth
+open subset in Y . When the local system ξ is not specified, it is chosen to be the
+constant sheaf κU and IC•
+X := IC•
+X,κU. We denote by IC•
+X
+Notations 3.10. The shifted intersection complexes are
+IC•
+X,ξ := IC•−dim X
+X,ξ
+and IC•
+X := IC•−dim X
+X
+.
+Definition 3.11 (Intersection cohomology). Let p : X → Spec K be the structural
+morphism and k an integer. The k-th intersection cohomology space of X is
+IHk(X, κ) := Hk−dim Xp∗IC•
+X = Hkp∗IC•
+X
+and the k-th compactly supported intersection cohomology space of X is
+IHk
+c (X, κ) := Hk−dim Xp!IC•
+X = Hkp!IC•
+X
+For K = C, Saito [Sai86] proved that the intersection cohomology spaces carry
+a mixed Hodge structure.
+Thus there exists on IHk
+c (X, Q) an increasing finite
+filtration called the weight filtration and denoted by W k
+• such that the complexified
+quotient C ⊗Q W k
+r /W k
+r−1 carries a pure Hodge structure of weight r. The Hodge
+numbers of this structure are denoted hi,j,k
+c
+(X) = dim IHi,j,k
+c
+(X, C) and satisfy i +
+j = r.
+14
+
+Definition 3.12. The mixed Hodge structure is encoded in the mixed Hodge poly-
+nomial,
+IHc (X; x, y, v) :=
+�
+i,j,k
+hi,j,k
+c
+(X)xiyjvk.
+(10)
+This polynomial has an important specialisation, the Poincaré polynomial
+Pc(X; v) := IHc (X; 1, 1, v) =
+�
+k
+dim IHk
+c (X, κ)vk.
+(11)
+In this article "Poincaré polynomial" always refers to "Poincaré polynomial for com-
+pactly supported intersection cohomology".
+4
+Main objects and notations
+4.1
+Adjoint orbits in gln
+4.1.1
+Notations for adjoint orbits
+The goal of this article is to relate some geometric objects to combinatorial data.
+The first step is the well-known labelling of adjoint orbits by their Jordan types,
+which we recall in order to fix the notations.
+For an integer r and for z ∈ K, we denote by Jr(z) ∈ glr the Jordan block of size r
+with eigenvalue z so that Jr(z)−z Idr is nilpotent of order r. Let µ = (µ1, µ2, . . . , µs)
+be a partition of an integer m and let z ∈ K. Let Jµ(z) be the matrix with eigenvalue
+z and Jordan blocks of sizes given by (µj)1≤j≤s,
+Jµ(z) :=
+
+
+
+
+
+Jµ1(z)
+Jµ2(z)
+...
+Jµs(z)
+
+
+
+
+ ∈ glm .
+Let ν = (ν1, . . . , νl) ∈ Pn be a partition of n, introduce the following notation
+Pν := Pν1 × Pν2 × · · · × Pνl.
+Consider a diagonal matrix σ,
+σ =
+
+
+
+
+
+σ1 Idν1
+σ2 Idν2
+...
+σl Idνl
+
+
+
+
+ ,
+(12)
+with σi ̸= σj for i ̸= j, so that νi is the multiplicity of the eigenvalue σi ∈ K.
+Notations 4.1. Consider an element µ =
+�
+µ1, . . . , µl�
+in Pν, we denote by Oµ,σ the
+adjoint orbit of the matrix
+Jµ,σ :=
+
+
+
+
+
+Jµ1(σ1)
+Jµ2(σ2)
+...
+Jµl(σl).
+
+
+
+
+ .
+15
+
+Let us recall a well-known proposition.
+Proposition 4.2. The Zariski closure of the adjoint orbit Oµ,σ is
+Oµ,σ =
+�
+ρ⪯µ
+Oρ,σ,
+the union is over the set of l-tuples ρ =
+�
+ρ1, . . . , ρl�
+with ρj ⪯ µj
+for 1 ≤ j ≤ l.
+The dominance order on partition was recalled in 2.1.
+Over Fq, there is a more precise description of adjoint orbits. Denote by F the
+Frobenius endomorphism of gln(Fq) raising the coefficients to the power q.
+Definition 4.3 (Type of an F-stable adjoint orbit). Let O be an F-stable adjoint
+orbit in gln(Fq), i.e., an orbit such that F(O) ⊂ O. Then the set OF of fixed points
+in O under the Frobenius is not empty. The characteristic polynomial of O has
+its coefficients in Fq so that its eigenvalues, which live in Fq, are permuted by the
+Frobenius. The spectrum of O, with multiplicities, reads
+
+
+
+
+�
+γ1, . . . , γqd1−1
+1
+�
+, . . . ,
+�
+γ1, . . . , γqd1−1
+1
+�
+�
+��
+�
+m1
+, . . . ,
+�
+γl, . . . , γqdl−1
+l
+�
+, . . . ,
+�
+γl, . . . , γqdl−1
+l
+�
+�
+��
+�
+ml
+
+
+
+ ,
+where γi ∈ Fq is such that γqdi−1
+i
+̸= γi, γqdi
+i
+= γi and γi ̸= γj for i ̸= j. Then the
+orbit O determines partitions ωi ∈ Pmi giving the size of the Jordan blocks of the
+Frobenius orbit of eigenvalues
+�
+γi, . . . , γqdi−1
+i
+�
+. Up to ordering, it defines a sequence
+ω = (d1, ω1) . . . (dl, ωl) in Z>0 × P called the type of the adjoint orbit.
+4.1.2
+Resolutions of Zariski closures of adjoint orbits
+In this section we recall the construction of resolutions of closures of adjoint orbits.
+The references for this construction are Kraft–Procesi [KP81], Nakajima [Nak98,
+Nak01], Crawley-Boevey [CB03a, CB03b] and Shmelkin [Shm09] (see also Letellier
+[Let11]).
+Using the notations from the previous section, consider an adjoint orbit Oµ,σ.
+The matrix σ ∈ gln is diagonal as in (12) and we denote by M its stabilizer in GLn,
+M =
+
+
+
+GLν1
+0
+0
+GLν2
+...
+0
+...
+
+
+ .
+Let µ = (µ1, . . . , µl) ∈ Pν so that µi is a partition of the integer νi. The transposed
+of the partition µi is denoted by µi′ =
+�
+µi
+1
+′, µi
+2
+′, . . .
+�
+. Let L be the subgroup of GLn
+formed by block diagonal matrices with blocks of size µi
+r
+′, it is a subgroup of M with
+16
+
+the following form
+L =
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ν1
+�
+��
+�
+GLµ1
+1
+′
+0
+0
+GLµ1
+2
+′
+...
+0
+...
+ν2
+�
+��
+�
+GLµ2
+1
+′
+0
+0
+GLµ2
+2
+′
+...
+0
+...
+...
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+.
+Notations 4.4. For a partition ν = (ν1, . . . , νl), define
+Sν := Sν1 × · · · × Sνl and GLν := GLν1 × · · · × GLνl .
+Now for ρ = (ρ1, . . . , ρl) ∈ Pν, we use the following notations,
+GLρ := GLρ1 × . . . GLρl =
+�
+r,s
+GLρrs
+and
+Sρ := Sρ1 × . . . Sρl =
+�
+r,s
+Sρrs.
+Then the previously introduced Levi subgroups satisfy M ∼= GLν and L ∼= GLµ′.
+Denote by P the parabolic subgroup of blocks upper triangular matrices having
+L as a Levi factor, then P = LUP with
+UP =
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ν1
+�
+��
+�
+Idµ1
+1
+′
+∗
+0
+Idµ1
+2
+′
+...
+0
+...
+*
+ν2
+�
+��
+�
+Idµ2
+1
+′
+∗
+0
+Idµ2
+2
+′
+...
+0
+...
+...
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+,
+and the Lie algebra counterpart of this Levi decomposition is p = l ⊕ uP.
+Proposition 4.5 (Resolutions of Zariski closures of conjugacy classes). Consider
+�YL,P,σ :=
+�
+(X, gP) ∈ gln × GLn /P
+��g−1Xg ∈ σ + uP
+�
+.
+The image of the projection to the first factor �YL,P,σ → gln is the Zariski closure of
+the adjoint orbit Oµ,σ. Moreover the following map is a resolution of singularities
+pσ
+:
+�YL,P,σ
+→
+Oµ,σ
+(X, gP)
+�→
+X
+.
+17
+
+Remark 4.6. If M (which is defined as the stabilizer of σ in GLn) is exactly L
+then the adjoint orbit Oµ,σ is semisimple. Semisimple orbits in gln are closed and
+smooth, for such orbits, pσ is an isomorphism.
+Remark 4.7. The decomposition theorem and Springer theory (or more generally
+Lusztig’s parabolic induction [Lus84, Lus85, Lus86]) provide more information about
+the previous resolution of singularities in terms of Weyl group representations. For
+G a reductive group and for T a maximal torus in G the Weyl group is denoted by
+WG := NG(T)/T.
+The Weyl group of M is WM ∼= �
+i Sνi and let Vρ be the representation �
+i Vρi of
+WM. The Weyl group of L is WL ∼= �
+i,j Sµi
+j
+′ and let ǫ be its sign representation.
+Lusztig’s parabolic induction provides the following decomposition
+H
+•+dµ
+c
+�
+�YL,P,σ; κ
+�
+=
+�
+ρ⪯µ
+HomWM
+�
+IndWM
+WL ǫ, Vρ
+�
+⊗ H
+•+dρ
+c
+�
+Oρ,σ; κ
+�
+,
+with dµ the dimension of Oµ,σ. Letellier [Let11] constructed an action of the relative
+Weyl group
+WM(L) = NM(L)/L
+on the spaces HomWM
+�
+IndWM
+WL ǫ, Vρ
+�
+.
+4.2
+The varieties QOµ,σ and their resolutions
+The varieties we are interested in are additive analogues of the character varieties
+classifying representations of the fundamental group of a compact Riemann surface
+with k punctures. They were studied by Crawley-Boevey [CB03b, CB06] in the case
+g = 0, by Hausel, Letellier and Rodriguez-Villegas [HLRV11] for semisimple adjoint
+orbits and by Letellier [Let11] in general.
+For each puncture an adjoint orbit Oµj,σj ⊂ gln is fixed. A bold symbol is used
+to represent k-tuple:
+µ
+:=
+�
+µ1, . . . , µk�
+,
+σ
+:=
+�
+σ1, . . . , σk�
+,
+Oµ,σ
+:=
+�
+Oµ1,σ1, . . . , Oµk,σk
+�
+.
+(13)
+Definition 4.8 (Comet-shaped quiver varieties). Consider the variety
+VOµ,σ :=
+�
+(A1, B1, . . . , Ag, Bg, X1, . . . , Xk) ∈ gl2g
+n ×Oµ1,σ1 × · · · × Oµk,σk
+���
+g
+�
+i=1
+[Ai, Bi] +
+k
+�
+j=1
+Xj = 0
+�
+.
+This is an affine variety acted upon by GLn by coordinate-wise adjoint action. The
+center of GLn acts trivialy so that the action factors through a PGLn-action. The
+main focus of the article is the following GIT quotient,
+QOµ,σ := VOµ,σ
+��
+PGLn = Spec
+�
+K
+�
+VOµ,σ
+�GLn�
+.
+(14)
+We call the varieties QOµ,σ the comet-shaped quiver varieties because of their inter-
+pretation as Nakajima’s quiver varieties that we will recall in Section 5.
+18
+
+Definition 4.9 (Generic adjoint orbits). Denote by ∆(σj) the multiset of eigenvalues
+of σj repeated according to multiplicities, σj
+r appears exactly νj
+r times in the multiset
+∆(σj). The k-tuple of adjoint orbits Oµ,σ is generic if and only if it satisfies the
+two following conditions:
+1.
+k
+�
+j=1
+�
+α∈∆(σj)
+α = 0,
+2. for any r ≤ n − 1 and for all (R1, . . . , Rk) with Rj ⊂ ∆(σj) of size r
+k
+�
+j=1
+�
+α∈Rj
+α ̸= 0.
+.
+Definition 4.10 (Generic conjugacy classes). If all the eigenvalues in σ are non-
+zero, then the adjoint orbits are also conjugacy classes in GLn. They are denoted
+by Cµ,σ =
+�
+Cµ1,σ1, . . . , Cµk,σk
+�
+instead of Oµ,σ. A k-tuple of conjugacy classes Cµ,σ
+is generic if it satisfies the two following conditions:
+1.
+k
+�
+j=1
+�
+α∈∆(σj)
+α = 1,
+2. for any r ≤ n − 1, for all (R1, . . . , Rk) with Rj ⊂ ∆(σj) of size r
+k
+�
+j=1
+�
+α∈Rj
+α ̸= 1.
+.
+Proposition 4.11 ([Let11] Proposition 5.2.4, Corollary 5.2.5). Let VOµ,σ := VOµ,σ ∩
+gl2g
+n ×Oµ1,σ1 × · · · × Oµk,σk, and let QOµ,σ be the image of VOµ,σ in QOµ,σ. Assume
+that Oµ,σ is generic, then
+QOµ,σ =
+�
+ρ⪯µ
+QOρ,σ
+is a stratification of QOµ,σ with smooth strata. Moreover, if it is non-empty,
+dim QOµ,σ = dµ = n2(2g − 2) + 2 +
+k
+�
+j=1
+dim Oµj,σj.
+As before, σj is a diagonal matrix with stabilizer Mj := ZGLn(σj) such that with
+Notations 4.4,
+Mj ∼= GLνj
+for some partition νj ∈ Pn given by the multiplicities of the eigenvalues of σj. The
+Jordan type of the eigenvalue σj
+i in the adjoint orbit Oµj,σj is µj,i ∈ Pνj
+i . Denote by
+19
+
+µj,i′ =
+�
+µj,i
+1
+′, µj,i
+2
+′, . . .
+�
+the transposed partition. Let Lj ⊂ Mj be the subgroup of
+block diagonal matrices as in 4.1.2,
+Lj ∼= GLµj,1
+1
+′ × GLµj,1
+2
+′ × . . .
+�
+��
+�
+⊂GLνj
+1
+× · · · × GL
+µ
+j,lj
+1
+′ × GL
+µ
+j,lj
+2
+′ × . . .
+�
+��
+�
+⊂GLνj
+lj
+.
+Let �YLj,P j,σj be a resolution of Oµj,σj as constructed in 4.1.2. Let L := �k
+j=1 Lj and
+let P := �k
+j=1 P j, then define
+�YL,P ,σ :=
+�
+1≤j≤k
+�YLj,P j,σj.
+Letellier [Let11] constructed resolutions of singularities of QOµ,σ
+Definition 4.12 (Resolutions of QOµj ,σj). Define
+�QL,P ,σ :=
+�
+(Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k ∈ gl2g
+n ×�YL,P ,σ
+�����
+g
+�
+i=1
+[Ai, Bi] +
+k
+�
+j=1
+Xj = 0
+�
+// PGLn .
+(15)
+The action of PGLn on gjP j is by left multiplication. The maps pσj : �YLj,P j,σj →
+Oµj,σj induce a map
+pσ : �QL,P ,σ → QOµ,σ,
+this morphism is a resolution of singularities.
+Remark 4.13. Similarly to Remark 4.6, if Lj = Mj for 1 ≤ j ≤ k, then the adjoint
+orbit Oµ,σ are semisimple and pσ is an isomorphism.
+Remark 4.14. Similarly to Remark 4.7,
+H•+dµ
+c
+�
+�QL,P ,σ, Ql
+�
+=
+�
+ρ⪯µ
+Vµ,ρ ⊗ IH•+dρ
+c
+�
+QOρ,σ, Ql
+�
+,
+with Vµ,ρ := �k
+j=1 HomWMj
+�
+Ind
+WMj
+WLj ǫ, Vρj
+�
+. Therefore, as in Remark 4.7, Letellier
+constructed an action of the relative Weyl group WM(L) = �k
+j=1 WMj(Lj) on the
+cohomology of the resolution �QL,P ,σ, we call this action the Springer action.
+Definition 4.15. Similar constructions exist in the multiplicative case (see Letellier
+[Let13]). If all the eigenvalues in σ are non-zero, then the adjoint orbits are actually
+conjugacy classes in GLn, they are denoted with a C instead of an O. For a generic
+k-tuple of conjugacy classes Cµ,σ, the character variety is defined by
+MCµ,σ :=
+�
+(A1, B1, . . . , Ag, Bg, X1, . . . , Xk) ∈ GL2g
+n ×Cµ1,σ1 × · · · × Cµk,σk
+���
+A1B1A−1
+1 B−1
+1
+. . . AgBgA−1
+g B−1
+g X1 . . . Xk = Id
+�
+// PGLn .
+20
+
+It admits a resolution of singularities
+�
+ML,P ,σ :=
+�
+(Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k ∈ GL2g
+n ×�YL,P ,σ
+��A1B1A−1
+1 B−1
+1
+. . . AgBgA−1
+g B−1
+g X1 . . . Xk = Id
+�
+// PGLn,
+and the cohomology of the resolution admits the following decomposition
+H•+dµ
+c
+�
+�
+ML,P ,σ, Ql
+�
+=
+�
+ρ⪯µ
+Vµ,ρ ⊗ IH•+dρ
+c
+�
+MCρ,σ, Ql
+�
+.
+(16)
+4.3
+Intersection cohomology of the varieties QOµ,σ
+Definition 4.16 (Hausel-Letellier-Villegas kernel). Let k ∈ Z>0 and let g ∈ Z≥0,
+the k-points, genus g Cauchy function is defined by
+Ωg
+k(z, w) :=
+�
+λ∈P
+Hλ(z, w)
+k
+�
+i=1
+˜Hλ
+�
+Xi, z2, w2�
+s|λ|,
+(17)
+with
+Hλ(z, w) :=
+�
+�
+z2a+1 − w2l+1�2g
+(z2a+2 − w2l) (z2a − w2l+2).
+(18)
+The degree n Hausel–Letellier–Villegas kernel is defined by
+HHLV
+n
+(z, w) := (z2 − 1)(1 − w2) Log Ωg
+k(z, w)
+��
+sn .
+This kernel was introduced to describe the cohomology of character varieties
+for genus g Riemann surfaces with k punctures. It also describes the cohomology
+of the varieties QOµ,σ.
+Consider a generic k-tuple of adjoint orbits Oµ,σ , with
+µ = (µ1, . . . , µk) and µj =
+�
+µj,1, . . . , µj,lj�
+. The transposed of the partition µj,i ∈ Pνj
+i
+is denoted by µj,i′ and we define the following symmetric function
+sµ′ :=
+k
+�
+j=1
+lj
+�
+i=1
+sµj,i′[Xj].
+(19)
+Theorem 4.17. Let Oµ,σ be a generic k-tuple of adjoint orbits.
+The Poincaré
+polynomial for compactly supported intersection cohomology of QOµ,σ is
+Pc
+�
+QOµ,σ, v
+�
+= vdµ �
+sµ′, HHLV
+n
+(0, v)
+�
+.
+Proof. For semisimple adjoint orbits, the variety is smooth, intersection cohomology
+coincides with usual cohomology and the theorem is proved by Hausel, Letellier and
+Rodriguez-Villegas [HLRV11]. The general case is proved by Letellier [Let11].
+Hausel–Letellier–Rodriguez-Villegas [HLRV11] proposed a conjecture for the mixed
+Hodge polynomial of character varieties with semisimple monodromies. It was gen-
+eralized by Letellier [Let13] to monodromies with any Jordan type.
+Conjecture 4.18 (Letellier [Let13], Conjecture 1.5). Let Cµ,σ be a generic k-tuple of
+conjugacy classes, the mixed Hodge polynomial (see Definition 3.12) of the character
+variety MCµ,σ is
+IHc(MCµ,σ; q, v) = (v√q)dµ
+�
+sµ′, HHLV
+n
+�−1
+√q, v√q
+��
+with q = xy.
+21
+
+5
+Construction in terms of Nakajima’s quiver vari-
+eties
+In order to study monodromic Weyl group action on the cohomology of the varieties
+QOµ,σ and �QL,P ,σ, they need to be put in a family for varying eigenvalues σ. One
+way to do that is to construct them as Nakajima’s quiver varieties [Nak94]. The
+family obtained is a fibration by the moment map. In this section we recall the
+construction of QOµ,σ as a comet-shaped quiver variety. In genus zero the construc-
+tion is due to Crawley-Boevey [CB03b], for any genus it is due to Hausel, Letellier,
+Rodriguez-Villegas [HLRV11] and Letellier [Let11].
+5.1
+Generalities about Nakajima’s quiver varieties
+In this section we recall the construction of Nakajima’s quiver varieties [Nak94] in
+order to fix the notations. Consider a quiver Γ with a set of vertices Ω0 and a set
+of edges Ω1. We denote by t(γ) the tail and by h(γ) the head of an edge γ ∈ Ω1. A
+dimension vector for Γ is an element v ∈ ZΩ0
+≥0. The space of quiver representations
+with dimension vector v is identified with a space of matrices,
+Rep (Γ, v) :=
+�
+γ∈Ω1
+MatK(vh(γ), vt(γ)).
+Its cotangent bundle T ∗ Rep (Γ, v) can be identified with the space of representations
+of an extended quiver �Γ. This extended quiver has the same set of vertices as Γ. It
+is obtained by adding an inverse γ to each edge γ ∈ Ω1:
+t(γ)•
+•h(γ).
+γ
+γ
+We denote by Ω1 the set of such inverted edges, then the set of edges of �Γ is �Ω :=
+Ω1 ⊔ Ω1. We have T ∗ Rep (Γ, v) ∼= Rep
+�
+�Γ, v
+�
+.
+For a dimension vector v ∈ ZΩ0
+≥0 consider the reductive group GLv := �
+i∈Ω0 GLvi.
+This group acts on Rep
+�
+�Γ, v
+�
+by
+(gvi)i∈Ω0 .(φγ)γ∈�Ω :=
+�
+gh(γ)xγg−1
+t(γ)
+�
+.
+The diagonal embedding of the multiplicative group K∗ in GLv acts trivially so that
+the action goes down to an action of the group Gv := GLv /K∗. The Lie algebra of
+GLv is
+glv :=
+�
+i∈Ω0
+glvi,
+and the Lie algebra of Gv is
+gv =
+�
+(xj)j∈Ω0 ∈ glv
+�����
+�
+j∈Ω0
+tr xj = 0
+�
+.
+22
+
+The center of this Lie algebra is given by
+Z(gv) =
+�
+(ξj Idvj)j∈Ω0
+�����(ξj)j∈Ω0 ∈ KΩ0 with
+�
+j∈Ω0
+vjξj = 0
+�
+.
+Consider an element θ ∈ ZΩ0 such that �
+i∈Ω0 θivi = 0. Such an element is called a
+stability parameter, it defines a character χθ of the group Gv by
+χθ ((gj)j∈Ω0) =
+�
+j∈Ω0
+det(gj)−θj.
+(20)
+We denote by Rep
+�
+�Γ, v
+�θ-ss
+and by Rep
+�
+�Γ, v
+�θ-s
+, the θ-semistable, respectively the
+θ-stable locus, in the sense of GIT, for the linearization χθ. Consider the moment
+map
+µ
+:
+Rep(�Γ, v)
+→
+gv
+(φγ)γ∈�Ω
+�→
+�
+γ∈�Ω ǫ(γ)φγφγ
+where ǫ(γ) = 1 for γ ∈ Ω1 and ǫ(γ) = −1 for γ ∈ Ω1.
+Definition 5.1 (Nakajima’s quiver variety). Let ξ be an element in Z(gv), the center
+of the Lie algebra of Gv. The Nakajima’s quiver variety Mθ
+v(ξ) is defined as the GIT
+quotient
+Mθ
+v(ξ) := µ−1(ξ) ∩ Rep
+�
+�Γ, v
+�θ-ss
+// Gv .
+We also need another kind of quiver varieties, the Nakajima’s framed quiver
+varieties. Fix a second dimension vector w ∈ NΩ0 and define
+Rep (v, w) :=
+�
+j∈Ω0
+MatK(vi, wi),
+Rep (w, v) :=
+�
+j∈Ω0
+MatK(wi, vi).
+An element g ∈ GLv acts on a = (aj)j∈Ω0 ∈ Rep (v, w) by
+g.a := (ajg−1
+j )j∈Ω0
+and it acts on b = (bj)j∈Ω0 ∈ Rep (v, w) by
+g.b := (gjbj)j∈Ω0.
+Introduce the space of framed quiver representations
+Rep
+�
+�Γ, v, w
+�
+:= Rep (v, w) ⊕ Rep (w, v) ⊕ Rep
+�
+�Γ, v
+�
+.
+In this context the moment map is
+µ′
+:
+Rep(�Γ, v, w)
+→
+glv
+(a, b, φ)
+�→
+(µ(φ)j − bjaj)j∈Ω0
+.
+For θ ∈ ZΩ0, consider the linearization
+χθ
+:
+GLv
+→
+K∗
+(gi)i∈Ω0
+�→
+�
+i∈Ω0 det(gi)−θi .
+Definition 5.2 (Nakajima’s framed quiver varieties). For ξ in the center of glv and
+for θ ∈ ZΩ0, the Nakajima’s framed quiver variety Mθ
+v,w(ξ) is defined as a GIT
+quotient with respect to the linearization χθ,
+Mθ
+v,w(ξ) := µ′−1(ξ) ∩ Rep
+�
+�Γ, v, w
+�θ-ss
+// GLv .
+23
+
+5.2
+Resolutions of Zariski closures of adjoint orbits as Naka-
+jima’s framed quiver varieties
+In this section we recall the construction of resolutions of closures of adjoint orbits as
+Nakajima’s framed quiver varieties. Those results come from Kraft-Procesi [KP81],
+Nakajima [Nak98, Nak01], Crawley-Boevey [CB03a, CB03b], Shmelkin [Shm09] and
+Letellier [Let11]. In this subsection and in 5.3, we fix the base field K = C.
+Let Oµ,σ be an adjoint orbit with semisimple part σ and Jordan type µ ∈ Pν as
+in 4.1.1. Consider the resolution �YL,P,σ → Oµ,σ as in 4.5. There is a Nakajima’s
+framed quiver variety realizing this resolution. Let d := �l
+i=1 µi
+1 and recall that
+L ∼=
+l�
+i=1
+µi
+1
+�
+r=1
+GLµir
+′ .
+The indices
+�
+µi
+r
+′�
+1≤i≤l
+1≤r≤µi
+1
+are relabelled (cs)1≤s≤d so that
+L ∼=
+d
+�
+s=1
+GLcs .
+Introduce the parameter ζ = (ζs)1≤s≤d such that ζs = σi if cs corresponds to µi
+r
+′ for
+some r. Consider the quiver ΓOµ,σ of type Ad−1 with summit indexed by integers
+between 1 and d − 1 and arrows going in the decreasing direction. Introduce the
+dimension vector vOµ,σ := (v1, ..., vd−1) with
+v1 := n − c1, vi := vi−1 − ci for i > 1.
+and w := (n, 0, . . . , 0). Define the parameter ξOµ,σ = (ξ1, ..., ξd−1) by ξi = ζi − ζi+1
+so that
+ξi :=
+�
+σk − σk+1
+if
+i = µ1
+1 + · · · + µk
+1
+0
+otherwise
+.
+The parameter ξOµ,σ will be identified with the element (ξj Idvj)1≤j≤d−1.
+We summarize everything in the following diagram showing the quiver, the di-
+mension vector, the parameter ζ and the parameter ξ.
+•1
+•2
+· · ·
+•µ1
+1+···+µk
+1
+· · ·
+•d−1
+n − c1
+n − c1 − c2
+· · ·
+n − ν1 − · · · − νk
+· · ·
+cr
+σ1
+σ1
+· · ·
+σk
+· · ·
+σr
+0
+0
+· · ·
+σk − σk+1
+· · ·
+0
+Remark 5.3. When writing the dimension vector under the quiver, we used the fact
+that |µi| = νi.
+24
+
+Consider a second dimension vector w = (n, 0, . . . , 0) and an extended represen-
+tation (a, b, φ) ∈ Rep
+�
+�ΓOµ,σ, vOµ,σ, w
+�
+. As wi = 0 unless i = 1, the component a is
+just a linear map a : V1 → W1 and b : W1 → V1 with W1 = Cn. For 1 ≤ i ≤ d − 2,
+denote by φi+1,i the linear map associated to the edge from i + 1 to i and by φi,i+1
+the map associated to the reverse edge from i to i+1. Such a representation belongs
+to µ′−1(ξOµ,σ) if and only if
+
+
+
+φ2,1φ1,2 − ba
+=
+(ζ1 − ζ2) Idv1
+φi+1,iφi,i+1 − φi−1,iφi,i−1
+=
+(ζi − ζi+1) Idvi
+for 2 ≤ i ≤ d − 2
+−φd−1,d−2φd−1,d−2
+=
+(ζd−1 − ζd) Idvd−1
+.
+(21)
+Those equations are called the preprojective relations.
+Example 5.4. The adjoint orbit of the matrix
+
+
+
+
+
+
+
+
+σ1
+1
+0
+0
+0
+0
+0
+σ1
+1
+0
+0
+0
+0
+0
+σ1
+0
+0
+0
+0
+0
+0
+σ1
+0
+0
+0
+0
+0
+0
+σ2
+0
+0
+0
+0
+0
+0
+σ2
+
+
+
+
+
+
+
+
+has Jordan type µ = ((3, 1), (1, 1)) ∈ P4 × P2 and we obtain
+W1
+V1
+V2
+V3
+vOµ,σ :
+4
+3
+2
+ζ :
+σ1
+σ1
+σ1
+ξOµ,σ :
+0
+0
+σ1 − σ2.
+b
+a
+φ1,2
+φ2,1
+φ2,3
+φ3,2
+Theorem 5.5. First consider the Nakajima’s framed quiver variety M0
+vO,w(ξOµ,σ)
+obtained from the previous data and stability parameter θ = 0. The following map
+is well-defined and is an isomorphism
+Ψ0
+:
+M0
+vOµ,σ ,w
+�
+ξOµ,σ
+�
+→
+Oµ,σ
+(a, b, φ)
+�→
+ab − σ1 Idn
+.
+Now take a stability parameter θ ∈ Zd−1
+>0 , the following map is an isomorphism
+Ψθ
+:
+Mθ
+vOµ,σ,w
+�
+ξOµ,σ
+�
+→
+�YL,P,σ
+(a, b, φ)
+�→
+(ab + σ1 Idn, fa,b,φ)
+,
+25
+
+where fa,b,φ is the flag 0 ⊂ Ed−1 ⊂ · · · ⊂ E1 ⊂ Cn defined by
+E1
+:=
+Im(a),
+Ei
+:=
+Im(a ◦ φ2,1 ◦ φ3,2 ◦ · · · ◦ φi,i−1)
+for 2 ≤ i ≤ d − 1.
+Moreover, the following diagram commutes
+Mθ
+vOµ,σ ,w
+�
+ξOµ,σ
+�
+�YL,P,σ
+M0
+vOµ,σ ,w
+�
+ξOµ,σ
+�
+Oµ,σ,
+Ψθ
+π
+pσ
+Ψ0
+where pσ is the resolution of Oµ,σ from Proposition 4.5 and π is the natural map
+from GIT theory.
+5.3
+Comet-shaped quiver varieties
+As in the previous subsection, we fix the base field K = C. Let Oµ,σ =
+�
+Oµ1,σ1, . . . , Oµk,σk
+�
+be a generic k-tuple of adjoint orbits in gln. We recall Crawley-Boevey’s result re-
+lating the variety QOµ,σ to a quiver variety. The idea is to glue together k quivers of
+type A corresponding to each adjoint orbit Oµj,σj to a central vertex 0 and to add
+g loops to this central vertex. We obtain the following comet-shaped quiver ΓOµ,σ,
+•[1,1]
+•[1,2]
+· · ·
+•[1,d1−1]
+•[2,1]
+•[2,2]
+· · ·
+•[2,d2−1]
+. . .
+•0
+. . .
+•[k,1]
+•[k,2]
+· · ·
+•[k,dk−1].
+The j-th leg is a quiver of type A with vertices labelled from [j, 1] to [j, dj − 1].
+The dimension vector vOµ,σ is defined such that its coordinate at the central vertex
+is n and its coordinates on the j-th leg coincide with the dimension vector vOµj ,σj
+described in the previous section. Similarly, the parameter ξOµ,σ is defined such that
+its coordinates on the j-th leg coincide with the parameter ξOµj,σj. The component
+at the central vertex ξOµ,σ,0 is defined such that vOµ,σ.ξOµ,σ = 0 hence
+nξOµ,σ,0 = −
+k
+�
+j=1
+dj−1
+�
+i=1
+vOµ,σ,[j,i]ξOµ,σ,[j,i].
+Consider a representation of the extended quiver φ ∈ Rep
+�
+�ΓOµ,σ, vOµ,σ
+�
+.
+• Denote by φ[j,i] the linear map associated to the arrow with tail [j, i] and φ[j,i]
+the linear map associated to the reversed arrow with head [j, i].
+26
+
+• For 1 ≤ i ≤ g the map associated to the i-th loop is denoted φi and the one
+associated to the reverse loop is denoted φi.
+As usual µ is the moment map and ξOµ,σ is identified with an element in the center
+of the Lie algebra gvOµ,σ and we let
+Xj := φ[j,1]φ[j,1] − ζ[j,1].
+If φ belongs to µ−1(ξOµ,σ) then Xj ∈ Oµj,σj. Indeed it follows from the previous
+description of closures of adjoint orbits as framed quiver varieties and the identifica-
+tion, for each leg, of the vector space at the central vertex with the framing vector
+space W1 from the previous section.
+Now if Ai is the linear map associated to the i-th loop of the quiver and Bi the
+map associated to the reversed loop, the preprojective relation at the central vertex
+is exactly the equation defining VOµ,σ. Then the following map is well-defined
+ΨOµ,σ
+:
+µ−1(ξOµ,σ)
+→
+VOµ,σ
+φ
+�→
+(A1, B1, . . . , Ag, Bg, X1, . . . , Xk) .
+Theorem 5.6. In the following diagram where the vertical arrows are quotient maps,
+the application ΨOµ,σ goes down to the quotient to an isomorphism ΦOµ,σ,
+µ−1(ξOµ,σ)
+VOµ,σ
+M0
+vOµ,σ(ξOµ,σ)
+QOµ,σ.
+ΨOµ,σ
+ΦOµ,σ
+Proof. It is proved by Crawley-Boevey [CB01, CB03b], see also Letellier [Let11,
+Proposition 5.2.2] for any genus.
+The resolution �QL,P ,σ of QOµ,σ, as introduced in 4.12, is also interpreted as a
+Nakajima’s quiver variety for the quiver ΓOµ,σ.
+Theorem 5.7. Consider a stability parameter θ associated to the quiver QOµ,σ such
+that θ[j,i] > 0 for each vertex [j, i]. There is an isomorphism ΦOµ,σ,θ : Mθ
+vOµ,σ(ξOµ,σ) →
+�QL,P ,σ and the following diagram commutes,
+Mθ
+vOµ,σ(ξOµ,σ)
+�
+QL,P ,σ
+M0
+vOµ,σ(ξOµ,σ)
+QOµ,σ,
+ΦOµ,σ,θ
+π
+pσ
+ΦOµ,σ
+where π is the natural projection from GIT theory.
+Proof. The map ΦOµ,σ,θ is constructed by Letellier [Let11, Section 5.3]. This map is
+induced by the map Ψθ of Theorem 5.5. Contrarily to Letellier’s article, we do not
+consider partial resolutions so that our parameter θ only has non-zero components.
+Therefore the dimension vector for the quiver variety Mθ
+vOµ,σ(ξOµ,σ) describing the
+resolution �QL,P ,σ is the same as the dimension vector of the quiver variety describing
+QOµ,σ.
+27
+
+The quiver variety point of view gives a criteria for non-emptiness. The question
+of emptiness of QOµ,σ (and its analogous character variety) is known as the Deligne-
+Simpson problem. See Kostov [Kos04] for a survey about this problem. Crawley-
+Boevey gave a solution to the problem in the generic case in terms of roots of quivers
+[CB03b], see also Letellier [Let11, Section 5.2]. Those results are summarized in the
+following theorem.
+Theorem 5.8. Let Oµ,σ be a generic k-tuple of adjoint orbits. The variety QOµ,σ is
+non-empty if and only QOµ,σ is not empty. This happens if and only if the dimension
+vector vOµ,σ is a root of the quiver ΓOµ,σ. This is always the case for g > 0.
+5.4
+Family of comet-shaped quiver varieties
+Now the field K is again either C or Fq. When the eigenvalues σ are varying, the
+varieties �QL,P ,σ fit in a family. First we describe explicitly this family, then we give
+an interpretation in terms of Nakajima’s quiver varieties and moment map.
+Notations 5.9. From now on the pair L, P is fixed. For short, let
+Z(l) := Z(l1) × · · · × Z(lk).
+Denote by B the subset of elements σ ∈ Z(l) such that the k-tuple of adjoint orbits
+Oµ,σ is generic. Note that the genericity condition depends only the semisimple part
+σ and not on the type µ. Then B is either empty or Zariski open in the hyperplane
+of Z(l) defined by the vanishing of the sum of the traces.
+Definition 5.10 (Family of varieties �
+QL,P ,σ). Define
+�VL,P :=
+��
+σ, (Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k
+� ��
+σ ∈ B, and (Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k ∈ VL,P,σ
+�
+,
+�QL,P := �VL,P // GLn .
+Denote by η the natural map η : �QL,P → B. The varieties �QL,P ,σ = η−1(σ) fit in a
+family �QL,P over B.
+The choice of L determines a unique quiver ΓOµ,σ and a unique dimension vector
+vOµ,σ independent of the choice of σ. Assume that the dimension vector is indivisible
+so that B is not empty. Then we can make the following assumption:
+Assumption 5.11 (Genericity of the stability parameter θ). The stability parameter
+θ is generic, i.e., it is a stability parameter for the quiver ΓOµ,σ with dimension vector
+vOµ,σ such that θ.vOµ,σ = 0 and θ.v ̸= 0 for a smaller dimension vector v.
+The construction of Theorem 5.7 extends to this family. It provides the following
+commutative diagram (the left vertical arrows is induced by the moment map µ)
+µ−1(zgen
+vOµ,σ)θ-ss//GvOµ,σ
+�QL,P
+zgen
+vOµ,σ
+B,
+Φ
+η
+(22)
+28
+
+where θ is a fixed generic stability parameter and zgen
+vOµ,σ is the subset of the center
+of the Lie algebra gvOµ,σ corresponding to the subset B under the correspondence
+between the parameters ξOµ,σ and the eigenvalues σ. Note that the correspondence
+between parameters of the quiver variety ξOµ,σ ∈ Z(gvOµ,σ) and Z(l) is not bijective,
+only difference of successive eigenvalues appear in the construction of the quiver
+variety.
+Thus the previous diagram relies on a choice of k − 1 eigenvalues.
+To
+σ ∈ Z(l) associate the element (ξOµ,σ, σ1
+1, . . . , σk−1
+1
+) in Z(gvOµ,σ) × Kk−1 this defines
+a bijective map
+h : Z(l)
+∼−→ zvOµ,σ × Kk−1.
+(23)
+Note that for a given parameter ξOµ,σ the genericity conditions is independant of
+the choice of the k − 1 eigenvalues, namely h−1(ξOµ,σ, σ1
+1, . . . , σk−1
+1
+) is generic if and
+only if h−1(ξOµ,σ, 0, . . . , 0) is generic. Therefore Diagram (22) can be modified in
+order to account for various choices of eigenvalues, then the horizontal arrows are
+isomorphisms and
+Kk−1 × µ−1(zgen
+vOµ,σ)θ-ss//GvOµ,σ
+�QL,P
+Kk−1 × zgen
+vOµ,σ
+B.
+Φ
+Id ×µ
+η
+(24)
+Theorem 5.12. If K = C, or if K = Fq and the characteristic is large enough, the
+cohomology sheaves Hiη!κ are constant sheaves.
+Proof. When K = C this follows from the quiver variety point of view of Diagram
+(24), it is a well-known fact used by Nakajima [Nak94] to construct a Weyl group
+action on the cohomology of quiver varieties (see also [Bal20]). To prove the result
+for K = Fq we can change characteristic as in [HLRV13, proof of Theorem 2.3]. This
+implies the result in large enough characteristic.
+6
+Monodromic Weyl group action
+The goal of this section is to construct and study the monodromic Weyl group action
+on the cohomology of the quiver varieties �QL,P ,σ. We use technics from Nakajima
+[Nak94], Lusztig (see Letellier [Let05, Proof of proposition 5.5.3]), Mellit [Mel19,
+Section 8] and Hausel–Letellier–Rodriguez-Villegas [HLRV13].
+The construction
+relies essentially on Theorem 5.12.
+6.1
+Family of resolutions of closures of adjoint orbits
+In this section we study a family formed by the varieties �YL,P,σ when σ is varying. It
+will be usefull in the next section to study a family of comet-shaped quiver varieties
+and to obtain some dimension estimates to prove Lemma 6.8.
+Let P be a parabolic subgroup of GLn and let L be a Levi factor of P, then L is
+isomorphic to a group of block diagonal matrices GLc1 × · · ·×GLcr. The Lie algebra
+of L and of UP are denoted by l respectivly by uP. At the level of the Lie algebras
+the Levi decomposition reads p = l ⊕ uP. The center of this Lie algebra l is denoted
+29
+
+by Z(l) and its regular locus is
+Z(l)reg = {x ∈ Z(l) |ZG(x) = L} .
+Define
+�Yreg
+L,P =
+�
+(x, gL) ∈ gln × GLn /L
+��g−1xg ∈ Z(l)reg�
+.
+Consider the projection on the first factor preg : �Yreg
+L,P → gln, denote by Yreg
+L,P its
+image. This image consists of semisimple elements with r distinct eigenvalues with
+multiplicities c1, . . . , cr. Consider the relative Weyl group WGLn(L) = NGLn(L)/L,
+and for each w ∈ WGLn(L) chose a representative ˙w ∈ NGLn(L). This relative Weyl
+group acts on Z(l) by
+w.σ := ˙wσ ˙w−1.
+Consider the fiber product
+Z(l)reg
+Yreg
+L,P ×Z(l)reg/WGLn(L) Z(l)reg
+Z(l)reg/WGLn(L)
+Yreg
+L,P,
+χ
+with χ the characteristic polynomial. Note that the following map is an isomorphism
+�Yreg
+L,P
+→
+Yreg
+L,P ×Z(l)reg/WGLn(L) Z(l)reg
+(x, gL)
+�→
+(x, g−1xg).
+(25)
+Therefore the WGLn(L)-action on Z(l)reg induces an action on �Yreg
+L,P.
+It is given
+explicitly by
+w.(x, gL) = (x, g ˙w−1L).
+Then the morphism
+�Yreg
+L,P
+preg
+−−→ Yreg
+L,P
+is a Galois cover with group WGLn(L). This relative Weyl group acts on the push
+forward of the constant sheaf preg
+∗ κ. Define
+�YL,P =
+�
+(x, gP) ∈ gln × GLn /P
+��g−1xg ∈ Z(l) ⊕ uP
+�
+.
+Remark 6.1. An element gP ∈ GLn /P is identified with a partial flag
+0 = Er ⊂ Er−1 ⊂ · · · ⊂ E1 ⊂ Kn
+such that dim Ei−1/Ei = ci for all 1 ≤ i ≤ r. Indeed GLn acts transitively on such
+flags and the stabilizer of any of them is isomorphic to P. Then a point (x, gP) in
+�YL,P consists of an endomorphism x ∈ gln and a partial flag gP preserved by x such
+that x acts as a scalar on Ei−1/Ei for all 1 ≤ i ≤ r.
+Denote by YL,P the image of the projection to the first factor p : �YL,P → gln.
+Note that the map p is proper. The following theorem is a particular case of [Lus84,
+Lemma 4.3 and Proposition 4.5].
+It can be seen as a generalization of Borho–
+MacPherson [BM83] approach to Springer theory.
+30
+
+Theorem 6.2. The subvariety Yreg
+L,P is open, smooth and dense in YL,P. The fol-
+lowing square is cartesian
+�Yreg
+L,P
+�YL,P
+Yreg
+L,P
+YL,P,
+i
+preg
+p
+(26)
+with i the map (x, gL) → (x, gP). Moreover p!κ = IC•
+YL,P , preg
+!
+κ so that WGLn(L) acts
+on p!κ.
+Remark 6.3. The morphism preg is a Galois cover and i is an open embedding so
+that the dimension can be easily computed
+dim YL,P = dim �YL,P = dim �Yreg
+L,P = dim GLn − dim L + dim Z(L).
+(27)
+Let us describe the relation with the resolutions of closures of adjoint orbits
+introduced in 4.5. Let σ ∈ Z(l) and let M := ZGLn(σ) be the stabilizer of σ in GLn.
+The notations from 4.1.2 are used so that M ∼= GLν for ν a partition of n. Moreover
+L ⊂ M and the integers (c1, c2, . . . , cr) are relabelled (µ1
+1
+′, µ1
+2
+′, . . . ) so that µi′ is a
+partition of νi. The inclusion L ⊂ M comes from the inclusions
+GLµi
+1
+′ × · · · × GLµi
+li
+′ ⊂ GLνi .
+The resolution of the closure of Oµ,σ fits in the following diagram
+�YL,P
+�YL,P,σ
+YL,P
+Oµ,σ = �
+ρ⪯µ Oρ,σ
+p
+pσ
+(28)
+The decomposition Oµ,σ = �
+ρ⪯µ Oρ,σ actualy comes from a decomposition of YL,P.
+Define
+Y
+M,ρ
+L,P :=
+�
+σ′∈Z(m)reg
+Oρ,σ′.
+This decomposition is similar to the one introduced by Shoji [Sho88].
+Proposition 6.4. The variety Y
+M,ρ
+L,P is smooth of dimension
+dim Y
+M,ρ
+L,P = dim Oρ,σ + dim Z(m).
+The variety YL,P admits the following decomposition
+YL,P =
+�
+M
+�
+ρ⪯µ
+Y
+M,ρ
+L,P .
+The first union is over the set of stabilizers of elements σ ∈ Z(l). In the second
+union, µ depends on M as previously described. The unique part indexed by M = L
+is Yreg
+L,P.
+31
+
+Proof. Denote by Zρ the stabilizer in GLn of the element Jρ,σ in Oρ,σ (see Notations
+4.1). There is a finite cover
+Z(m)reg × GLn /Zρ
+→
+Y
+M,ρ
+L,P
+�
+σ′, gZρ
+�
+�→
+gJρ,σ′g−1,
+hence Y
+M,ρ
+L,P is smooth and
+dim Y
+M,ρ
+L,P = dim Oρ,σ + dim Z(m).
+6.2
+Decomposition of the family QL,P
+In this section we study a family related to the family �QL,P introduced briefly in
+5.4. We compute some dimensions which will be useful to prove Lemma 6.8.
+Notations 6.5. First we recall the notations from 6.1 in this context. For 1 ≤ j ≤ k,
+�YLj,P j :=
+�
+(X, gjP j) ∈ gln × GLn /P j ��g−1
+j Xgj ∈ Z(lj) ⊕ uP j �
+and define
+�YL,P := �YL1,P 1 × · · · × �YLk,P k.
+Then YL,P is the image in glk
+n of the map p forgetting the partial flags gjP j,
+p
+:
+�YL,P
+→
+glk
+n
+(Xj, gjP j)1≤j≤k
+�→
+(Xj)1≤j≤k.
+Similarly VL,P, respectively QL,P, is obtained from �VL,P , respectively �
+QL,P , by for-
+getting the partial flags.
+In this section a decomposition of the family QL,P is deduced from the decom-
+position Oµ,σ = �
+ρ⪯µ Oρ,σ and from the decomposition introduced in Proposition
+6.4,
+YL,P =
+�
+M
+�
+ρ⪯µ
+Y
+M,ρ
+L,P .
+The decomposition is used in the next section (Lemma 6.8) in order to define a Weyl
+group action.
+Let YB
+L,P be the subset of elements in YL,P with generic semisimple parts, i.e.,
+with a k-tuple of semisimple parts belonging to B. The set YB
+L,P is assumed to be
+non-empty. The dimension of YB
+L,P is computed similarly to dim YL,P in Remark
+6.3,
+dim YB
+L,P = kn2 + dim B −
+k
+�
+j=0
+dim Lj.
+The decomposition YL,P = �
+M
+�
+ρ⪯µ Y
+M,ρ
+L,P induces a similar decomposition for YB
+L,P ,
+YB
+L,P =
+�
+M
+�
+ρ⪯µ
+YB,M,ρ
+L,P
+,
+32
+
+where M = (M1, . . . , Ml) and YB,M,ρ
+L,P
+is the subset of elements in
+Y
+M1,ρ1
+L1,P 1 × · · · × Y
+Mk,ρk
+Lk,P k
+with a generic k-tuple of semisimple parts. From the computation of the dimension
+of Y
+M,ρ
+L,P in Proposition 6.4, we deduce that when Z(m) ∩ B is not empty
+dim YB,M,ρ
+L,P
+=
+n
+�
+j=1
+dim Oρj,σj + dim Z(m) ∩ B.
+(29)
+Now the decomposition of YB
+L,P induces a decomposition of the family of quiver
+varieties QL,P and we define
+QM,ρ
+L,P :=
+�
+VL,P ×YB
+L,P YB,M,ρ
+L,P
+�
+// PGLn .
+Proposition 6.6. The variety QL,P admits the following decomposition
+QL,P =
+�
+M
+�
+ρ⪯µ
+QM,ρ
+L,P .
+When non-empty, the dimension of a part is
+dim QM,ρ
+L,P = n2(2g − 2) + 2 + dim Z(m) ∩ B +
+k
+�
+j=1
+dim Oρj,σj.
+(30)
+Proof. The dimension of QM,ρ
+L,P can be computed just like the dimension of QOµ,σ
+(see Hausel, Letellier, Rodriguez-Villegas [HLRV11, Theorem 2.2.4] and Letellier
+[Let11, Corollary 5.2.3]). The computation relies on the smoothness of YB,M,ρ
+L,P
+which
+follows from the smoothness of Y
+B,Mj,ρj
+Lj,P j
+and on the expression (29) for the dimension
+of YB,M,ρ
+L,P
+.
+6.3
+W-equivariant structure on the cohomology of the fibers
+of the family �QL,P
+In this section we use the family �QL,P → B in order to construct a Weyl group action
+on the cohomology of the varieties �QL,P ,σ for σ ∈ B. The Weyl group studied in
+this section is
+W := WGLn(L1) × · · · × WGLn(Lk).
+Each WGLn(Lj) is isomorphic to a symmetric group and acts on Z(lj) by permuting
+the distinct eigenvalues with the same multiplicities. Then the Weyl group W acts
+on B, for w = (w1, . . . , wk) ∈ W and σ =
+�
+σ1, . . . , σk�
+∈ B,
+w.σ :=
+�
+˙w1σ1 ˙w−1
+1 , . . . , ˙wkσk ˙w−1
+k
+�
+,
+where ˙wj is a representative in GLn of wj ∈ WGLn(Lj). Consider the diagram
+B
+�QL,P
+B/W
+QL,P .
+π0
+η
+p
+χ
+(31)
+33
+
+Thanks to the quiver variety point of view, the cohomology sheaves Hiη!κ are con-
+stant (Theorem 5.12). In this section a W-equivariant structure on those cohomol-
+ogy sheaves is constructed. The Weyl group actions on the cohomology of quiver
+varieties with such constant sheaves were introduced by Nakajima [Nak94]. Here
+we also use a method from Lusztig (see [Let05, Proof of Proposition 5.5.3]), this
+method is also applied by Laumon-Letellier [LL19, Section 5.2]. This approach al-
+lows to extend the equivariant structure away from a regular locus. Mellit obtained
+a similar result with a different construction for character varieties [Mel19, Section
+8].
+Before constructing the equivariant structure, let us define the regular locus.
+Denote by Breg the subset of regular elements, i.e. elements
+�
+σ1, . . . , σk�
+∈ B such
+that ZGLn(σj) = Lj. It is the locus of B where the W-action is free. Diagram (31)
+is pulled back to the regular locus
+Breg
+�Qreg
+L,P
+Breg/W
+Qreg
+L,P .
+πreg
+ηreg
+preg
+χreg
+(32)
+Similarly to (25), notice that
+Qreg
+L,P ×Breg/W Breg ∼= �Qreg
+L,P.
+(33)
+Theorem 6.7. The cohomology sheaves Hiη!κ admit a W-equivariant structure over
+B.
+Proof. Consider the diagram
+�QL,P
+B
+QL,P ×B/W B
+B/W
+QL,P
+η
+p
+c
+π0
+a
+b
+χ
+,
+(34)
+the group W acts on QL,P ×B/W B and the morphism a is W-equivariant. The
+variety Qreg
+L,P ×Breg/W Breg is smooth, dense and open in QL,P ×B/W B. The constant
+sheaf κ over Qreg
+L,P ×Breg/W Breg is W-equivariant. Indeed for w ∈ W we can define a
+morphism
+φw : w∗κ → κ
+which is the identity on the stalks.
+It satisfies the conditions of Definition 3.4.
+Applying the continuation principle from Remark 3.8, this W-equivariant structure
+extends to a W-equivariant structure on IC•
+QL,P ×B/W B. Notice that η!κ ∼= a!c!κ. We
+shall see in Lemma 6.8 that
+c!κ ∼= IC•
+QL,P ×B/W B.
+Then the W-equivariant structure on c!κ induces a W-equivariant structure on η!κ.
+Up to the isomorphism c!κ ∼= IC•
+QL,P ×B/W B, the theorem is proved.
+34
+
+It remains to prove the following lemma.
+Lemma 6.8. There is an isomorphism c!κ ∼= IC•
+QL,P ×B/W B.
+Proof. Because of the isomorphism (33), the restriction of c!κ to the smooth locus
+Qreg
+L,P ×Breg/W Breg is the constant sheaf κ.
+In order to verify the hypothesis of
+Definition 3.7 it remains to prove that the map c is small, i.e. that it satisfies the
+following inequality
+dim
+�
+x ∈ QL,P ×B/W B
+��dim c−1(x) ≥ d
+�
+≤ dim QL,P ×B/W B − 2d for all d > 0.
+We use dimension estimates from Lusztig [Lus84, 1.2], see also [Sho88, Theorem
+1.4]. In the Lie algebra gln the estimate becomes, for X in an adjoint orbit O,
+dim
+�
+gP ∈ GLn /P
+��g−1Xg ∈ σ + uP
+�
+≤ 1
+2
+�
+n2 − dim L − dim O
+�
+.
+(35)
+The proof is then standard in Springer theory. Let d > 0 and let x be such that
+dim c−1(x) ≥ d,
+the element x belongs to QOρ,σ for an element σ ∈ B and for some adjoint orbits
+Oρ1,σ1, . . . , Oρk,σk. The dimension estimate (35) implies
+d ≤ 1
+2
+�
+kn2 −
+k
+�
+j=1
+dim Lj − dim Oρj,σj
+�
+,
+so that
+k
+�
+j=1
+dim Oρj,σj ≤ kn2 −
+k
+�
+j=1
+dim Lj − 2d.
+Using the decomposition from Proposition 6.6, we have that x ∈ QB,M,ρ
+L,P
+.
+The
+previous inequality together with the expression (30) for the dimension of QB,M,ρ
+L,P
+give
+dim QB,M,ρ
+L,P
+≤ n2(2g − 2) + 2 + dim Z(m) ∩ B + kn2 −
+k
+�
+j=1
+dim Lj − 2d.
+(36)
+Moreover
+dim QB,M,ρ
+L,P
+×B/W B = dim QB,M,ρ
+L,P
+(37)
+and
+dim QL,P ×B/W B = dim QL,P = n2(2g − 2) + 2 + dim B + kn2 −
+k
+�
+j=1
+dim Lj. (38)
+Combining (36),(37) and (38),
+dim QB,M,ρ
+L,P
+×B/W B ≤ dim QL,P ×B/W B + 2d + dim Z(m) ∩ B − dim B.
+(39)
+As d is assumed to be strictly positive, necessarily the inclusion L ⊊ M is strict, so
+that
+dim Z(m) ∩ B < dim B.
+(40)
+35
+
+Now (39) and (40) provide the estimate
+dim QB,M,ρ
+L,P
+×B/W B < dim QL,P ×B/W B − 2d.
+(41)
+To conclude, the set
+�
+x ∈ QL,P ×B/W B |dim c−1(x) ≥ d
+�
+is a finite union of varieties
+QB,M,ρ
+L,P
+×B/W B with dimensions satisfying the previous estimate (41).
+Remark 6.9. Let us study the restriction of the W-equivariant sheaves Hiη!κ to the
+regular locus. Recall that Qreg
+L,P ×Breg/W Breg ∼= �Qreg
+L,P, then for σ ∈ Breg
+Hi
+ση!κ ∼= Hi
+c( �QL,P ,σ, κ).
+For w ∈ W, the W-equivariant structure is given by the functoriality of the compactly
+supported cohomology (see Proposition 3.6 and Remark 3.3)
+w∗ : Hi
+c
+�
+�QL,P ,w.σ, κ
+�
+→ Hi
+c
+�
+�QL,P ,σ, κ
+�
+.
+Therefore the construction of Theorem 6.7 gives a canonical extension over B of this
+natural Weyl group action over Breg.
+6.4
+Monodromic Weyl group action on the cohomology of
+�QL,P ,σ
+We saw in 5.12 that the cohomology sheaves Hiη!κ are constant sheaves over B.
+Together with the W-equivariant structure, this allows to construct a Weyl group
+action on the cohomology of the varieties �QL,P ,σ for any σ ∈ B, this is called the
+monodromic Weyl group action. Note that the fiber over σ of this constant sheaf is
+Hi
+c( �QL,P ,σ; κ). Thus for any σ, τ ∈ B, there is an isomorphism
+fσ,τ : Hi
+c( �QL,P ,σ; κ) → Hi
+c( �QL,P,τ; κ),
+such that for any ω ∈ B
+fσ,τ = fω,τ ◦ fσ,ω.
+The W-equivariance of the local system Hiη!κ implies the following theorem. It can
+also be proved directly, without referring to equivariance of the local system (see
+Maffei [Maf02, Section 5]).
+Theorem 6.10. Let σ, τ ∈ B, then the following diagram commutes
+Hi
+c( �QL,P ,σ; κ)
+Hi
+c( �QL,P,w−1.σ; κ)
+Hi
+c( �QL,P,τ; κ)
+Hi
+c( �QL,P,w−1.τ; κ).
+w∗
+fσ,τ
+fw−1.σ,w−1.τ
+w∗
+Remark 6.11. Note that if σ ∈ B is not regular, then the map
+w∗ : Hi
+c
+�
+�QL,P ,σ, κ
+�
+→ Hi
+c
+�
+�
+QL,P ,w−1.σ, κ
+�
+is only the map coming from the W-equivariant structure of the constant sheaf Hiη!κ.
+It does not necessarily come by functoriality from a morphism of variety.
+36
+
+This theorem allows to define a W-action on the compactly supported cohomol-
+ogy space Hi
+c( �QL,P ,σ; κ).
+Theorem 6.12. For σ ∈ B and for w ∈ W let
+ρi(w) = fw.σ,σ ◦ (w−1)∗.
+This defines an action of W on Hi
+c( �QL,P ,σ; κ), it is called the monodromic Weyl
+group action.
+Proof. For w1 and w2 in W, the following diagram commutes by Theorem 6.10.
+Hi
+c( �QL,P ,σ; κ)
+Hi
+c( �QL,P,w2.σ; κ)
+Hi
+c( �QL,P,w1w2.σ; κ)
+Hi
+c( �QL,P ,σ; κ)
+Hi
+c( �QL,P,w1.σ; κ)
+Hi
+c( �QL,P ,σ; κ)
+(w−1
+2
+)
+∗
+(w−1
+1
+)
+∗
+fw2.σ,σ
+fw1w2.σ,w1.σ
+(w−1
+1
+)
+∗
+fw1.σ,σ
+Going from the top left corner to bottom right corner by the top right corner is
+ρ(w1w2). Going by the middle gives ρ(w1)ρ(w2). Therefore ρ(w1w2) = ρ(w1)ρ(w2).
+6.5
+Frobenius morphism and monodromic action
+The techniques in this section come from Hausel, Letellier and Rodriguez-Villegas
+[HLRV13], however we do no consider regular semisimple values of the moment
+map. Instead each component of the moment map is central and each leg of the
+comet-shaped quiver corresponds to a particular adjoint orbit. Comet-shaped quiver
+varieties were also studied in this context by Letellier [Let12]. A slightly more general
+situation is considered here, as a leg can represents any adjoint orbit and not only
+a semisimple regular one.
+The representation defined in Theorem 6.12 when K = C is isomorphic to the
+representation obtained for K = Fq and large enough characteristic. Indeed this can
+be proved by base change exactly like in [HLRV13, Theorem 2.5]. Therefore from
+now on we assume:
+Assumption 6.13. K = Fq and the characteristic is large enough.
+This assumption is very convenient as it allows to introduce Frobenius endo-
+morphism and use Grothendiek’s trace formula to compute the traces of the action
+obtained.
+We denote by F the Frobenius endomorphism on gln raising the coefficients to
+the power q. The set of F-fixed points in gln is gln(Fq) and similarly for the group
+GLn. Assume that the Lj are subgroups of bock diagonal matrices, and that the
+P j are subgroups of block upper triangular matrices, then they are F-stable. The
+morphism F induces a Frobenius endomorphism on �Qreg
+L,P and on Breg also denoted
+by F,
+F
+�
+σ, (Ai, Bi)1≤i≤g , (Xj, gjLj)1≤j≤k
+�
+=
+�
+F(σ), (F(Ai), F(Bi))1≤i≤g , (F(Xj), F(gj)Lj)1≤j≤k
+�
+.
+37
+
+This Frobenius endomorphism can be twisted by an element w = (w1, . . . , wk) in
+the Weyl group W. For σ ∈ Breg, define
+wF(σ1, . . . , σk) := (w1.F(σ1), . . . , wk.F(σk)) .
+The set of points fixed by wF is (Breg)wF. Similarly, the w-twisted Frobenius on
+�Qreg
+L,P is
+wF := w ◦ F.
+They are compatible, preg ◦ wF = wF ◦ preg so that for σ, τ ∈ Breg the following
+diagram commutes
+Hi
+c( �QL,P,σ; κ)
+Hi
+c( �QL,P,F −1(σ); κ)
+Hi
+c( �QL,P,τ; κ)
+Hi
+c( �QL,P,F −1(τ); κ).
+F ∗
+fσ,τ
+fF −1(σ),F −1(τ)
+F ∗
+Theorem 6.14. For τ ∈ (Breg)F and for σ ∈ (Breg)wF, the cardinal of the set of
+wF fixed points in �QL,P ,σ is
+♯ �QwF
+L,P ,σ =
+�
+i
+tr
+�
+ρ2i(w), H2i
+c ( �QL,P,τ; κ)
+�
+qi.
+Proof. Consider the commutative diagram
+Hi
+c( �QL,P,τ; κ)
+Hi
+c( �QL,P,w−1.τ; κ)
+Hi
+c( �QL,P,τ; κ)
+Hi
+c( �QL,P,τ; κ)
+Hi
+c( �QL,P,σ; κ)
+Hi
+c( �QL,P,F (σ); κ)
+Hi
+c( �QL,P,σ; κ).
+w∗
+ρ(w−1)
+fw−1.τ,τ
+F ∗
+fσ,τ
+w∗
+fF (σ),τ
+F ∗
+fσ,τ
+Apply Grothendieck trace formula to wF,
+♯ �QwF
+L,P ,σ
+=
+�
+i
+(−1)i tr
+�
+(wF)∗, Hi
+c( �QL,P ,σ; κ)
+�
+=
+�
+i
+(−1)i tr
+�
+F ∗ ◦ ρi(w−1), Hi
+c( �QL,P,τ; κ)
+�
+.
+The varieties QL,P,τ are pure and polynomial count (see Remark 4.13 and [HLRV11,
+Theorem 1.3.1]) and ρ(w−1) commutes with F so that
+♯ �QwF
+L,P ,σ
+=
+�
+i
+tr
+�
+F ∗ ◦ ρ2i(w−1), H2i
+c ( �QL,P,τ; κ)
+�
+=
+�
+i
+tr
+�
+ρ2i(w−1), H2i
+c ( �QL,P,τ; κ)
+�
+qi.
+Now as W is isomorphic to a product of symmetric groups, w is conjugated to its
+inverse w−1 and
+♯ �QwF
+L,P ,σ =
+�
+i
+tr
+�
+ρ2i(w), H2i
+c ( �QL,P,τ; κ)
+�
+qi.
+38
+
+Notations 6.15. The j-th part of L is
+Lj ∼= GLcj
+1 × · · · × GLcj
+1
+�
+��
+�
+mj
+1
+× · · · × GLcj
+lj × · · · × GLcj
+lj
+�
+��
+�
+mj
+lj
+,
+with cj
+r ̸= cj
+s for r ̸= s. Then the j-th part of the relative Weyl group W is
+WGLn(Lj) ∼= Smj
+1 × · · · × Smj
+lj
+The symmetric group Smj
+r acts by permuting the blocks of size cj
+r.
+Take w =
+(w1, . . . , wk) in W and choose σw = (σ1, . . . , σk) in (Breg)wF. The conjugacy class
+of the element wj is determined by a lj-tuple (ηj,1, . . . , ηj,lj) with ηj,r ∈ Pmj
+r. Let Oσj
+be the adjoint orbit of σj. This orbit is semisimple, F-stable and of the following
+type (as defined in 4.3),
+�
+ηj,1
+1 , 1cj
+1
+�
+. . .
+�
+ηj,1
+l(ηj,1), 1cj
+1
+�
+. . .
+�
+η
+j,lj
+1 , 1
+cj
+lj
+�
+. . .
+�
+η
+j,lj
+l(ηj,lj ), 1
+cj
+lj
+�
+.
+Define Ow := (Oσ1, . . . , Oσk).
+Lemma 6.16. With the previous notations we have the following identity between
+cardinals,
+♯ �QwF
+L,P ,σ = ♯QF
+Ow.
+(42)
+Proof. As the orbits Ow = (Oσ1, . . . , Oσk) are semisimple (hence they are closed),
+the map �QL,P ,σ → QOw is an isomorphism compatible with the Frobenius wF on
+the source and the Frobenius F on the target.
+Letellier [Let11] computed the number of points of comet shaped quiver varieties,
+in particular of QF
+Ow.
+Theorem 6.17. With Notations 6.15, the cardinal of QF
+Ow is given by
+♯QF
+Ow = (−1)r(η)q
+dµ
+2
+�
+�hη, HHLV
+n
+(0, q
+1
+2)
+�
+,
+where �hη is a particular case of the generalized Schur function from [Let11]. This
+symmetric function can be expressed in terms of complete symmetric functions hn,
+�hη :=
+k
+�
+j=1
+lj
+�
+r=1
+l(ηj,i)
+�
+s=1
+hcj
+r
+�
+Xηj,r
+s
+j
+�
+,
+and
+r(η) :=
+k
+�
+j=1
+lj
+�
+r=1
+cj
+r
+l(ηj,i)
+�
+s=1
+(ηj,r
+s − 1).
+Proof. As the orbits Oσj are semisimple, the variety QOw is smooth so that the
+characteristic function of the intersection complex is constant with value 1. The
+result follows from Letellier [Let11, Theorem 6.9.1, Theorem 7.4.1 and Corollary
+7.4.3].
+39
+
+Corollary 6.18. For σ ∈ B and η representing a conjugacy class in the Weyl group
+as described in Notations 6.15, the η-twisted Poincaré polynomial of �QL,P ,σ is
+�
+i
+tr
+�
+η, Hi
+c( �QL,P ,σ, κ)
+�
+vi = (−1)r(η)vdµ �
+�hη, HHLV
+n
+(0, v)
+�
+.
+Proof. The action comes from the W-equivariant structure of the constant sheaves
+Hiη!κ. Therefore, up to isomorphism, the representation does not depend on the
+choice of σ ∈ B so that the twisted Poincaré polynomial can be computed for
+τ ∈ (Breg)F. Then from Theorem 6.14 and from (42),
+�
+i
+tr
+�
+ρ2i(η), H2i
+c
+�
+�QL,P ,τ, κ
+��
+qi = (−1)r(η)q
+dµ
+2
+�
+�hη, HHLV
+n
+(0, q
+1
+2)
+�
+.
+This equality remains true after substituting qn for q for n > 0. Thus it is an equality
+between two polynomials and the corollary is proved.
+Remark 6.19 (Comparison between monodromic and Springer action). Let σ =
+(σ1, . . . , σk) ∈ B, as before Mj is the stabilizer of σj in GLn. The relative Weyl
+group is
+WM(L) =
+k
+�
+j=1
+WMj(Lj)
+with WMj(Lj) = NMj(Lj)/Lj. Then WM (L) is a subgroup of the Weyl group W
+studied in this section. The group WM (L) is exactly the subgroup of elements w ∈ W
+such that w.σ = σ. The monodromic Weyl group action from Theorem 6.12 induces
+an action of WM (L) on Hi
+c
+�
+�QL,P ,σ, κ
+�
+. Interestingly, this action comes only from
+the W-equivariant structure, it does not rely on the constant property of the sheaf:
+it is given explicitly by ρi(w) = (w−1)∗.
+There is another action of WM (L) on Hi
+c
+�
+�QL,P ,σ, κ
+�
+, the Springer action con-
+structed by Letellier and mentioned in 4.14. Letellier computed the twisted Poincaré
+polynomial for this Springer action [Let11, Corollary 7.4.3], it coincides with the
+Poincaré polynomial obtained from the monodromic action, therefore both action are
+isomorphic. It would be interesting to have a direct proof of this fact. We proved it
+in the character variety setting for just one orbit, regular, with a unique eigenvalue
+[Bal21, Chapter 5].
+It is also interesting to consider the monodromic action over the regular locus
+Breg as an action on the cohomology of a quiver variety with semisimple adjoint
+orbits. For σ ∈ Breg consider the associated generic k-tuple of semisimple adjoint
+orbits Oσ = (Oσ1, . . . , Oσk). The Weyl group WGLn(Lj) is the group of permutation
+of the distinct eigenvalues of Oσj with the same multiplicities. This provides another
+formulation of Corollary 6.18.
+Corollary 6.20. For η representing a conjugacy class in the Weyl group as described
+in Notations 6.15, the η-twisted Poincaré polynomial of QOσ is
+�
+i
+tr
+�
+η, Hi
+c(QOσ, κ)
+�
+vi = (−1)r(η)vdη �
+�hη′, HHLV
+n
+(0, v)
+�
+.
+40
+
+The interpretation over the regular locus in terms of semisimple quiver varieties
+together with Remark 6.19 show the advantages of extending the W-equivariant
+structure from Breg to B (Theorem 6.10). This provides a uniform description of the
+Springer action on the cohomology of some resolution �QL,P ,σ and the monodromic
+action on the cohomology of semisimple quiver varieties QOσ.
+Remark 6.21. It is also interesting to study the action of a Weyl group relative to a
+particuar leg 1 ≤ j ≤ k, for instance relative to the first one. This will be used in 7.2
+to describe some structure coefficients of the algebra spanned by Kostka polynomial.
+A particularly interesting case is when L1 is a maximal torus and M1 = GLn. Then
+the component of the Weyl group relative to the first leg is WM1(L1) ∼= Sn and
+WM(L) ∼= Sn ×
+k
+�
+j=2
+WMj(Lj).
+According to this decomposition, consider an element (w, 1, . . . , 1) ∈ WM(L) with
+w ∈ Sn an element of cycle type λ ∈ Pn. Then
+�hη = pλ[X1]hµ′2[X2] . . . hµ′k[Xk],
+and (−1)r(η) = ǫ(λ) is the sign of the permutation w with cycle type λ. Corollary
+6.18 reads
+P η
+c
+�
+�QL,P ,σ, v
+�
+= vdµǫ(λ)
+�
+pλ[X1]hµ′2[X2] . . . hµ′k[Xk], HHLV
+n
+(0, v)
+�
+.
+This can be understood in terms of Frobenius characteristic, see Definition 2.12.
+Consider the representation of Sn on the cohomology of �QL,P ,σ twisted by the sign,
+H•( �QL,P ,σ, κ) ⊗ ǫ. Its v-graded Frobenius characteristic is given by the following
+symmetric function in X1
+vdµ �
+hµ′2[X2] . . . hµ′k[Xk], HHLV
+n
+(0, v)
+�
+X2,...,Xk
+.
+Notice that Vρ ⊗ ǫ ∼= Vρ′, then by Remark 2.11, the multiplicity of the irreducible
+representation Vρ in H•( �QL,P ,σ, κ) is given by
+vdµ �
+sρ′[X1]hµ′2[X2] . . . hµ′k[Xk], HHLV
+n
+(0, v)
+�
+.
+7
+Geometric interpretations in the algebra spanned
+by Kostka polynomials
+7.1
+Description of the algebra
+In this section an algebra spanned by Kostka polynomials is studied and some struc-
+ture coefficients are related to traces of Weyl group action on the cohomology of
+quiver varieties. Define a linear map ∆# : Sym[X] → Sym[X, Y ] such that on the
+basis of modified Macdonald polynomials,
+∆# �
+˜Hλ[X]
+�
+:= ˜Hλ[X] ˜Hλ[Y ] for all λ ∈ P.
+41
+
+As in 2.20, the variables (q, t) are implicit. Now as the Hall pairing is non-degenerate,
+there is a uniquely determined bilinear map . . . # . . . such that for all F, G and H
+in Sym[X],
+⟨F[X]#G[X], H[X]⟩ =
+�
+F[X]G[Y ], ∆# (H[X])
+�
+.
+The product # defines an associative and commutative algebra structure on Sym[X].
+Definition 7.1. For a k-tuple of partitions µ =
+�
+µ1, . . . , µk�
+∈ Pk
+n and for λ ∈ Pn
+we denote by cλ
+µ the structure coefficients of the product # in the basis of Schur
+functions
+sµ1#sµ2 . . . #sµk =
+�
+|µ|=n
+cλ
+µsλ.
+(43)
+Lemma 7.2. For µ = (µ, ν), the coefficients cλ
+µ,ν coincide with those defined in the
+introduction, i.e., the following relation is satisfied
+�Kµ,ρ �Kν,ρ =
+�
+λ
+cλ
+µ,ν �Kλ,ρ.
+(44)
+Proof. First let
+�
+�Lη,λ
+�
+λ,η∈Pn be the inverse of the matrix of Kostka polynomials
+�
+�Kη,λ
+�
+λ,η∈Pn (see Definition 2.19), then
+sλ =
+�
+η∈Pn
+�Lη,λ ˜Hη[X].
+Now the coefficient cλ
+µ,ν is defined by
+cλ
+µ,ν
+=
+⟨sµ#sν, sλ⟩
+=
+�
+sµ#sν,
+�
+η∈Pn
+�Lη,λ ˜Hη[X]
+�
+.
+By definition of the product # and of the coproduct ∆#,
+cλ
+µ,ν
+=
+�
+η∈Pn
+�Lη,λ
+�
+sµ[X]sν[Y ], ˜Hη[X] ˜Hη[Y ]
+�
+,
+cλ
+µ,ν
+=
+�
+η∈Pn
+�Lη,λ �Kµ,η �Kν,η.
+Multiply the last equation by �Kλ,ρ and sum over λ ∈ Pn,
+�Kµ,ρ �Kν,ρ =
+�
+λ
+cλ
+µ,ν �Kλ,ρ.
+This last relation is exactly the one used in the introduction to define the coefficients
+cλ
+µ,ν.
+Example 7.3. We computed some coefficients with the software SageMath
+c(2,1,1)
+(2,2),(2,1,1)
+=
+−q3t − q2t2 − qt3 − q2t − t2q + q2 + qt + t2,
+c(1,1,1,1)
+(2,2),(2,1,1)
+=
+q3 + q2t + qt2 + t3 + q2 + 2qt + t2 + q + t.
+42
+
+The next conjecture comes from unpublished notes by Rodriguez-Villegas.
+Conjecture 7.4. The structure coefficients cλ
+µ lie in Z[q, t].
+Some evidences supporting this conjecture will be provided. The following defi-
+nition and remark were suggested by François Bergeron.
+Definition 7.5. Let F be a symmetric function, consider the operator
+F# . . .
+:
+Sym [X]
+→
+Sym [X]
+G
+�→
+F#G.
+We denote ψF its adjoint with respect to the Hall pairing so that for any G, H ∈
+Sym [X]
+⟨F#G, H⟩ = ⟨G, ψF(H)⟩
+(45)
+Those operators are diagonal in the basis of modified Macdonald polynomials
+ψF( ˜Hλ[X; q, t]) =
+�
+F, ˜Hλ[X; q, t]
+�
+˜Hλ[X; q, t].
+(46)
+Remark 7.6. Applying the relation (46) with en,
+ψen
+�
+˜Hλ[X; q, t]
+�
+= qn(λ′)tn(λ) ˜Hλ[X; q, t],
+and we recognize the usual expression of the operator ∇ introduced by Bergeron-
+Garsia [BG98]. The higher (q, t)-Catalan sequence from Garsia–Haiman [GH96]
+(see also Haiman [Hai02, p.95]) is defined by
+C(m)
+n
+(q, t) = ⟨en, ∇men⟩ ,
+but ∇ = ψen is the adjoint of en# . . . and s1n = en so that
+C(m)
+n
+(q, t) = c1n
+1n, . . . , 1n
+�
+��
+�
+m+1
+.
+The higher (q, t)-Catalan sequences are particular cases of the coefficients c1n
+µ .
+We recall an important theorem which was first conjectured by Garsia–Haiman
+[GH96].
+Theorem 7.7 ([Hai02] theorem 4.2.5). The symmetric function ∇(en) is obtained
+as the Frobenius characteristic (see definition 2.12) of a bigraded representation of
+Sn called the diagonal harmonics. In particular,
+⟨∇(en), sλ⟩ ∈ Z≥0[q, t].
+Corollary 7.8. For any µ ∈ Pn, the structure coefficients c1n
+1n,µ gives the multiplicity
+of the irreducible representation of type µ in the bigraded representation of Sn on
+diagonal harmonics. In particular c1n
+1n,µ(q, t) ∈ Z≥0[q, t] so that the conjecture 7.4 is
+true for those particular coefficients.
+43
+
+Proof. According to Remark 7.6 and to the adjonction relation (45),
+⟨sµ, ∇(en)⟩ = ⟨en#sµ, en⟩ .
+(47)
+By definition of the structure coefficients cλ
+µ,ν and as en = s1n, we have
+en#sµ =
+�
+λ∈Pn
+cλ
+1nµsλ,
+substituting in (47) we obtain
+c1n
+1n,µ(q, t) = ⟨sµ, ∇(en)⟩ .
+We conclude with the interpretation of ∇(en) as a Frobenius characteristic from
+Theorem 7.7.
+The next theorem and the following corollary come from unpublished notes by
+Rodriguez-Villegas. They relate particular structure coefficients c1n
+µ to the kernel
+HHLV
+n
+.
+Consider the generating function from Definition 4.16 for genus g = 0, k + 2
+punctures and with variable z = q
+1
+2, w = t
+1
+2. It is given by
+Ω0
+k+2 :=
+�
+λ∈P
+�k+2
+i=1 ˜Hλ [Xi; q, t]
+aλ(q, t)
+s|λ|,
+with aλ(q, t) =
+�
+˜Hλ[X; q, t], ˜Hλ[X; q, t]
+�q,t
+as in 2.18.
+Theorem 7.9. The following relation holds,
+�
+p(n)[Xk+1]h(n−1,1)[Xk+2], Log
+�
+Ωg
+k+2
+��
+Xk+1,Xk+2 =
+�
+|λ|=n
+φλΠ′
+λ
+aλ
+k
+�
+i=1
+˜Hλ[Xi]s|λ|,
+with
+φλ
+=
+�
+i,j∈λ
+qj−1ti−1,
+Π′
+λ
+=
+�
+i,j∈λ\(1,1)
+(1 − qj−1ti−1).
+Proof. According to Lemma 2.23, to take the Hall pairing with h(n−1,1)[Xk+2] is
+equivalent to do plethystic substitution Xk+2 = 1 + u and to take the degree n
+coefficient in front of u. As the plethystic substitution and the plethystic logarithm
+commute, we can perform this substitution inside the plethystic logarithm.
+We
+consider terms of order 1 in u using Lemma 2.22,
+Log
+�
+Ω0
+k+2
+�
+= Log
+�
+Ω0
+k+1 + u
+�
+λ∈P∗
+φλ
+aλ
+k+1
+�
+i=1
+˜Hλ[Xi]s|λ| + O(u2)
+�
+= Log
+�
+Ω0
+k+1
+�
+1 + u
+1
+Ω0
+k+1
+�
+λ∈P∗
+φλ
+aλ
+k+1
+�
+i=1
+˜Hλ[Xi]s|λ| + O(u2)
+��
+= Log
+�
+Ω0
+k+1
+�
++ Log
+�
+1 + u
+1
+Ω0
+k+1
+�
+λ∈P∗
+φλ
+aλ
+k+1
+�
+i=1
+˜Hλ[Xi]s|λ| + O(u2)
+�
+.
+44
+
+We used that plethystic logarithm turns products into sums. From the definition of
+the plethystic logarithm, as pn[u] = un, we easily see the coefficient in front of u in
+the previous expression
+Log
+�
+Ω0
+k+2
+���
+u =
+1
+Ω0
+k+1
+�
+λ∈P∗
+φλ
+aλ
+k+1
+�
+i=1
+˜Hλ[Xi]s|λ|.
+Keeping the terms of degree n we obtain
+�
+h(n−1,1)[Xk+2], Log
+�
+Ω0
+k+2
+��
+Xk+2 =
+1
+Ω0
+k+1
+�
+λ∈P∗
+φλ
+aλ
+k+1
+�
+i=1
+˜Hλ[Xi]s|λ|
+�����
+sn
+.
+Inverting Ω0
+k+1 is licit, it is defined by
+1
+Ω0
+k+1
+=
+1
+1 +
+�
+Ω0
+k+1 − 1
+� =
+�
+k
+�
+1 − Ω0
+k+1
+�k .
+Now we just have to take the Hall pairing with the power sum p(n) [Xk+1]. This is
+equivalent to picking the coefficient in front of n−1p(n) [Xk+1]. But p(n) cannot be
+written as the product of two symmetric functions of degree strictly smaller than n
+so that the contribution of Ω0
+k+1 in the denominator is irrelevant for the coefficient
+in front of n−1p(n) [Xk+1] and
+�
+p(n)[Xk+1]h(n−1,1)[Xk+2], Log
+�
+Ω0
+k+2
+��
+Xk+1,Xk+2 =
+�
+p(n)[Xk+1],
+�
+λ∈P∗
+φλ
+aλ
+k+1
+�
+i=1
+˜Hλ[Xi]s|λ|
+�
+Xk+1
+.
+We conclude with Lemma 2.24 and (4).
+The following corollary allows to obtain a geometric interpretation of the coeffi-
+cients. Indeed, it relates the coefficient c(1n)
+µ
+to the generating serie Ω0
+k+2 known to
+encode cohomological information about comet-shaped quiver varieties and charac-
+ter varieties.
+Corollary 7.10. With the notations of the previous theorem and Definition 7.1,
+(−1)n−1c(1n)
+µ
+= (q−1)(1−t)
+� k
+�
+j=1
+sµj[Xj]p(n)[Xk+1]h(n−1,1)[Xk+2], Log
+�
+Ω0
+k+2
+�
+�
+X1,...,Xk+2
+.
+(48)
+Proof. We apply Theorem 7.9 to express the right hand side of (48) as
+(q − 1)(1 − t)
+�
+sµ1[X1] . . . sµk[Xk],
+�
+|λ|=n
+φλΠ′
+λ
+aλ
+k
+�
+i=1
+˜Hλ[Xi]
+�
+X1,...,Xk
+.
+By definition of the product #,
+(q − 1)(1 − t)
+�
+sµ1# . . . #sµk[X],
+�
+|λ|=n
+φλΠ′
+aλ
+˜Hλ[X]
+�
+X
+.
+45
+
+Here we recognize the expression of Theorem 2.25
+�
+sµ1# . . . #sµk[X], (−1)n−1s(1n)
+�
+X
+so that if we write
+sµ1# . . . #sµk[X] =
+�
+λ
+cλ
+µsλ[X]
+the result follows from orthonormality of Schur functions.
+7.2
+Interpretation of certain coefficients as traces of Weyl
+group actions on the intersection cohomology of quiver
+varieties
+In this section a cohomological interpretation is given for the coefficients c1n
+µ . In
+order to lighten the notations, the description is only given for the coefficient c1n
+µ,ν.
+The generalization to any µ is straightforward.
+First let us detail the data to describe the relevant variety �QL,P ,σ. The Levi sub-
+groups are torus of diagonal matrices Lj = T for 1 ≤ j ≤ 3 and L4 = GL1 × GLn−1.
+The semisimple part σ = (σ1, . . . , σ4) is such that:
+• σ1 = ζ1 Id is central,
+• σ2 = ζ2 Id is central,
+• σ3 =
+
+
+
+
+
+α1
+α2
+...
+αn
+
+
+
+
+ with αr ̸= αs for r ̸= s,
+• σ4 =
+
+
+
+
+
+β
+γ
+...
+γ
+
+
+
+
+ has two eigenvalues β ̸= γ. The multiplicity of β is one
+and the multiplicity of γ is n − 1.
+Notice that such a choice can be made in the generic locus, i.e., with σ ∈ B.
+First we consider Letellier’s construction of the action à la Springer in order to
+compute isotypical components. Let M = M1 × · · · × M4 with Mj the stabilizer in
+GLn of σj. Then WM(L) ∼= S2
+n. Letellier’s construction (Remark 4.14) provides an
+action of WM(L) on the cohomology of �QL,P ,σ. Moreover for Vµ′, respectively Vν′,
+the irreducible representation of Sn associated to the transpose of some partition
+µ, respectively ν,
+HomWM(L)
+�
+Vµ′ ⊗ Vν′, H
+i+d �
+QL,P ,σ
+c
+�
+�QL,P ,σ, κ
+��
+= H
+i+dQO
+c
+(QO, κ) .
+(49)
+With O = (O1, . . . , O4) the 4-tuple of generic adjoint orbits defined by,
+• O1 has Jordan type µ′ and eigenvalue ζ1,
+• O2 has Jordan type ν′ and eigenvalue ζ2,
+46
+
+• O3 is the orbit of σ3,
+• O4 is the orbit of σ4.
+Now with the construction from Theorem 6.12, there is an action of the whole
+group W ∼= S3
+n on the cohomology of �QL,P ,σ. The restriction of this W-action to
+WM(L) ∼= S2
+n is isomorphic to the Springer action (see Remark 6.19). First take
+the Vµ′ ⊗ Vν′ isotypical component with respect to the S2
+n-action. There remains an
+action of the Weyl group Sn relative to the third leg on the intersection cohomology
+IHi
+c (QO, κ).
+Theorem 7.11. Let w be a n-cycle in the Weyl group relative to the third leg (this
+terminology comes from the comet-shaped quiver). The coefficient c1n
+µ,ν, after special-
+ization q = 0, is given by the w-twisted Poincaré polynomial of QO, namely
+c1n
+µ,ν(0, t) = t−
+dO
+2
+�
+i
+tr
+�
+w, IH2i
+c (QO, κ)
+�
+ti.
+Proof. Combining (49), Corollary 6.18 and Remark 6.21,
+�
+i
+tr
+�
+w, IHi
+c (QO, κ)
+�
+vi = (−1)n−1vdO �
+sµ[X1]sν[X2]p(n)[X3]h(n−1,1)[X4], HHLV
+n
+(0, v)
+�
+.
+The theorem now follows from Corollary 7.10.
+7.3
+Cohomological interpretation in the multiplicative case
+There are similar interpretations in the multiplicative case. A conjectural one in-
+volving c1n
+µ,ν(q, t) which is a theorem after specializing to c1n
+µ,ν(1, t). Unfortunately in
+the multiplicative case the monodromic action is not defined in the general case so
+that we have to rely only on the Springer action. Therefore the statements involve
+partial resolutions of character varieties instead of actual character varieties.
+First introduce the relevant parameters. The Levi subgroups are torus of diagonal
+matrices Lj = T for 1 ≤ j ≤ 3 and L4 = GL1 × GLn−1.
+The semisimple part
+σ = (σ1, . . . , σ4) is such that:
+• σ1 = ζ1 Id is central,
+• σ2 = ζ2 Id is central,
+• σ3 = ζ3 Id is central,
+• σ4 =
+
+
+
+
+
+β
+γ
+...
+γ
+
+
+
+
+ has two eigenvalues β ̸= γ.
+This 4-tuple is chosen to be generic (in the multiplicative sense of Definition 4.10).
+This is the case for instance if ζ1ζ2ζ3 = 1 and γn−1 = β−1 ̸= 1. The relative Weyl
+group is WM(L) ∼= S3
+n. Now consider the following conjugacy classes
+• C1 has Jordan type µ′ and eigenvalue ζ1,
+47
+
+• C2 has Jordan type ν′ and eigenvalue ζ2,
+• C3 has one Jordan block of size n with eigenvalue ζ3,
+• C4 is the conjugacy class of σ4.
+Then �
+ML,P ,σ is the resolution of MC with C = (C1, . . . , C4) (see Definition 4.15).
+An intermediate in between �
+ML,P ,σ and MC is given by the variety
+Mµ,ν =
+�
+(X1, . . . , X4) ∈ C1 × · · · × C4, gB ∈ GLn /B
+��g−1X3g ∈ ζ3U
+X1 . . . X4 = Id} // PGLn,
+with B the Borel subgroup of upper triangular matrices in GLn and U its unipotent
+radical.
+Then the resolution �
+ML,P ,σ → MC factors through Mµ,ν.
+This is a
+particular case of the partial resolutions of character varieties studied by Letellier
+[Let13]. The result we recalled about Springer theory for resolutions of character
+varieties (16) admit a more general version for partial resolutions.
+In particular
+considering the action of S2
+n with respect to the first two punctures and taking the
+Vµ′ ⊗ Vν′ isotypical component of the cohomology H•
+c
+�
+�
+ML,P ,σ, κ
+�
+we obtain,
+HomWM(L)
+�
+Vµ′ ⊗ Vν′, H
+i+d �
+ML,P ,σ
+c
+�
+�
+ML,P ,σ, κ
+��
+= H
+i+dMµ,ν
+c
+(Mµ,ν, κ) .
+There remains an action of the Weyl group Sn relative to the third puncture. For
+w in this Weyl group define the w-twisted mixed Hodge polynomial by
+IHw
+c (Mµ,ν; u, v) :=
+�
+i,r
+urvi tr
+�
+w, IHr,r,i
+c
+(Mµ,ν, κ)
+�
+.
+Conjecture 4.18 admits a generalization describing the Weyl group action on the
+intersection cohomology of partial resolutions of character varieties (Letellier [Let13,
+Conjecture 5.5]). In particular this conjecture predicts the following formula for the
+w-twisted mixed Hodge polynomial for w a n-cycle,
+IHw
+c (Mµ,ν; u, v) =
+(−1)n−1 �
+v√u
+�dim Mµ,ν
+�
+sµ[X1]sν[X2]p(n)[X3]h(n−1,1)[X4], HHLV
+n
+� −1
+√u, v√u
+��
+.
+The next conjecture follows from this conjectural formula, just like Theorem 7.11
+is deduced from Corollary 6.18.
+Conjecture 7.12. Let w be a n-cycle in the Weyl group relative to the third punc-
+ture. The coefficient c1n
+µ,ν relates to the w-twisted mixed Hodge polynomial of Mµ′,ν′
+by
+c1n
+µ,ν(q, t) = t
+− dim Mµ,ν
+2
+IHw
+c
+�
+Mµ,ν, 1
+q, √qt
+�
+.
+We will prove that the right handside of this conjecture is indeed a polynomial
+in q, t thus supporting Conjecture 7.4.
+48
+
+Lemma 7.13. The Poincaré polynomial (for compactly supported intersection co-
+homology) of the variety Mµ,ν is
+�
+i
+vi dim IHi
+c (Mµ,ν, κ) = vd �
+sµ[X1]sν[X2]h1n[X3]h(n−1,1)[X4], HHLV
+n
+(−1, v)
+�
+,
+(50)
+and for w a n-cycle in the Weyl group relative to the third puncture the w-twisted
+Poincaré polynomial is
+�
+i
+vi dim tr
+�
+w, IHi
+c (Mµ,ν, κ)
+�
+= vd �
+sµ[X1]sν[X2]p(n)[X3]h(n−1,1)[X4], HHLV
+n
+(−1, v)
+�
+.
+Proof. Letellier [Let13, Theorem 5.4] studied Springer theory for partial resolutions
+of character varieties. This allows to describe the intersection cohomology of the
+partial resolutions, together with its Weyl group action, in terms of intersection
+cohomology of character varieties. The Poincaré polynomials for intersection coho-
+mology of character varieties are computed in [Bal22]. The formula for (twisted)
+Poincaré polynomials of the resolutions are therefore a consequence of the Poincaré
+polynomial specialization of [Let13, Proposition 5.7].
+Proposition 7.14. The following expression
+t
+− dim Mµ,ν
+2
+IHw
+c
+�
+Mµ,ν, 1
+q, √qt
+�
+,
+which is the value of c1n
+µ,ν(q, t) according to Conjecture 7.12, is a polynomial in q, t
+with integer coefficients.
+Proof. First note that only integer powers of q and t appear because of (50) and the
+fact that in genus g = 0 the kernel HHLV
+n
+(z, w) contains only terms in z2, w2.
+Let S be a regular semisimple conjugacy class such that the 4-tuple C′ =
+(C1, C2, S, C4) is generic. By [Bal22] and (50), the Poincaré polynomial of the char-
+acter variety MC
+′ is the same as the Poincaré polynomial of the variety Mµ,ν. But
+the variety MC
+′ is affine, hence its compactly supported intersection cohomology
+vanishes in degree strictly smaller than its dimension and only positive power of t
+appear in the expression. Only positive power of q appear because the weight on
+the compactly supported intersection cohomology is smaller than the cohomological
+degree.
+The second equality in Lemma 7.13 implies the Poincaré polynomial specialisa-
+tion of Conjecture 7.12.
+Theorem 7.15. Let w be a n-cycle in the Weyl group relative to the third puncture.
+The coefficient c1n
+µ,ν relates to the w-twisted Poincaré polynomial of Mµ,ν:
+c1n
+µ,ν(1, t) = t
+− dim Mµ,ν
+2
+�
+i
+t
+i
+2 tr
+�
+w, IHi
+c (Mµ,ν, κ)
+�
+.
+References
+[Bal20]
+Mathieu Ballandras. Trivializations of moment maps. arXiv.2010.08294,
+2020.
+49
+
+[Bal21]
+Mathieu Ballandras. Weyl group actions on the cohomology of character
+and quiver varieties. PhD thesis, Université de Paris – Scuola Inter-
+nazionale Superiore di Studi Avanzati, 2021.
+[Bal22]
+Mathieu Ballandras. Intersection cohomology of character varieties for
+punctured riemann surfaces. arXiv.2201.08795, 2022.
+[BBDG18] A. A. Beilinson, J. Bernstein, P. Deligne, and O. Gabber. Faisceaux
+pervers, volume 100 of Astérisque.
+Société mathématique de France,
+2018.
+[BG98]
+F. Bergeron and G. Garsia. Science fiction and Macdonald’s polynomials.
+CRM Proc. Lecture Notes, 22, 10 1998.
+[BM83]
+W. Borho and R. MacPherson.
+Partial resolutions of nilpotent vari-
+eties, Analysis and topology on singular spaces, II, III, volume 101 of
+Astérisque. Société mathématique de France, 1983.
+[CB01]
+William Crawley-Boevey. Geometry of the moment map for representa-
+tions of quivers. Compositio Mathematica, 126(3):257–293, 2001.
+[CB03a]
+William Crawley-Boevey.
+Indecomposable parabolic bundles and the
+existence of matrices in prescribed conjugacy class closures with product
+equal to the identity. Publications mathématiques, 100, 08 2003.
+[CB03b]
+William Crawley-Boevey. On matrices in prescribed conjugacy classes
+with no common invariant subspace and sum zero.
+Duke Math. J.,
+118(2):339–352, 06 2003.
+[CB06]
+William Crawley-Boevey.
+Quiver algebras, weighted projective lines,
+and the Deligne-Simpson problem.
+International Congress of Mathe-
+maticians, ICM 2006, 2, 05 2006.
+[GH93]
+A. M. Garsia and M. Haiman. A graded representation model for Mac-
+donald’s polynomials. Proceedings of the National Academy of Sciences,
+90(8):3607–3610, 1993.
+[GH96]
+A.M. Garsia and M. Haiman. A remarkable q,t-catalan sequence and
+q-lagrange inversion.
+Journal of Algebraic Combinatorics, 5:191–244,
+1996.
+[Hai01]
+Mark Haiman. Hilbert schemes, polygraphs and the Macdonald positiv-
+ity conjecture. Journal of the American mathematical society, 14(4):941–
+1006, 2001.
+[Hai02]
+Mark Haiman.
+Combinatorics, symmetric functions, and Hilbert
+schemes. Current Developments in Mathematics, 2002:39–111, 2002.
+[HLRV11] Tamás Hausel, Emmanuel Letellier, and Fernando Rodriguez-Villegas.
+Arithmetic harmonic analysis on character and quiver varieties. Duke
+Math. J., 160(2):323–400, 11 2011.
+50
+
+[HLRV13] Tamás Hausel, Emmanuel Letellier, and Fernando Rodriguez-Villegas.
+Positivity for Kac polynomials and DT-invariants of quivers. Ann. of
+Math., 177(3):1147–1168, 2013.
+[Kos04]
+Vladimir Petrov Kostov. The Deligne–Simpson problem—a survey. Jour-
+nal of Algebra, 281(1):83–108, 2004.
+[KP81]
+Hanspeter Kraft and Claudio Procesi.
+Minimal singularities in GLn.
+Inventiones mathematicae, 62:503–515, 1980/81.
+[KW01]
+Reinhardt Kiehl and Rainer Weissauer.
+Weil Conjectures, Perverse
+Sheaves and l’adic Fourier Transform. Springer, 01 2001.
+[Let05]
+Emmanuel Letellier. Fourier Transforms of Invariant Functions on Fi-
+nite Reductive Lie Algebras. Lecture Notes in Mathematics. Springer
+Berlin Heidelberg, Berlin, Heidelberg, 2005.
+[Let11]
+Emmanuel Letellier. Quiver varieties and the character ring of general
+linear groups over finite fields. Journal of the European Mathematical
+Society, 15, 03 2011.
+[Let12]
+Emmanuel Letellier. Tensor products of unipotent characters of general
+linear groups over finite fields. Transformation Groups, 18, 04 2012.
+[Let13]
+Emmanuel Letellier. Character varieties with Zariski closures of GLn
+conjugacy classes at punctures. Selecta Mathematica, 21, 09 2013.
+[LL19]
+Gérard Laumon and Emmanuel Letellier.
+Notes on a conjecture of
+Braverman-Kazhdan, 2019.
+[Lus84]
+G. Lusztig. Intersection cohomology complexes on a reductive group.
+Inventiones mathematicae, 75:205–272, 1984.
+[Lus85]
+G. Lusztig. Character sheaves I. Advances in Mathematics, 56(3):193–
+237, June 1985.
+[Lus86]
+G. Lusztig. On the character values of finite Chevalley groups at unipo-
+tent elements. J. Algebra, 104:146–194, 1986.
+[Lus00]
+G. Lusztig.
+Quiver varieties and Weyl group actions.
+Annales de
+l’Institut Fourier, 50(2):461–489, 2000.
+[Mac88]
+I.G. Macdonald.
+A new class of symmetric functions.
+Séminaire
+Lotharingien de Combinatoire [electronic only], 20:B20a, 41 p.–B20a,
+41 p., 1988.
+[Mac15]
+I. G. Macdonald. Symmetric functions and Hall polynomials. Oxford
+Classic Texts in the Physical Sciences. The Clarendon Press, Oxford
+University Press, New York, second edition, 2015. With contribution by
+A. V. Zelevinsky and a foreword by Richard Stanley, Reprint of the 2008
+paperback edition [ MR1354144].
+51
+
+[Maf02]
+Andrea Maffei. A remark on quiver varieties and Weyl groups. An-
+nali della Scuola Normale Superiore di Pisa - Classe di Scienze, Ser. 5,
+1(3):649–686, 2002.
+[Mel18]
+Anton Mellit.
+Integrality of Hausel–Letellier–Villegas kernels.
+Duke
+Math. J., 167(17):3171–3205, 11 2018.
+[Mel19]
+Anton Mellit. Cell decompositions of character varieties, 2019.
+[Mel20]
+A. Mellit. Poincaré polynomials of character varieties, Macdonald poly-
+nomials and affine Springer fibers. Annals of Mathematics, 192(1):165 –
+228, 2020.
+[Nak94]
+Hiraku Nakajima. Instantons on ALE spaces, quiver varieties, and Kac-
+Moody algebras. Duke Math. J., 76(2):365–416, 11 1994.
+[Nak98]
+Hiraku Nakajima. Quiver varieties and Kac-Moody algebras. Duke Math-
+ematical Journal, 91(3):515 – 560, 1998.
+[Nak00]
+Hiraku Nakajima. Reflection functors for quiver varieties and Weyl group
+actions. Mathematische Annalen, 327, 08 2000.
+[Nak01]
+Hiraku Nakajima. Quiver varieties and finite dimensional representa-
+tions of quantum affine algebras. Journal of the American Mathematical
+Society, 14(1):145–238, 2001.
+[Sai86]
+Morihiko Saito. Mixed Hodge modules. Proc. Japan Acad. Ser. A Math.
+Sci., 62(9):360–363, 1986.
+[Shm09]
+D. A. Shmelkin. Some remarks on Nakajima’s quiver varieties of type
+A, 2009.
+[Sho88]
+Toshiaki Shoji.
+Geometry of orbits and springer correspondence.
+In
+Orbites unipotentes et représentations - I. Groupes finis et Algèbres de
+Hecke, number 168 in Astérisque. Société mathématique de France, 1988.
+[Spr76]
+T. Springer. Trigonometric sums, Green functions of finite groups and
+representations of Weyl groups. Inventiones Mathematicae, 36:173–207,
+1976.
+52
+
diff --git a/fNE1T4oBgHgl3EQfygV0/content/tmp_files/load_file.txt b/fNE1T4oBgHgl3EQfygV0/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d2016f8b3ec34352cd888854b2b96c087f450e19
--- /dev/null
+++ b/fNE1T4oBgHgl3EQfygV0/content/tmp_files/load_file.txt
@@ -0,0 +1,2406 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf,len=2405
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='03434v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='RT] 9 Jan 2023 Comet-shaped quiver varieties, Weyl group actions, and modified Kostka polynomials Mathieu Ballandras Université de Paris Scuola Internazionale Superiore di Studi Avanzati Instituto de Ciencias Matemáticas mballandras@imj-prg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='fr January 10, 2023 Abstract We study an algebra spanned by modified Kostka polynomials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Particular structure coefficients of this algebra are interpreted as traces of some Weyl group actions on the intersection cohomology of comet-shaped quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Contents 1 Introduction 2 2 Symmetric functions 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Generalities .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Characters of the symmetric group and symmetric functions .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 7 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Orthogonality and Macdonald polynomials .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 8 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Generalities about scalar products on Sym [X] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 8 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Hall pairing and (q, t)-deformations .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 9 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 A result of Garsia–Haiman .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 10 3 Geometric background 12 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Notations and generalities on the bounded derived category of con- structible sheaves .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 12 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Intersection cohomology .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 14 4 Main objects and notations 15 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Adjoint orbits in gln .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 15 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Notations for adjoint orbits .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 15 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Resolutions of Zariski closures of adjoint orbits .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 16 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 The varieties QOµ,σ and their resolutions .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 18 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Intersection cohomology of the varieties QOµ,σ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 21 1 5 Construction in terms of Nakajima’s quiver varieties 22 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Generalities about Nakajima’s quiver varieties .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 22 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Resolutions of Zariski closures of adjoint orbits as Nakajima’s framed quiver varieties .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 24 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Comet-shaped quiver varieties .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 26 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 Family of comet-shaped quiver varieties .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 28 6 Monodromic Weyl group action 29 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Family of resolutions of closures of adjoint orbits .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 29 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Decomposition of the family QL,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 32 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 W-equivariant structure on the cohomology of the fibers of the family �QL,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 33 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 Monodromic Weyl group action on the cohomology of �QL,P ,σ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 36 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5 Frobenius morphism and monodromic action .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 37 7 Geometric interpretations in the algebra spanned by Kostka poly- nomials 41 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Description of the algebra .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 41 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Interpretation of certain coefficients as traces of Weyl group actions on the intersection cohomology of quiver varieties .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 46 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Cohomological interpretation in the multiplicative case .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 47 1 Introduction The modified Kostka polynomials �Kλ,ρ(q, t) form a family of two-variable polyno- mials indexed by pairs of partitions of some integer n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They are a two-parameter deformation of the Kostka numbers and appear in the theory of symmetric func- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They were introduced by Garsia–Haiman [GH96] in the expression of the modified Macdonald polynomials in terms of Schur functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The fact that they are polynomials with non-negative integer coefficients is an important result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This is known as the Macdonald conjecture [Mac88] which is a consequence of the n!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='- conjecture of Garsia–Haiman [GH93], proved by Haiman [Hai01].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this article we study an aglebra spanned by those polynomials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Its structure coefficients cλ µ,ν(q, t) were introduced by Rodriguez-Villegas in unpublished notes, they are defined by �Kµ,ρ �Kν,ρ = � λ∈Pn cλ µ,ν �Kλ,ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' with Pn the set of partitions of the integer n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We focus in particular on the co- efficients c1n µ,ν(q, t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They generalize the q, t-Catalan sequence of Garsia–Haiman [GH96].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We give an interpretation of the coefficients c1n µ,ν(0, t) in terms of Weyl group action on the cohomology of a comet-shaped quiver variety.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We also give a conjectural interpretation of the coefficients c1n µ,ν(q, t) in terms of cohomology of a partial resolution of a character variety.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We prove the q = 1 specialization of the conjecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The geometric framework is the following.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The base field K is either C or Fq an algebraic closure of a field Fq with q elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We consider cohomology with coefficients in κ which is either C or Ql with l and q coprime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Fix some integers n > 0, g ≥ 0 and k > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let O = (O1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ok) be a k-tuple of adjoint orbits in 2 gln, the Lie algebra of GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by Oj the Zariski closure of the class Oj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Some genericity conditions are imposed on the eigenvalues of the k-tuple O (see Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The main object in this article is the following variety: QO := � (A1, B1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ag, Bg, X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk) ∈ gl2g n ×O1 × · · · × Ok �� g � i=1 [Ai, Bi] + k � j=1 Xj = 0 � // GLn, with [Ai, Bi] := AiBi − BiAi the Lie bracket.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The quotient is a Geometric Invariant Theory (GIT) quotient with respect to the overall adjoint action of GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Those varieties were studied by Crawley-Boevey [CB03b, CB06] in genus g = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For any genus and semisimple adjoint orbits, they were studied by Letellier, Hausel and Rodriguez-Villegas [HLRV11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier [Let11] generalized to any type of adjoint orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Those varieties are called comet-shaped quiver varieties due to their interpre- tation as Nakajima’s quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They are additive analogues of the following character varieties: MC := � (A1, B1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ag, Bg, X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk) ∈ GL2g n ×C1 × · · · × Ck �� A1B1A−1 1 B−1 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' AgBgA−1 g B−1 g X1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Xk = Id � // GLn with C = (C1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ck) a k-tuple of conjugacy classes in GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The action of GLn is by overall conjugation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Those varieties classify representations of the fundamental group of a genus g Riemann surface with k punctures and prescribed monodromies around those punctures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They were extensively studied by Hausel, Letellier and Rodriguez-Villegas [HLRV11] for semisimple classes and by Letellier [Let13] for any Jordan type.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly to the additive case there is a genericity condition imposed on the eigenvalues of C in order for the quotient to be well behaved (see Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In general the varieties QO and MC are singular, therefore it is interesting to study their intersection cohomology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier [Let11, Let13] studied intersection cohomology of those varieties by constructing resolutions of singularities �QL,P ,σ → QO (see Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Those resolutions of singularities originate in Springer theory [Spr76, BM83] and Lusztig’s theory of parabolic induction [Lus84, Lus85, Lus86].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Thanks to this origin, the resolutions of singularities (in both the additive and multiplicative cases) come with a Weyl group action on their cohomology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This action is called the Springer action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this article we focus on the additive version (the comet-shaped quiver varieties).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We come back to character varieties, at the end, in order to give a conjectural interpretation of the coefficient c1n µ,ν(q, t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In the additive case, thanks to the quiver variety point of view, Weyl group ac- tions other than the Springer action can be constructed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Weyl group actions on the cohomology of Nakajima’s quiver varieties were studied extensively by Nakajima [Nak94, Nak00], Lusztig [Lus00] and Maffei [Maf02].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They were used to prove Kac conjecture by Letellier, Hausel, Rodriguez-Villegas [HLRV13] and to study unipo- tent characters of GLn(Fq) by Letellier [Let12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Nakajima’s construction [Nak94] of the Weyl group action relies on the hyperkähler structure of the quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We use this construction, together with technics from Lusztig (see Letellier [Let05, Proof of proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3]) and ideas from Mellit [Mel19, Section 8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This allows to describe uniformly a monodromic action on the cohomology of quiver varieties with 3 semisimple adjoint orbits and the Springer action on the cohomology of resolutions of quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Thanks to this construction we can obtain an interpretation of the coefficients c1n µ,ν(0, t) in terms of the Weyl group action on the compactly supported intersection cohomology of some comet-shaped quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider a generic 4-tuple of adjoint orbits of the following type: O1 has one eigenvalue with Jordan type µ′ ∈ Pn, O2 has one eigenvalue with Jordan type ν′ ∈ Pn, O3 is semisimple regular, it has n distinct eigenvalues, O4 is semisimple with one eigenvalue of multiplicity n − 1 and the other of multiplicity 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the Weyl group with respect to O3 is the symmetric group Sn and it acts on the compactly supported intersection cohomology of QO.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let w be a n-cycle in this Weyl group, then c1n µ,ν (0, t) = t −dO 2 � r tr � w, IH2r c (QO, κ) � tr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In general, in the multiplicative case, only the Springer action exists.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The mon- odromic action was constructed by Mellit [Mel19, Section 8] only with respect to one puncture for a regular monodromy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Springer action is defined by Letellier [Let13] on the cohomology of partial resolutions of character varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The partial resolution relevant to describe the coefficient c1n µ,ν is the variety Mµ,ν introduced in 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The group Sn acts on the compactly supported intersection cohomology of Mµ,ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For w ∈ Sn, the w-twisted mixed Hodge polynomial of Mµ,ν is IHw c (Mµ,ν;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' u, v) := � i,r urvi tr � w, IH2r,i c (Mµ,ν, κ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' with IH2r,i c (Mµ,ν, κ) the weight 2r graded part of the degree i compactly supported intersection cohomology of Mµ,ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Conjecture 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let w be a n-cycle in Sn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The coefficient c1n µ,ν(q, t) is related to the w-twisted mixed Hodge polynomial of Mµ,ν by c1n µ,ν(q, t) = t − dim Mµ,ν 2 IHw c � Mµ,ν, 1 q, √qt � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In [Bal22] we compute Poincaré polynomial for compactly supported intersec- tion cohomology of the character varieties MC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This allows to prove the following theorem which is the Poincaré polynomial specialization of Conjecture 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let w be a n-cycle in Sn, the coefficient c1n µ,ν is related to the w- twisted Poincaré polynomial of Mµ,ν by c1n µ,ν(1, t) = t − dim Mµ,ν 2 � i t i 2 tr � w, IHi c (Mµ,ν, κ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 4 Aknowledgement This work is part of my PhD thesis under the supervision of Emmanuel Letellier and Fernando Rodriguez-Villegas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' I am very grateful to both of them for introducing me to the interesting topic of Weyl group actions on the cohomology of character varieties and quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' I am aslo thankful to François Bergeron for interesting comments and suggestions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 2 Symmetric functions 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Generalities Notations 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 (Partitions).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A partition of an integer n ∈ N is a decreasing sequence of non-negative integers λ = (λ1, λ2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , λl(λ)) with |λ| := λ1 + λ2 + · · · + λl(λ) = n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The length of λ is the number l(λ) of non-zero terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The set of partitions of n is denoted by Pn and P := � n∈N Pn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The dominance ordering on P is defined by λ ⪯ µ if and only if |λ| = |µ| and k � i=1 λi ≤ k � i=1 µi for all k ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For λ = (λ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , λl) a partition, we introduce the following notation Pλ := Pλ1 × · · · × Pλl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 (Young diagrams).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' To a partition λ, we associate the following set {(i, j) |1 ≤ i ≤ l(λ) and 1 ≤ j ≤ λi} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This set is called the Young diagram of λ, it gives a graphical way to think about partitions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The transpose of a Young diagram is obtained by permuting i and j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The transpose λ′ of a partition λ is the partition with Young diagram the transpose of the Young diagram of λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Young diagram of the partition λ = (6, 4, 2) has the following graphical representation x .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The box x has coordinates (i, j) = (1, 3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The arm length of x is number of box to the right of x, in this case a(x) = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The leg length is the number of box under x, we have l(x) = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 5 Notations 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 (Symmetric functions).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let X = (x1, x2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ) be an infinite set of variable and let Sym[X] be the ring of symmetric functions in (x1, x2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We use the usual notations from Macdonald’s book [Mac15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In particular, the usual basis of symmetric functions indexed by partitions: mλ, eλ, hλ, pλ and sλ are respectively the monomial, elementary, complete, power sum and Schur symmetric functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Hall pairing is denoted by ⟨.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ⟩ and is defined by ⟨pλ, pµ⟩ = δλ,µzλ, (1) the symbol δλ,µ is 1 if λ and µ are the same partition and 0 otherwise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The order of the stabilizer of a permutation of cycle type λ is denoted by zλ, namely zλ = k � l=1 iml l ml!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' for a partition λ = (i1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' i1 � �� � m1 , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , ik, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ik � �� � mk ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 (Adams operator).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For n ∈ Z>0, the Adams operator pn is a ring endomorphism of Sym[X].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It can be defined by its values on the generating family of power sums, pm [pn[X]] := pmn[X] for m ∈ N>0 and n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following notation is commonly used for Adams operators F [Xn] := pn [F[X]] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' More generally, Adams operator are defined in any lambda ring.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this article, the only lambda rings appearing are rings of symmetric functions and polynomial rings such as K[u].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' On such polynomial rings, the Adams operator pn is defined by pn[u] := un.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let k be a positive integer, we consider the space of multivariate symmetric functions in k infinite sets of variables over Q(q, t) Sym [X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk] := Q(q, t) ⊗ Sym[X1] ⊗ · · · ⊗ Sym[Xk].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A series with coefficients in this ring of multivariate symmetric functions will conve- niently encode cohomological information about comet shaped quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The ring of such series is denoted by Sym [X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk] [[s]], it is a lambda ring, and the Adams operators extend to ring endomorphisms of Sym [X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk] [[s]] defined by pn � f(q, t)F1 [X1] ⊗ · · · ⊗ Fk [Xk] sl� = f(qn, tn)F1 [Xn 1 ] ⊗ · · · ⊗ Fk [Xn k ] snl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6 (Plethystic substitution).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let F be a symmetric function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The ring of symmetric functions Sym [X] is freely generated by power sums, so that F can be uniquely obtained as a polynomial expression in the power sums (pn)n∈N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Interpreting pn as the Adams operator, the same polynomial expression defines an operator acting on any lambda ring Λ, this operator is denoted by F[.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For G ∈ Λ, the expression F[G] is called a plethystic substitution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6 Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly to Adams operators, the operator F[.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ] naturally extends to Sym [X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk] [[s]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8 (Plethystic exponential and logarithm).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The plethystic exponential Exp : s Sym[X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk][[s]] → 1 + Sym[X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk][[s]] is defined by Exp[G] := exp �� n≥1 pn[G] n � , its inverse, the plethystic logarithm Log : 1 + s Sym[X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk][[s]] → Sym[X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk][[s]], is defined by Log[1 + G] := � n≥1 µ(n) n pn [log(1 + G)] , where µ is the usual Mobius function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Contrarily to the ordinary ones, the plethystic exponential and logarithm are written with an uppercase character.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Plethystics operations satisfy the relations Exp[F + G] = Exp[F] Exp[G], Log[(1 + F)(1 + G)] = Log[1 + F] + Log[1 + G], Log[Exp[G]] = G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Characters of the symmetric group and symmetric func- tions Let Rn be the vector space spanned by characters of the symmetric group Sn and let R := � n≤0 Rn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There is a natural ring structure on R and a pairing such that R is naturally isomorphic to Sym[X].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let us recall this well-known fact, see [Mac15] for more details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let χ and η be two characters of Sn, and let Vχ, respectively Vη be the associated representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The pairing is defined by ⟨χ, η⟩ = dim HomSn (Vχ, Vη) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The spaces Rm and Rn are orthogonal if m ̸= n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The product of two characters χ ∈ Rm and η ∈ Rn is the character of the representation IndSm+n Sm×Sn Vχ ⊗ Vη, it is denoted by χ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='η.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The irreducible characters of the symmetric group Sn are indexed by partitions of n, they are denoted by (χλ)λ∈Pn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The irreducible representation associated to the character χλ is denoted by Vλ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The indexing is the same as in Macdonald’s book [Mac15], so that V(n) is the trivial representation and V(1n) the sign representation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define the characteristic map ch : R → Sym[X] by ch(χλ) := sλ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is an isomorphism between R and Sym[X] compatible with the products and the pairings, Sym[X] being endowed with the Hall pairing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' See Macdonald [Mac15, I-7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 7 Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The last proposition gives a representation theoretic meaning to symmetric functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider V a representation of Sn with character χV , then The pairing ⟨sλ, ch(χV )⟩ gives the multiplicity of the irreducible representation Vλ in the representation V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The pairing ⟨pµ, ch(χV )⟩ gives the trace of an element in Sn with cycle type µ on the representation V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12 (Frobenius characteristic).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We extend the characteristic map ch to bigraded representations of Sn by adding variables q and t to keep track of the degree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' To a bigraded representation of the symmetric group V = � (i,j)∈N2 Vi,j is associated a symmetric function over Z(q, t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This symmetric function is given by ch(V ) = � λ∈Pn � (i,j)∈N2 ⟨Vi,j, χλ⟩ qitjsλ, (2) where we have identified the representation Vi,j with its character so that ⟨Vi,j, χλ⟩ is the multiplicity of the irreducible representation of type λ in Vi,j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The symmet- ric function ch(V ) is called the q, t-graded Frobenius characteristic of the bigraded representation V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Orthogonality and Macdonald polynomials In this section we recall the characterization of modified Macdonald polynomials following Mellit [Mel20, Mel18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Generalities about scalar products on Sym [X] A scalar product on Sym [X] is a non-degenerate Q(q, t)-bilinear form (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' )S : Sym [X] × Sym [X] → Q(q, t) F, G �→ (F[X], G[X])S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It can be extended to multivariate symmetric functions by specifying the variable acted upon with a lower index (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' )S X : Sym[X, Y1,· · · , Yk] × Sym[X, Z1,· · · , Zl] → Sym[Y1,· · · , Yk, Z1,· · · , Zl], on pure tensors it reads (F[X]⊗F ′[Y1,· · · , Yk], G[X]⊗G′[Z1,· · ·, Zl])S X := (F[X], G[X])SG′[Z1,· · · , Zl]F ′[Y1,· · · , Yk].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='13 (Homogeneity).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When considering families of symmetric func- tions indexed by partitions such as (uλ)λ∈P, the symmetric function uλ is always assumed to be homogeneous of degree |λ|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14 (Reproducing kernel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let (uλ)λ∈P and (vµ)µ∈P be two basis of Sym[X] dual with respect to a scalar product (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' )S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the element KS[X, Y ] ∈ Sym[X, Y ] defined by KS[X, Y ] := � λ∈P uλ[X]vλ[Y ], it is called the reproducing kernel of the scalar product (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' )S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It depends only on the scalar product but not on the choice of dual basis, it satisfies (KS[X, Y ], F[X])S X = F[Y ].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 8 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Hall pairing and (q, t)-deformations Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Recall that the Hall pairing satisfies ⟨pλ, pµ⟩ = δλ,µzλ, hence (pλ)λ∈P and � z−1 µ pµ � µ∈P are dual basis with respect to this pairing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This gives the reproducing kernel of the Hall pairing: Exp[XY ] = � λ∈P pλ[X]pλ[Y ] zλ = � n hn[XY ].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='16 ((q, t)-Hall pairing).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The (q, t)-deformation of the Hall pairing is defined by (F[X], G[X])q,t := ⟨F[X], G[(q − 1)(1 − t)X]⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The reproducing kernel of the (q, t)-Hall pairing is Exp � XY (q − 1)(1 − t) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let M⪯λ be the subspace of Sym [X] spanned by monomial sym- metric functions mµ[X] with µ ⪯ λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Macdonald polynomials � ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] � λ∈P are uniquely determined by: Orthogonality ( ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t], ˜Hµ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t])q,t = 0 if λ ̸= µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' One of the triangularity condition ˜Hλ[X(t−1)] ∈ M⪯λ or ˜Hλ[X(q−1)] ∈ M⪯λ′.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Normalization ˜H[1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover aλ(q, t) := � ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t], ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] �q,t = � x∈λ (qa(x)+1 − tl(x))(qa(x) − tl(x)+1), (3) where the product is over the Young diagram of λ and a(x) is the arm length and l(x) the leg length (see Notations 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Mel20] corollary 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The modified Macdonald polynomials ˜Hλ [X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] were first introduced by Garsia– Haiman [GH96] as a deformation of other polynomials defined by Macdonald [Mac15].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The definition recalled here comes from [Mel20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='19 (Modified Kostka polynomials).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The modified Kostka polynomials � �Kλ,ρ(q, t) � λ,ρ∈Pn are defined as the coefficients of the transition matrix between the basis of Schur functions and the basis of modified Macdonald polynomials ˜Hρ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = � λ∈Pn �Kλ,ρ(q, t)sλ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variables (q, t) will often be omitted and the modified Kostka polynomial denoted by �Kλ,ρ and the modified Macdonald polynomial by ˜Hλ[X].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 9 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 A result of Garsia–Haiman The remaining of this section is devoted to the presentation of a result of Garsia– Haiman [GH96, Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This result is important to study the coefficients c1n µ,ν in the next sections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Even though there are no new results, some proofs are included for convenience of the reader.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The operator ∆1 is defined by ∆1F [X] := F[X] − F � X + (1 − q)(1 − t) z � Exp [−zX] |z0 , where |z0 means taking the coefficient in front of z0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This operator acts on modified Macdonal polynomials by ∆1 ˜Hλ [X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = (1 − t)(1 − q) � (i,j)∈λ qj−1ti−1 ˜Hλ [X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover the following relation holds ˜Hλ [1 − u;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = � (i,j)∈λ � 1 − uqj−1ti−1� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (4) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [GH96, Corollary 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 and Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2] Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' At first order in u ˜Hλ [1 + u;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = 1 + u � (i,j)∈λ qj−1ti−1 + O(u2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (5) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' One should be careful with plethystic substitutions, to compute the left hand side of (5) one cannot just substitute −u for u in (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indeed pn[1 −u] = 1 −un and pn[1 + u] = 1 + un so that substituting −u for u in the latter gives back the former only when n is odd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote by dλ,µ the coefficient of pµ in the expansion of ˜Hλ in the basis of power sums (pκ)κ∈Pn, then ˜Hλ [1 − u;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = � |µ|=|λ| dλ,µ � i (1 − uµi), ˜Hλ [1 + u;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] = � |µ|=|λ| dλ,µ � i (1 + uµi).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We conclude by comparing the coefficient in front of u and using (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let F ∈ Symn [X] be a symmetric function of degree n ≥ 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the coefficient in front of u in F[1 + u] is given by the Hall pairing with a complete symmetric function F[1 + u]|u = � h(n−1,1)[X], F[X] � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The coefficient of mλ in the monomial expansion of F is denoted by cλ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The plethystic substitution F[1 + u] corresponds to the evaluation of the symmetric function F on the set of variables (1, u, 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ), moreover F[1 + u] = � |λ|=n cλmλ[1 + u].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 10 Therefore the only mλ contributing are the one with λ of length at most two and the coefficient in front of u is c(n−1,1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The conclusion follows as complete symmet- ric functions and monomial symmetric functions are dual with respect to the Hall pairing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let F ∈ Symn [X] be a symmetric function of degree n then F[1 − u] 1 − u ���� u=1 = ⟨F[X], pn[X]⟩ , where |u=1 means setting u = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let dλ be the coefficient in front of pλ in the power sum expansion of F, F[1 − u] = � |λ|=n dλpλ[1 − u] = � |λ|=n dλ � i (1 − uλi).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When dividing by (1 − u) and setting u = 1 all terms coming from partitions of length at least two will vanish as (1 − u)2 divides them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore we have F[1 − u] 1 − u ���� u=1 = d(n) 1 − un 1 − u ���� u=1 = nd(n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The size of the centralizer of a n-cycle in Sn is z(n) = n, the conclusion follows by orthogonality of power sums (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now we can state and recall the proof of an important theorem of Garsia– Haiman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='25 (Garsia–Haiman [GH96] Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote by �′ (i,j)∈λ a product over the young diagram of a partition λ omitting the top left corner with (i, j) = (1, 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following identity holds (−1)n−1s(1n)[X] = (q − 1)(1 − t) � |λ|=n � (i,j)∈λ qj−1ti−1 �′ (i,j)∈λ(1 − qj−1ti−1) ˜Hλ[X] aλ(q, t) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (6) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The reproducing kernel of the (q, t)-Hall pairing is given in Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The degree n term of Exp[Z] is hn[Z].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The basis � ˜Hλ[X] � λ∈P and � ˜Hλ[X] aλ � λ∈P are dual with respect to this scalar product.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Following Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14 and Remarks 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15, 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='17, the degree n term of the reproducing kernel of the (q, t)-Hall pairing is hn � XY (q − 1)(1 − t) � = � |λ|=n ˜Hλ[X] ˜Hλ[Y ] aλ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now expand hn in the basis of power sums, proceed to the substitution Y = 1 − u and apply (4) � |µ|=n z−1 µ pµ � X(1 − u) (q − 1)(1 − t) � = � |λ|=n ˜Hλ[X] � (i,j)∈λ (1 − uqj−1ti−1) aλ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 11 Divide by (1 − u), set u = 1, apply Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='24 to the left hand side and compute explicitly the right hand side: � |µ|=n z−1 µ � pµ � XY (q − 1)(1 − t) � , p(n)[Y ] � Y = � |λ|=n ˜Hλ[X] �′ (i,j)∈λ (1 − qj−1ti−1) aλ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As Adams operator are ring morphisms, we have pµ � XY (q − 1)(1 − t) � = pµ � X (q − 1)(1 − t) � pµ[Y ], then by orthogonality of power sums (1) p(n) � X (q − 1)(1 − t) � = � |λ|=n ˜Hλ[X] �′ (i,j)∈λ (1 − qj−1ti−1) aλ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (7) Apply the operator ∆1 to (7).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' According to Proposition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='21, the operator ∆1 is diagonal in the basis of Macdonald polynomials and we obtain, up to a sign, the right hand side of (6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let us compute the left hand side ∆1p(n) � X (q−1)(1−t) � = p(n) � X (q−1)(1−t) � − p(n) � X (q−1)(1−t) − 1 z � Exp[−zX] |z0 = p(n) � X (q−1)(1−t) � − p(n) � X (q−1)(1−t) � Exp[−zX] |z0 + p(n) � 1 z � Exp[−zX] |z0 = 1 zn Exp[−zX] |z0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In the second line we used that the Adams operator pn is a ring morphism and in the last line that it acts on z by raising to the power n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now Exp[−zX] is the inverse of Exp[zX] so that if X is the infinite set of variables (x1, x2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ), then Exp[−zX] = � i (1 − zxi).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The coefficient in front of zn is (−1)nen[X] so that (−1)nen[X] = −(q − 1)(1 − t) � |λ|=n � (i,j)∈λ qj−1ti−1 �′ (i,j)∈λ(1 − qj−1ti−1) ˜Hλ[X] aλ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' To conclude, notice that en = s(1n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 3 Geometric background In this section we recall classical results about intersection cohomology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The main reference is Beilinson–Bernstein–Deligne–Gabber [BBDG18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Notations and generalities on the bounded derived cate- gory of constructible sheaves The field K is either C or an algebraic closure Fq of a finite field Fq with q elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let X be an algebraic variety over K and let l be a prime different from the char- acteristic of K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote by κX the constant l-adic sheaf on X with coefficients in 12 Ql.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When there are no risk of confusion we just write κ instead of κX.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For K = C we can also consider the constant sheaf with complex coefficients, in the analytic topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The bounded derived category of κ-constructible sheaves on X is denoted by Db c (X).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Its objects are represented by complexes of sheaves K such that the cohomology sheaves HiK are κ-constructible sheaves on X and finitely many of them are non-zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Y be a variety over K and let f : X → Y be a morphism, we have the usual four functors f ∗, f !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' : Db c (Y ) → Db c (X) , f∗, f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' : Db c (X) → Db c (Y ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 (Base change).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let K ∈ Db c (Y ′) and consider a cartesian square X′ Y ′ X Y, g b a f (8) then the natural morphism f ∗a!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='K → b!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='g∗K is an isomorphism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let α ֒→ X be a geometric point of X and let β be its image by f.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variety Xα := X′ ×X α is the fiber of b over α and the variety Yβ := Y ′ ×Y β is the fiber of a over β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The cartesian square (8) induces the following cartesian square where h is an isomorphism Xα Yβ α β.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' h The base change isomorphism for this diagram identifies with the stalk at α of the base change isomorphism of Diagram (8), f ∗a!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='Kα → b!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='g∗Kα.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This morphism is nothing but the morphism obtained by functoriality of the com- pactly supported cohomology H c(Yβ, K) h∗ −→ H c(Xα, h∗K).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let W be a finite group acting from the left on a variety X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For all w ∈ W there is a morphism w : X → X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' An action of W on an element K ∈ Db c (X) is the data of isomorphisms φw : w∗K ∼= K such that for all w, w′ ∈ W, φw′w = φww∗(φw′), (9) and such that φ1 = Id.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We say that the complex K is W-equivariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When the action of W on X is trivial, an action of W on K ∈ Db c (X) is just a group morphism from the opposite group W op to the group of automorphism Aut(K).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 13 Proposition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let f : X → Y be a W-equivariant morphism between varieties with left W-action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let W act on K by morphisms φw : w∗K ∼= K, then W acts on f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Base change formulas allow to define the action, for all w ∈ W they provide an isomorphism w∗f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='K → f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='w∗K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Compose this isomorphism with f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='φw to obtain an isomorphism �φw : w∗f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='K → f!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The compatibility (9) follows from functoriality of base change.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Intersection cohomology Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7 (Intersection complex).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Y ֒→ X be a closed embedding and let j : U ֒→ Y be an open embedding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Assume that U is smooth, irreducible and that U = Y .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let ξ be a local system on U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The intersection complex IC• Y,ξ is the unique (up to isomorphism) element K in Db c (Y ) characterized by HiK = 0 if i < − dim Y, H− dim Y K|U = ξ, dim � Supp HiK � < −i if i > − dim Y, dim � Supp HiDY K � < −i if i > − dim Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We also denote by IC• Y,ξ its extension j∗IC• Y,ξ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8 (Continuation principle).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The intersection complex of ξ can also be defined as the intermediate extension IC• Y,ξ = j!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='∗ξ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover the functor j!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='∗ is fully faithful (see Kiehl-Weissauer [KW01, III - Corollary 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The intersection complex does not depend on the choice of smooth open subset in Y .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When the local system ξ is not specified, it is chosen to be the constant sheaf κU and IC• X := IC• X,κU.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote by IC• X Notations 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The shifted intersection complexes are IC• X,ξ := IC•−dim X X,ξ and IC• X := IC•−dim X X .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11 (Intersection cohomology).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let p : X → Spec K be the structural morphism and k an integer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The k-th intersection cohomology space of X is IHk(X, κ) := Hk−dim Xp∗IC• X = Hkp∗IC• X and the k-th compactly supported intersection cohomology space of X is IHk c (X, κ) := Hk−dim Xp!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='IC• X = Hkp!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='IC• X For K = C, Saito [Sai86] proved that the intersection cohomology spaces carry a mixed Hodge structure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Thus there exists on IHk c (X, Q) an increasing finite filtration called the weight filtration and denoted by W k such that the complexified quotient C ⊗Q W k r /W k r−1 carries a pure Hodge structure of weight r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Hodge numbers of this structure are denoted hi,j,k c (X) = dim IHi,j,k c (X, C) and satisfy i + j = r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 14 Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The mixed Hodge structure is encoded in the mixed Hodge poly- nomial, IHc (X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' x, y, v) := � i,j,k hi,j,k c (X)xiyjvk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (10) This polynomial has an important specialisation, the Poincaré polynomial Pc(X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' v) := IHc (X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 1, 1, v) = � k dim IHk c (X, κ)vk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (11) In this article "Poincaré polynomial" always refers to "Poincaré polynomial for com- pactly supported intersection cohomology".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 4 Main objects and notations 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Adjoint orbits in gln 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Notations for adjoint orbits The goal of this article is to relate some geometric objects to combinatorial data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The first step is the well-known labelling of adjoint orbits by their Jordan types, which we recall in order to fix the notations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For an integer r and for z ∈ K, we denote by Jr(z) ∈ glr the Jordan block of size r with eigenvalue z so that Jr(z)−z Idr is nilpotent of order r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let µ = (µ1, µ2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µs) be a partition of an integer m and let z ∈ K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Jµ(z) be the matrix with eigenvalue z and Jordan blocks of sizes given by (µj)1≤j≤s, Jµ(z) := \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed Jµ1(z) Jµ2(z) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Jµs(z) \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 ∈ glm .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let ν = (ν1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , νl) ∈ Pn be a partition of n, introduce the following notation Pν := Pν1 × Pν2 × · · · × Pνl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider a diagonal matrix σ, σ = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed σ1 Idν1 σ2 Idν2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' σl Idνl \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 , (12) with σi ̸= σj for i ̸= j, so that νi is the multiplicity of the eigenvalue σi ∈ K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider an element µ = � µ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µl� in Pν, we denote by Oµ,σ the adjoint orbit of the matrix Jµ,σ := \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed Jµ1(σ1) Jµ2(σ2) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Jµl(σl).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 15 Let us recall a well-known proposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Zariski closure of the adjoint orbit Oµ,σ is Oµ,σ = � ρ⪯µ Oρ,σ, the union is over the set of l-tuples ρ = � ρ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , ρl� with ρj ⪯ µj for 1 ≤ j ≤ l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The dominance order on partition was recalled in 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Over Fq, there is a more precise description of adjoint orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by F the Frobenius endomorphism of gln(Fq) raising the coefficients to the power q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 (Type of an F-stable adjoint orbit).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let O be an F-stable adjoint orbit in gln(Fq), i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', an orbit such that F(O) ⊂ O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the set OF of fixed points in O under the Frobenius is not empty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The characteristic polynomial of O has its coefficients in Fq so that its eigenvalues, which live in Fq, are permuted by the Frobenius.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The spectrum of O, with multiplicities, reads \uf8eb \uf8ec \uf8ec \uf8ed � γ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , γqd1−1 1 � , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , � γ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , γqd1−1 1 � � �� � m1 , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , � γl, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , γqdl−1 l � , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , � γl, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , γqdl−1 l � � �� � ml \uf8f6 \uf8f7 \uf8f7 \uf8f8 , where γi ∈ Fq is such that γqdi−1 i ̸= γi, γqdi i = γi and γi ̸= γj for i ̸= j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the orbit O determines partitions ωi ∈ Pmi giving the size of the Jordan blocks of the Frobenius orbit of eigenvalues � γi, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , γqdi−1 i � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Up to ordering, it defines a sequence ω = (d1, ω1) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (dl, ωl) in Z>0 × P called the type of the adjoint orbit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Resolutions of Zariski closures of adjoint orbits In this section we recall the construction of resolutions of closures of adjoint orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The references for this construction are Kraft–Procesi [KP81], Nakajima [Nak98, Nak01], Crawley-Boevey [CB03a, CB03b] and Shmelkin [Shm09] (see also Letellier [Let11]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Using the notations from the previous section, consider an adjoint orbit Oµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The matrix σ ∈ gln is diagonal as in (12) and we denote by M its stabilizer in GLn, M = \uf8eb \uf8ec \uf8ed GLν1 0 0 GLν2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' \uf8f6 \uf8f7 \uf8f8 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let µ = (µ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µl) ∈ Pν so that µi is a partition of the integer νi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The transposed of the partition µi is denoted by µi′ = � µi 1 ′, µi 2 ′, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let L be the subgroup of GLn formed by block diagonal matrices with blocks of size µi r ′, it is a subgroup of M with 16 the following form L = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ed ν1 � �� � GLµ1 1 ′ 0 0 GLµ1 2 ′ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ν2 � �� � GLµ2 1 ′ 0 0 GLµ2 2 ′ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f8 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For a partition ν = (ν1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , νl), define Sν := Sν1 × · · · × Sνl and GLν := GLν1 × · · · × GLνl .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now for ρ = (ρ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , ρl) ∈ Pν, we use the following notations, GLρ := GLρ1 × .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' GLρl = � r,s GLρrs and Sρ := Sρ1 × .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Sρl = � r,s Sρrs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the previously introduced Levi subgroups satisfy M ∼= GLν and L ∼= GLµ′.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by P the parabolic subgroup of blocks upper triangular matrices having L as a Levi factor, then P = LUP with UP = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ed ν1 � �� � Idµ1 1 ′ ∗ 0 Idµ1 2 ′ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ν2 � �� � Idµ2 1 ′ ∗ 0 Idµ2 2 ′ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f8 , and the Lie algebra counterpart of this Levi decomposition is p = l ⊕ uP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5 (Resolutions of Zariski closures of conjugacy classes).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider �YL,P,σ := � (X, gP) ∈ gln × GLn /P ��g−1Xg ∈ σ + uP � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The image of the projection to the first factor �YL,P,σ → gln is the Zariski closure of the adjoint orbit Oµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover the following map is a resolution of singularities pσ : �YL,P,σ → Oµ,σ (X, gP) �→ X .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 17 Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' If M (which is defined as the stabilizer of σ in GLn) is exactly L then the adjoint orbit Oµ,σ is semisimple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Semisimple orbits in gln are closed and smooth, for such orbits, pσ is an isomorphism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The decomposition theorem and Springer theory (or more generally Lusztig’s parabolic induction [Lus84, Lus85, Lus86]) provide more information about the previous resolution of singularities in terms of Weyl group representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For G a reductive group and for T a maximal torus in G the Weyl group is denoted by WG := NG(T)/T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Weyl group of M is WM ∼= � i Sνi and let Vρ be the representation � i Vρi of WM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Weyl group of L is WL ∼= � i,j Sµi j ′ and let ǫ be its sign representation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lusztig’s parabolic induction provides the following decomposition H +dµ c � �YL,P,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ � = � ρ⪯µ HomWM � IndWM WL ǫ, Vρ � ⊗ H +dρ c � Oρ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ � , with dµ the dimension of Oµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier [Let11] constructed an action of the relative Weyl group WM(L) = NM(L)/L on the spaces HomWM � IndWM WL ǫ, Vρ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 The varieties QOµ,σ and their resolutions The varieties we are interested in are additive analogues of the character varieties classifying representations of the fundamental group of a compact Riemann surface with k punctures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They were studied by Crawley-Boevey [CB03b, CB06] in the case g = 0, by Hausel, Letellier and Rodriguez-Villegas [HLRV11] for semisimple adjoint orbits and by Letellier [Let11] in general.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For each puncture an adjoint orbit Oµj,σj ⊂ gln is fixed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A bold symbol is used to represent k-tuple: µ := � µ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µk� , σ := � σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk� , Oµ,σ := � Oµ1,σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Oµk,σk � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (13) Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8 (Comet-shaped quiver varieties).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the variety VOµ,σ := � (A1, B1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ag, Bg, X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk) ∈ gl2g n ×Oµ1,σ1 × · · · × Oµk,σk ��� g � i=1 [Ai, Bi] + k � j=1 Xj = 0 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This is an affine variety acted upon by GLn by coordinate-wise adjoint action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The center of GLn acts trivialy so that the action factors through a PGLn-action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The main focus of the article is the following GIT quotient, QOµ,σ := VOµ,σ �� PGLn = Spec � K � VOµ,σ �GLn� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (14) We call the varieties QOµ,σ the comet-shaped quiver varieties because of their inter- pretation as Nakajima’s quiver varieties that we will recall in Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 18 Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9 (Generic adjoint orbits).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by ∆(σj) the multiset of eigenvalues of σj repeated according to multiplicities, σj r appears exactly νj r times in the multiset ∆(σj).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The k-tuple of adjoint orbits Oµ,σ is generic if and only if it satisfies the two following conditions: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' k � j=1 � α∈∆(σj) α = 0, 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' for any r ≤ n − 1 and for all (R1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Rk) with Rj ⊂ ∆(σj) of size r k � j=1 � α∈Rj α ̸= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10 (Generic conjugacy classes).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' If all the eigenvalues in σ are non- zero, then the adjoint orbits are also conjugacy classes in GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They are denoted by Cµ,σ = � Cµ1,σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Cµk,σk � instead of Oµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A k-tuple of conjugacy classes Cµ,σ is generic if it satisfies the two following conditions: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' k � j=1 � α∈∆(σj) α = 1, 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' for any r ≤ n − 1, for all (R1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Rk) with Rj ⊂ ∆(σj) of size r k � j=1 � α∈Rj α ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11 ([Let11] Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4, Corollary 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let VOµ,σ := VOµ,σ ∩ gl2g n ×Oµ1,σ1 × · · · × Oµk,σk, and let QOµ,σ be the image of VOµ,σ in QOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Assume that Oµ,σ is generic, then QOµ,σ = � ρ⪯µ QOρ,σ is a stratification of QOµ,σ with smooth strata.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover, if it is non-empty, dim QOµ,σ = dµ = n2(2g − 2) + 2 + k � j=1 dim Oµj,σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As before, σj is a diagonal matrix with stabilizer Mj := ZGLn(σj) such that with Notations 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4, Mj ∼= GLνj for some partition νj ∈ Pn given by the multiplicities of the eigenvalues of σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Jordan type of the eigenvalue σj i in the adjoint orbit Oµj,σj is µj,i ∈ Pνj i .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by 19 µj,i′ = � µj,i 1 ′, µj,i 2 ′, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � the transposed partition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Lj ⊂ Mj be the subgroup of block diagonal matrices as in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2, Lj ∼= GLµj,1 1 ′ × GLµj,1 2 ′ × .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � �� � ⊂GLνj 1 × · · · × GL µ j,lj 1 ′ × GL µ j,lj 2 ′ × .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � �� � ⊂GLνj lj .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let �YLj,P j,σj be a resolution of Oµj,σj as constructed in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let L := �k j=1 Lj and let P := �k j=1 P j, then define �YL,P ,σ := � 1≤j≤k �YLj,P j,σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier [Let11] constructed resolutions of singularities of QOµ,σ Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12 (Resolutions of QOµj ,σj).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define �QL,P ,σ := � (Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k ∈ gl2g n ×�YL,P ,σ ����� g � i=1 [Ai, Bi] + k � j=1 Xj = 0 � // PGLn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (15) The action of PGLn on gjP j is by left multiplication.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The maps pσj : �YLj,P j,σj → Oµj,σj induce a map pσ : �QL,P ,σ → QOµ,σ, this morphism is a resolution of singularities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly to Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6, if Lj = Mj for 1 ≤ j ≤ k, then the adjoint orbit Oµ,σ are semisimple and pσ is an isomorphism.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly to Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7, H•+dµ c � �QL,P ,σ, Ql � = � ρ⪯µ Vµ,ρ ⊗ IH•+dρ c � QOρ,σ, Ql � , with Vµ,ρ := �k j=1 HomWMj � Ind WMj WLj ǫ, Vρj � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore, as in Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7, Letellier constructed an action of the relative Weyl group WM(L) = �k j=1 WMj(Lj) on the cohomology of the resolution �QL,P ,σ, we call this action the Springer action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similar constructions exist in the multiplicative case (see Letellier [Let13]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' If all the eigenvalues in σ are non-zero, then the adjoint orbits are actually conjugacy classes in GLn, they are denoted with a C instead of an O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For a generic k-tuple of conjugacy classes Cµ,σ, the character variety is defined by MCµ,σ := � (A1, B1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ag, Bg, X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk) ∈ GL2g n ×Cµ1,σ1 × · · · × Cµk,σk ��� A1B1A−1 1 B−1 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' AgBgA−1 g B−1 g X1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Xk = Id � // PGLn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 20 It admits a resolution of singularities � ML,P ,σ := � (Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k ∈ GL2g n ×�YL,P ,σ ��A1B1A−1 1 B−1 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' AgBgA−1 g B−1 g X1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Xk = Id � // PGLn, and the cohomology of the resolution admits the following decomposition H•+dµ c � � ML,P ,σ, Ql � = � ρ⪯µ Vµ,ρ ⊗ IH•+dρ c � MCρ,σ, Ql � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (16) 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Intersection cohomology of the varieties QOµ,σ Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='16 (Hausel-Letellier-Villegas kernel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let k ∈ Z>0 and let g ∈ Z≥0, the k-points, genus g Cauchy function is defined by Ωg k(z, w) := � λ∈P Hλ(z, w) k � i=1 ˜Hλ � Xi, z2, w2� s|λ|, (17) with Hλ(z, w) := � � z2a+1 − w2l+1�2g (z2a+2 − w2l) (z2a − w2l+2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (18) The degree n Hausel–Letellier–Villegas kernel is defined by HHLV n (z, w) := (z2 − 1)(1 − w2) Log Ωg k(z, w) �� sn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This kernel was introduced to describe the cohomology of character varieties for genus g Riemann surfaces with k punctures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It also describes the cohomology of the varieties QOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider a generic k-tuple of adjoint orbits Oµ,σ , with µ = (µ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µk) and µj = � µj,1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µj,lj� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The transposed of the partition µj,i ∈ Pνj i is denoted by µj,i′ and we define the following symmetric function sµ′ := k � j=1 lj � i=1 sµj,i′[Xj].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (19) Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Oµ,σ be a generic k-tuple of adjoint orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Poincaré polynomial for compactly supported intersection cohomology of QOµ,σ is Pc � QOµ,σ, v � = vdµ � sµ′, HHLV n (0, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For semisimple adjoint orbits, the variety is smooth, intersection cohomology coincides with usual cohomology and the theorem is proved by Hausel, Letellier and Rodriguez-Villegas [HLRV11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The general case is proved by Letellier [Let11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Hausel–Letellier–Rodriguez-Villegas [HLRV11] proposed a conjecture for the mixed Hodge polynomial of character varieties with semisimple monodromies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It was gen- eralized by Letellier [Let13] to monodromies with any Jordan type.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Conjecture 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18 (Letellier [Let13], Conjecture 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Cµ,σ be a generic k-tuple of conjugacy classes, the mixed Hodge polynomial (see Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12) of the character variety MCµ,σ is IHc(MCµ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, v) = (v√q)dµ � sµ′, HHLV n �−1 √q, v√q �� with q = xy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 21 5 Construction in terms of Nakajima’s quiver vari- eties In order to study monodromic Weyl group action on the cohomology of the varieties QOµ,σ and �QL,P ,σ, they need to be put in a family for varying eigenvalues σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' One way to do that is to construct them as Nakajima’s quiver varieties [Nak94].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The family obtained is a fibration by the moment map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this section we recall the construction of QOµ,σ as a comet-shaped quiver variety.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In genus zero the construc- tion is due to Crawley-Boevey [CB03b], for any genus it is due to Hausel, Letellier, Rodriguez-Villegas [HLRV11] and Letellier [Let11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Generalities about Nakajima’s quiver varieties In this section we recall the construction of Nakajima’s quiver varieties [Nak94] in order to fix the notations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider a quiver Γ with a set of vertices Ω0 and a set of edges Ω1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote by t(γ) the tail and by h(γ) the head of an edge γ ∈ Ω1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A dimension vector for Γ is an element v ∈ ZΩ0 ≥0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The space of quiver representations with dimension vector v is identified with a space of matrices, Rep (Γ, v) := � γ∈Ω1 MatK(vh(γ), vt(γ)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Its cotangent bundle T ∗ Rep (Γ, v) can be identified with the space of representations of an extended quiver �Γ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This extended quiver has the same set of vertices as Γ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is obtained by adding an inverse γ to each edge γ ∈ Ω1: t(γ)• h(γ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' γ γ We denote by Ω1 the set of such inverted edges, then the set of edges of �Γ is �Ω := Ω1 ⊔ Ω1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We have T ∗ Rep (Γ, v) ∼= Rep � �Γ, v � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For a dimension vector v ∈ ZΩ0 ≥0 consider the reductive group GLv := � i∈Ω0 GLvi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This group acts on Rep � �Γ, v � by (gvi)i∈Ω0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (φγ)γ∈�Ω := � gh(γ)xγg−1 t(γ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The diagonal embedding of the multiplicative group K∗ in GLv acts trivially so that the action goes down to an action of the group Gv := GLv /K∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Lie algebra of GLv is glv := � i∈Ω0 glvi, and the Lie algebra of Gv is gv = � (xj)j∈Ω0 ∈ glv ����� � j∈Ω0 tr xj = 0 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 22 The center of this Lie algebra is given by Z(gv) = � (ξj Idvj)j∈Ω0 �����(ξj)j∈Ω0 ∈ KΩ0 with � j∈Ω0 vjξj = 0 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider an element θ ∈ ZΩ0 such that � i∈Ω0 θivi = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Such an element is called a stability parameter, it defines a character χθ of the group Gv by χθ ((gj)j∈Ω0) = � j∈Ω0 det(gj)−θj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (20) We denote by Rep � �Γ, v �θ-ss and by Rep � �Γ, v �θ-s , the θ-semistable, respectively the θ-stable locus, in the sense of GIT, for the linearization χθ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the moment map µ : Rep(�Γ, v) → gv (φγ)γ∈�Ω �→ � γ∈�Ω ǫ(γ)φγφγ where ǫ(γ) = 1 for γ ∈ Ω1 and ǫ(γ) = −1 for γ ∈ Ω1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 (Nakajima’s quiver variety).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let ξ be an element in Z(gv), the center of the Lie algebra of Gv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Nakajima’s quiver variety Mθ v(ξ) is defined as the GIT quotient Mθ v(ξ) := µ−1(ξ) ∩ Rep � �Γ, v �θ-ss // Gv .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We also need another kind of quiver varieties, the Nakajima’s framed quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Fix a second dimension vector w ∈ NΩ0 and define Rep (v, w) := � j∈Ω0 MatK(vi, wi), Rep (w, v) := � j∈Ω0 MatK(wi, vi).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' An element g ∈ GLv acts on a = (aj)j∈Ω0 ∈ Rep (v, w) by g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='a := (ajg−1 j )j∈Ω0 and it acts on b = (bj)j∈Ω0 ∈ Rep (v, w) by g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='b := (gjbj)j∈Ω0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Introduce the space of framed quiver representations Rep � �Γ, v, w � := Rep (v, w) ⊕ Rep (w, v) ⊕ Rep � �Γ, v � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this context the moment map is µ′ : Rep(�Γ, v, w) → glv (a, b, φ) �→ (µ(φ)j − bjaj)j∈Ω0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For θ ∈ ZΩ0, consider the linearization χθ : GLv → K∗ (gi)i∈Ω0 �→ � i∈Ω0 det(gi)−θi .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 (Nakajima’s framed quiver varieties).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For ξ in the center of glv and for θ ∈ ZΩ0, the Nakajima’s framed quiver variety Mθ v,w(ξ) is defined as a GIT quotient with respect to the linearization χθ, Mθ v,w(ξ) := µ′−1(ξ) ∩ Rep � �Γ, v, w �θ-ss // GLv .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 23 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Resolutions of Zariski closures of adjoint orbits as Naka- jima’s framed quiver varieties In this section we recall the construction of resolutions of closures of adjoint orbits as Nakajima’s framed quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Those results come from Kraft-Procesi [KP81], Nakajima [Nak98, Nak01], Crawley-Boevey [CB03a, CB03b], Shmelkin [Shm09] and Letellier [Let11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this subsection and in 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3, we fix the base field K = C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Oµ,σ be an adjoint orbit with semisimple part σ and Jordan type µ ∈ Pν as in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the resolution �YL,P,σ → Oµ,σ as in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There is a Nakajima’s framed quiver variety realizing this resolution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let d := �l i=1 µi 1 and recall that L ∼= l� i=1 µi 1 � r=1 GLµir ′ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The indices � µi r ′� 1≤i≤l 1≤r≤µi 1 are relabelled (cs)1≤s≤d so that L ∼= d � s=1 GLcs .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Introduce the parameter ζ = (ζs)1≤s≤d such that ζs = σi if cs corresponds to µi r ′ for some r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the quiver ΓOµ,σ of type Ad−1 with summit indexed by integers between 1 and d − 1 and arrows going in the decreasing direction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Introduce the dimension vector vOµ,σ := (v1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', vd−1) with v1 := n − c1, vi := vi−1 − ci for i > 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' and w := (n, 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define the parameter ξOµ,σ = (ξ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', ξd−1) by ξi = ζi − ζi+1 so that ξi := � σk − σk+1 if i = µ1 1 + · · · + µk 1 0 otherwise .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The parameter ξOµ,σ will be identified with the element (ξj Idvj)1≤j≤d−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We summarize everything in the following diagram showing the quiver, the di- mension vector, the parameter ζ and the parameter ξ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 1 2 · · µ1 1+···+µk 1 · · d−1 n − c1 n − c1 − c2 · · n − ν1 − · · · − νk · · cr σ1 σ1 · · σk · · σr 0 0 · · σk − σk+1 · · 0 Remark 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When writing the dimension vector under the quiver, we used the fact that |µi| = νi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 24 Consider a second dimension vector w = (n, 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , 0) and an extended represen- tation (a, b, φ) ∈ Rep � �ΓOµ,σ, vOµ,σ, w � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As wi = 0 unless i = 1, the component a is just a linear map a : V1 → W1 and b : W1 → V1 with W1 = Cn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For 1 ≤ i ≤ d − 2, denote by φi+1,i the linear map associated to the edge from i + 1 to i and by φi,i+1 the map associated to the reverse edge from i to i+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Such a representation belongs to µ′−1(ξOµ,σ) if and only if \uf8f1 \uf8f2 \uf8f3 φ2,1φ1,2 − ba = (ζ1 − ζ2) Idv1 φi+1,iφi,i+1 − φi−1,iφi,i−1 = (ζi − ζi+1) Idvi for 2 ≤ i ≤ d − 2 −φd−1,d−2φd−1,d−2 = (ζd−1 − ζd) Idvd−1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (21) Those equations are called the preprojective relations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Example 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The adjoint orbit of the matrix \uf8eb \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ed σ1 1 0 0 0 0 0 σ1 1 0 0 0 0 0 σ1 0 0 0 0 0 0 σ1 0 0 0 0 0 0 σ2 0 0 0 0 0 0 σ2 \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f8 has Jordan type µ = ((3, 1), (1, 1)) ∈ P4 × P2 and we obtain W1 V1 V2 V3 vOµ,σ : 4 3 2 ζ : σ1 σ1 σ1 ξOµ,σ : 0 0 σ1 − σ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' b a φ1,2 φ2,1 φ2,3 φ3,2 Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First consider the Nakajima’s framed quiver variety M0 vO,w(ξOµ,σ) obtained from the previous data and stability parameter θ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following map is well-defined and is an isomorphism Ψ0 : M0 vOµ,σ ,w � ξOµ,σ � → Oµ,σ (a, b, φ) �→ ab − σ1 Idn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now take a stability parameter θ ∈ Zd−1 >0 , the following map is an isomorphism Ψθ : Mθ vOµ,σ,w � ξOµ,σ � → �YL,P,σ (a, b, φ) �→ (ab + σ1 Idn, fa,b,φ) , 25 where fa,b,φ is the flag 0 ⊂ Ed−1 ⊂ · · · ⊂ E1 ⊂ Cn defined by E1 := Im(a), Ei := Im(a ◦ φ2,1 ◦ φ3,2 ◦ · · · ◦ φi,i−1) for 2 ≤ i ≤ d − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover, the following diagram commutes Mθ vOµ,σ ,w � ξOµ,σ � �YL,P,σ M0 vOµ,σ ,w � ξOµ,σ � Oµ,σ, Ψθ π pσ Ψ0 where pσ is the resolution of Oµ,σ from Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5 and π is the natural map from GIT theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Comet-shaped quiver varieties As in the previous subsection, we fix the base field K = C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Oµ,σ = � Oµ1,σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Oµk,σk � be a generic k-tuple of adjoint orbits in gln.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We recall Crawley-Boevey’s result re- lating the variety QOµ,σ to a quiver variety.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The idea is to glue together k quivers of type A corresponding to each adjoint orbit Oµj,σj to a central vertex 0 and to add g loops to this central vertex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We obtain the following comet-shaped quiver ΓOµ,σ, [1,1] [1,2] · · [1,d1−1] [2,1] [2,2] · · [2,d2−1] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [k,1] [k,2] · · [k,dk−1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The j-th leg is a quiver of type A with vertices labelled from [j, 1] to [j, dj − 1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The dimension vector vOµ,σ is defined such that its coordinate at the central vertex is n and its coordinates on the j-th leg coincide with the dimension vector vOµj ,σj described in the previous section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly, the parameter ξOµ,σ is defined such that its coordinates on the j-th leg coincide with the parameter ξOµj,σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The component at the central vertex ξOµ,σ,0 is defined such that vOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='ξOµ,σ = 0 hence nξOµ,σ,0 = − k � j=1 dj−1 � i=1 vOµ,σ,[j,i]ξOµ,σ,[j,i].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider a representation of the extended quiver φ ∈ Rep � �ΓOµ,σ, vOµ,σ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by φ[j,i] the linear map associated to the arrow with tail [j, i] and φ[j,i] the linear map associated to the reversed arrow with head [j, i].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 26 For 1 ≤ i ≤ g the map associated to the i-th loop is denoted φi and the one associated to the reverse loop is denoted φi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As usual µ is the moment map and ξOµ,σ is identified with an element in the center of the Lie algebra gvOµ,σ and we let Xj := φ[j,1]φ[j,1] − ζ[j,1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' If φ belongs to µ−1(ξOµ,σ) then Xj ∈ Oµj,σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indeed it follows from the previous description of closures of adjoint orbits as framed quiver varieties and the identifica- tion, for each leg, of the vector space at the central vertex with the framing vector space W1 from the previous section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now if Ai is the linear map associated to the i-th loop of the quiver and Bi the map associated to the reversed loop, the preprojective relation at the central vertex is exactly the equation defining VOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the following map is well-defined ΨOµ,σ : µ−1(ξOµ,σ) → VOµ,σ φ �→ (A1, B1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ag, Bg, X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Xk) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In the following diagram where the vertical arrows are quotient maps, the application ΨOµ,σ goes down to the quotient to an isomorphism ΦOµ,σ, µ−1(ξOµ,σ) VOµ,σ M0 vOµ,σ(ξOµ,σ) QOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ΨOµ,σ ΦOµ,σ Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is proved by Crawley-Boevey [CB01, CB03b], see also Letellier [Let11, Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2] for any genus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The resolution �QL,P ,σ of QOµ,σ, as introduced in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12, is also interpreted as a Nakajima’s quiver variety for the quiver ΓOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider a stability parameter θ associated to the quiver QOµ,σ such that θ[j,i] > 0 for each vertex [j, i].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There is an isomorphism ΦOµ,σ,θ : Mθ vOµ,σ(ξOµ,σ) → �QL,P ,σ and the following diagram commutes, Mθ vOµ,σ(ξOµ,σ) � QL,P ,σ M0 vOµ,σ(ξOµ,σ) QOµ,σ, ΦOµ,σ,θ π pσ ΦOµ,σ where π is the natural projection from GIT theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The map ΦOµ,σ,θ is constructed by Letellier [Let11, Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This map is induced by the map Ψθ of Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Contrarily to Letellier’s article, we do not consider partial resolutions so that our parameter θ only has non-zero components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore the dimension vector for the quiver variety Mθ vOµ,σ(ξOµ,σ) describing the resolution �QL,P ,σ is the same as the dimension vector of the quiver variety describing QOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 27 The quiver variety point of view gives a criteria for non-emptiness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The question of emptiness of QOµ,σ (and its analogous character variety) is known as the Deligne- Simpson problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' See Kostov [Kos04] for a survey about this problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Crawley- Boevey gave a solution to the problem in the generic case in terms of roots of quivers [CB03b], see also Letellier [Let11, Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Those results are summarized in the following theorem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Oµ,σ be a generic k-tuple of adjoint orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variety QOµ,σ is non-empty if and only QOµ,σ is not empty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This happens if and only if the dimension vector vOµ,σ is a root of the quiver ΓOµ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This is always the case for g > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 Family of comet-shaped quiver varieties Now the field K is again either C or Fq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When the eigenvalues σ are varying, the varieties �QL,P ,σ fit in a family.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First we describe explicitly this family, then we give an interpretation in terms of Nakajima’s quiver varieties and moment map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' From now on the pair L, P is fixed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For short, let Z(l) := Z(l1) × · · · × Z(lk).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by B the subset of elements σ ∈ Z(l) such that the k-tuple of adjoint orbits Oµ,σ is generic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Note that the genericity condition depends only the semisimple part σ and not on the type µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then B is either empty or Zariski open in the hyperplane of Z(l) defined by the vanishing of the sum of the traces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10 (Family of varieties � QL,P ,σ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define �VL,P := �� σ, (Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k � �� σ ∈ B, and (Ai, Bi)1≤i≤g, (Xj, gjP j)1≤j≤k ∈ VL,P,σ � , �QL,P := �VL,P // GLn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by η the natural map η : �QL,P → B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The varieties �QL,P ,σ = η−1(σ) fit in a family �QL,P over B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The choice of L determines a unique quiver ΓOµ,σ and a unique dimension vector vOµ,σ independent of the choice of σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Assume that the dimension vector is indivisible so that B is not empty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then we can make the following assumption: Assumption 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11 (Genericity of the stability parameter θ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The stability parameter θ is generic, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', it is a stability parameter for the quiver ΓOµ,σ with dimension vector vOµ,σ such that θ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='vOµ,σ = 0 and θ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='v ̸= 0 for a smaller dimension vector v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The construction of Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7 extends to this family.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It provides the following commutative diagram (the left vertical arrows is induced by the moment map µ) µ−1(zgen vOµ,σ)θ-ss//GvOµ,σ �QL,P zgen vOµ,σ B, Φ η (22) 28 where θ is a fixed generic stability parameter and zgen vOµ,σ is the subset of the center of the Lie algebra gvOµ,σ corresponding to the subset B under the correspondence between the parameters ξOµ,σ and the eigenvalues σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Note that the correspondence between parameters of the quiver variety ξOµ,σ ∈ Z(gvOµ,σ) and Z(l) is not bijective, only difference of successive eigenvalues appear in the construction of the quiver variety.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Thus the previous diagram relies on a choice of k − 1 eigenvalues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' To σ ∈ Z(l) associate the element (ξOµ,σ, σ1 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk−1 1 ) in Z(gvOµ,σ) × Kk−1 this defines a bijective map h : Z(l) ∼−→ zvOµ,σ × Kk−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (23) Note that for a given parameter ξOµ,σ the genericity conditions is independant of the choice of the k − 1 eigenvalues, namely h−1(ξOµ,σ, σ1 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk−1 1 ) is generic if and only if h−1(ξOµ,σ, 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , 0) is generic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore Diagram (22) can be modified in order to account for various choices of eigenvalues, then the horizontal arrows are isomorphisms and Kk−1 × µ−1(zgen vOµ,σ)θ-ss//GvOµ,σ �QL,P Kk−1 × zgen vOµ,σ B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Φ Id ×µ η (24) Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' If K = C, or if K = Fq and the characteristic is large enough, the cohomology sheaves Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ are constant sheaves.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When K = C this follows from the quiver variety point of view of Diagram (24), it is a well-known fact used by Nakajima [Nak94] to construct a Weyl group action on the cohomology of quiver varieties (see also [Bal20]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' To prove the result for K = Fq we can change characteristic as in [HLRV13, proof of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This implies the result in large enough characteristic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6 Monodromic Weyl group action The goal of this section is to construct and study the monodromic Weyl group action on the cohomology of the quiver varieties �QL,P ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We use technics from Nakajima [Nak94], Lusztig (see Letellier [Let05, Proof of proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3]), Mellit [Mel19, Section 8] and Hausel–Letellier–Rodriguez-Villegas [HLRV13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The construction relies essentially on Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Family of resolutions of closures of adjoint orbits In this section we study a family formed by the varieties �YL,P,σ when σ is varying.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It will be usefull in the next section to study a family of comet-shaped quiver varieties and to obtain some dimension estimates to prove Lemma 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let P be a parabolic subgroup of GLn and let L be a Levi factor of P, then L is isomorphic to a group of block diagonal matrices GLc1 × · · ·×GLcr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Lie algebra of L and of UP are denoted by l respectivly by uP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' At the level of the Lie algebras the Levi decomposition reads p = l ⊕ uP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The center of this Lie algebra l is denoted 29 by Z(l) and its regular locus is Z(l)reg = {x ∈ Z(l) |ZG(x) = L} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define �Yreg L,P = � (x, gL) ∈ gln × GLn /L ��g−1xg ∈ Z(l)reg� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the projection on the first factor preg : �Yreg L,P → gln, denote by Yreg L,P its image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This image consists of semisimple elements with r distinct eigenvalues with multiplicities c1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , cr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the relative Weyl group WGLn(L) = NGLn(L)/L, and for each w ∈ WGLn(L) chose a representative ˙w ∈ NGLn(L).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This relative Weyl group acts on Z(l) by w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ := ˙wσ ˙w−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the fiber product Z(l)reg Yreg L,P ×Z(l)reg/WGLn(L) Z(l)reg Z(l)reg/WGLn(L) Yreg L,P, χ with χ the characteristic polynomial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Note that the following map is an isomorphism �Yreg L,P → Yreg L,P ×Z(l)reg/WGLn(L) Z(l)reg (x, gL) �→ (x, g−1xg).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (25) Therefore the WGLn(L)-action on Z(l)reg induces an action on �Yreg L,P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is given explicitly by w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='(x, gL) = (x, g ˙w−1L).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the morphism �Yreg L,P preg −−→ Yreg L,P is a Galois cover with group WGLn(L).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This relative Weyl group acts on the push forward of the constant sheaf preg ∗ κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define �YL,P = � (x, gP) ∈ gln × GLn /P ��g−1xg ∈ Z(l) ⊕ uP � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' An element gP ∈ GLn /P is identified with a partial flag 0 = Er ⊂ Er−1 ⊂ · · · ⊂ E1 ⊂ Kn such that dim Ei−1/Ei = ci for all 1 ≤ i ≤ r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indeed GLn acts transitively on such flags and the stabilizer of any of them is isomorphic to P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then a point (x, gP) in �YL,P consists of an endomorphism x ∈ gln and a partial flag gP preserved by x such that x acts as a scalar on Ei−1/Ei for all 1 ≤ i ≤ r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by YL,P the image of the projection to the first factor p : �YL,P → gln.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Note that the map p is proper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following theorem is a particular case of [Lus84, Lemma 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 and Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It can be seen as a generalization of Borho– MacPherson [BM83] approach to Springer theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 30 Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The subvariety Yreg L,P is open, smooth and dense in YL,P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The fol- lowing square is cartesian �Yreg L,P �YL,P Yreg L,P YL,P, i preg p (26) with i the map (x, gL) → (x, gP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover p!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ = IC• YL,P , preg !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ so that WGLn(L) acts on p!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The morphism preg is a Galois cover and i is an open embedding so that the dimension can be easily computed dim YL,P = dim �YL,P = dim �Yreg L,P = dim GLn − dim L + dim Z(L).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (27) Let us describe the relation with the resolutions of closures of adjoint orbits introduced in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let σ ∈ Z(l) and let M := ZGLn(σ) be the stabilizer of σ in GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The notations from 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 are used so that M ∼= GLν for ν a partition of n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover L ⊂ M and the integers (c1, c2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , cr) are relabelled (µ1 1 ′, µ1 2 ′, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' ) so that µi′ is a partition of νi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The inclusion L ⊂ M comes from the inclusions GLµi 1 ′ × · · · × GLµi li ′ ⊂ GLνi .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The resolution of the closure of Oµ,σ fits in the following diagram �YL,P �YL,P,σ YL,P Oµ,σ = � ρ⪯µ Oρ,σ p pσ (28) The decomposition Oµ,σ = � ρ⪯µ Oρ,σ actualy comes from a decomposition of YL,P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define Y M,ρ L,P := � σ′∈Z(m)reg Oρ,σ′.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This decomposition is similar to the one introduced by Shoji [Sho88].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variety Y M,ρ L,P is smooth of dimension dim Y M,ρ L,P = dim Oρ,σ + dim Z(m).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variety YL,P admits the following decomposition YL,P = � M � ρ⪯µ Y M,ρ L,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The first union is over the set of stabilizers of elements σ ∈ Z(l).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In the second union, µ depends on M as previously described.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The unique part indexed by M = L is Yreg L,P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 31 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by Zρ the stabilizer in GLn of the element Jρ,σ in Oρ,σ (see Notations 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There is a finite cover Z(m)reg × GLn /Zρ → Y M,ρ L,P � σ′, gZρ � �→ gJρ,σ′g−1, hence Y M,ρ L,P is smooth and dim Y M,ρ L,P = dim Oρ,σ + dim Z(m).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Decomposition of the family QL,P In this section we study a family related to the family �QL,P introduced briefly in 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We compute some dimensions which will be useful to prove Lemma 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notations 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First we recall the notations from 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 in this context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For 1 ≤ j ≤ k, �YLj,P j := � (X, gjP j) ∈ gln × GLn /P j ��g−1 j Xgj ∈ Z(lj) ⊕ uP j � and define �YL,P := �YL1,P 1 × · · · × �YLk,P k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then YL,P is the image in glk n of the map p forgetting the partial flags gjP j, p : �YL,P → glk n (Xj, gjP j)1≤j≤k �→ (Xj)1≤j≤k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly VL,P, respectively QL,P, is obtained from �VL,P , respectively � QL,P , by for- getting the partial flags.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this section a decomposition of the family QL,P is deduced from the decom- position Oµ,σ = � ρ⪯µ Oρ,σ and from the decomposition introduced in Proposition 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4, YL,P = � M � ρ⪯µ Y M,ρ L,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The decomposition is used in the next section (Lemma 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8) in order to define a Weyl group action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let YB L,P be the subset of elements in YL,P with generic semisimple parts, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', with a k-tuple of semisimple parts belonging to B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The set YB L,P is assumed to be non-empty.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The dimension of YB L,P is computed similarly to dim YL,P in Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3, dim YB L,P = kn2 + dim B − k � j=0 dim Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The decomposition YL,P = � M � ρ⪯µ Y M,ρ L,P induces a similar decomposition for YB L,P , YB L,P = � M � ρ⪯µ YB,M,ρ L,P , 32 where M = (M1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Ml) and YB,M,ρ L,P is the subset of elements in Y M1,ρ1 L1,P 1 × · · · × Y Mk,ρk Lk,P k with a generic k-tuple of semisimple parts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' From the computation of the dimension of Y M,ρ L,P in Proposition 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4, we deduce that when Z(m) ∩ B is not empty dim YB,M,ρ L,P = n � j=1 dim Oρj,σj + dim Z(m) ∩ B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (29) Now the decomposition of YB L,P induces a decomposition of the family of quiver varieties QL,P and we define QM,ρ L,P := � VL,P ×YB L,P YB,M,ρ L,P � // PGLn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variety QL,P admits the following decomposition QL,P = � M � ρ⪯µ QM,ρ L,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' When non-empty, the dimension of a part is dim QM,ρ L,P = n2(2g − 2) + 2 + dim Z(m) ∩ B + k � j=1 dim Oρj,σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (30) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The dimension of QM,ρ L,P can be computed just like the dimension of QOµ,σ (see Hausel, Letellier, Rodriguez-Villegas [HLRV11, Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4] and Letellier [Let11, Corollary 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The computation relies on the smoothness of YB,M,ρ L,P which follows from the smoothness of Y B,Mj,ρj Lj,P j and on the expression (29) for the dimension of YB,M,ρ L,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 W-equivariant structure on the cohomology of the fibers of the family �QL,P In this section we use the family �QL,P → B in order to construct a Weyl group action on the cohomology of the varieties �QL,P ,σ for σ ∈ B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Weyl group studied in this section is W := WGLn(L1) × · · · × WGLn(Lk).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Each WGLn(Lj) is isomorphic to a symmetric group and acts on Z(lj) by permuting the distinct eigenvalues with the same multiplicities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the Weyl group W acts on B, for w = (w1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , wk) ∈ W and σ = � σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk� ∈ B, w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ := � ˙w1σ1 ˙w−1 1 , .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , ˙wkσk ˙w−1 k � , where ˙wj is a representative in GLn of wj ∈ WGLn(Lj).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the diagram B �QL,P B/W QL,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' π0 η p χ (31) 33 Thanks to the quiver variety point of view, the cohomology sheaves Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ are con- stant (Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In this section a W-equivariant structure on those cohomol- ogy sheaves is constructed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Weyl group actions on the cohomology of quiver varieties with such constant sheaves were introduced by Nakajima [Nak94].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Here we also use a method from Lusztig (see [Let05, Proof of Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3]), this method is also applied by Laumon-Letellier [LL19, Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This approach al- lows to extend the equivariant structure away from a regular locus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Mellit obtained a similar result with a different construction for character varieties [Mel19, Section 8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Before constructing the equivariant structure, let us define the regular locus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Denote by Breg the subset of regular elements, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' elements � σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk� ∈ B such that ZGLn(σj) = Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is the locus of B where the W-action is free.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Diagram (31) is pulled back to the regular locus Breg �Qreg L,P Breg/W Qreg L,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' πreg ηreg preg χreg (32) Similarly to (25), notice that Qreg L,P ×Breg/W Breg ∼= �Qreg L,P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (33) Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The cohomology sheaves Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ admit a W-equivariant structure over B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the diagram �QL,P B QL,P ×B/W B B/W QL,P η p c π0 a b χ , (34) the group W acts on QL,P ×B/W B and the morphism a is W-equivariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The variety Qreg L,P ×Breg/W Breg is smooth, dense and open in QL,P ×B/W B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The constant sheaf κ over Qreg L,P ×Breg/W Breg is W-equivariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indeed for w ∈ W we can define a morphism φw : w∗κ → κ which is the identity on the stalks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It satisfies the conditions of Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Applying the continuation principle from Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8, this W-equivariant structure extends to a W-equivariant structure on IC• QL,P ×B/W B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notice that η!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ ∼= a!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='c!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We shall see in Lemma 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8 that c!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ ∼= IC• QL,P ×B/W B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the W-equivariant structure on c!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ induces a W-equivariant structure on η!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Up to the isomorphism c!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ ∼= IC• QL,P ×B/W B, the theorem is proved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 34 It remains to prove the following lemma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lemma 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There is an isomorphism c!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ ∼= IC• QL,P ×B/W B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Because of the isomorphism (33), the restriction of c!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ to the smooth locus Qreg L,P ×Breg/W Breg is the constant sheaf κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In order to verify the hypothesis of Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7 it remains to prove that the map c is small, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' that it satisfies the following inequality dim � x ∈ QL,P ×B/W B ��dim c−1(x) ≥ d � ≤ dim QL,P ×B/W B − 2d for all d > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We use dimension estimates from Lusztig [Lus84, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2], see also [Sho88, Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In the Lie algebra gln the estimate becomes, for X in an adjoint orbit O, dim � gP ∈ GLn /P ��g−1Xg ∈ σ + uP � ≤ 1 2 � n2 − dim L − dim O � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (35) The proof is then standard in Springer theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let d > 0 and let x be such that dim c−1(x) ≥ d, the element x belongs to QOρ,σ for an element σ ∈ B and for some adjoint orbits Oρ1,σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Oρk,σk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The dimension estimate (35) implies d ≤ 1 2 � kn2 − k � j=1 dim Lj − dim Oρj,σj � , so that k � j=1 dim Oρj,σj ≤ kn2 − k � j=1 dim Lj − 2d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Using the decomposition from Proposition 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6, we have that x ∈ QB,M,ρ L,P .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The previous inequality together with the expression (30) for the dimension of QB,M,ρ L,P give dim QB,M,ρ L,P ≤ n2(2g − 2) + 2 + dim Z(m) ∩ B + kn2 − k � j=1 dim Lj − 2d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (36) Moreover dim QB,M,ρ L,P ×B/W B = dim QB,M,ρ L,P (37) and dim QL,P ×B/W B = dim QL,P = n2(2g − 2) + 2 + dim B + kn2 − k � j=1 dim Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (38) Combining (36),(37) and (38), dim QB,M,ρ L,P ×B/W B ≤ dim QL,P ×B/W B + 2d + dim Z(m) ∩ B − dim B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (39) As d is assumed to be strictly positive, necessarily the inclusion L ⊊ M is strict, so that dim Z(m) ∩ B < dim B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (40) 35 Now (39) and (40) provide the estimate dim QB,M,ρ L,P ×B/W B < dim QL,P ×B/W B − 2d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (41) To conclude, the set � x ∈ QL,P ×B/W B |dim c−1(x) ≥ d � is a finite union of varieties QB,M,ρ L,P ×B/W B with dimensions satisfying the previous estimate (41).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let us study the restriction of the W-equivariant sheaves Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ to the regular locus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Recall that Qreg L,P ×Breg/W Breg ∼= �Qreg L,P, then for σ ∈ Breg Hi ση!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ ∼= Hi c( �QL,P ,σ, κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For w ∈ W, the W-equivariant structure is given by the functoriality of the compactly supported cohomology (see Proposition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6 and Remark 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3) w∗ : Hi c � �QL,P ,w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ, κ � → Hi c � �QL,P ,σ, κ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore the construction of Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7 gives a canonical extension over B of this natural Weyl group action over Breg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 Monodromic Weyl group action on the cohomology of �QL,P ,σ We saw in 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12 that the cohomology sheaves Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ are constant sheaves over B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Together with the W-equivariant structure, this allows to construct a Weyl group action on the cohomology of the varieties �QL,P ,σ for any σ ∈ B, this is called the monodromic Weyl group action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Note that the fiber over σ of this constant sheaf is Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Thus for any σ, τ ∈ B, there is an isomorphism fσ,τ : Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) → Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ), such that for any ω ∈ B fσ,τ = fω,τ ◦ fσ,ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The W-equivariance of the local system Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ implies the following theorem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It can also be proved directly, without referring to equivariance of the local system (see Maffei [Maf02, Section 5]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let σ, τ ∈ B, then the following diagram commutes Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,w−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,w−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' w∗ fσ,τ fw−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ,w−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='τ w∗ Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Note that if σ ∈ B is not regular, then the map w∗ : Hi c � �QL,P ,σ, κ � → Hi c � � QL,P ,w−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ, κ � is only the map coming from the W-equivariant structure of the constant sheaf Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It does not necessarily come by functoriality from a morphism of variety.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 36 This theorem allows to define a W-action on the compactly supported cohomol- ogy space Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For σ ∈ B and for w ∈ W let ρi(w) = fw.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ,σ ◦ (w−1)∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This defines an action of W on Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ), it is called the monodromic Weyl group action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For w1 and w2 in W, the following diagram commutes by Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,w2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,w1w2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,w1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) (w−1 2 ) ∗ (w−1 1 ) ∗ fw2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ,σ fw1w2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ,w1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ (w−1 1 ) ∗ fw1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ,σ Going from the top left corner to bottom right corner by the top right corner is ρ(w1w2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Going by the middle gives ρ(w1)ρ(w2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore ρ(w1w2) = ρ(w1)ρ(w2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5 Frobenius morphism and monodromic action The techniques in this section come from Hausel, Letellier and Rodriguez-Villegas [HLRV13], however we do no consider regular semisimple values of the moment map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Instead each component of the moment map is central and each leg of the comet-shaped quiver corresponds to a particular adjoint orbit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Comet-shaped quiver varieties were also studied in this context by Letellier [Let12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A slightly more general situation is considered here, as a leg can represents any adjoint orbit and not only a semisimple regular one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The representation defined in Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12 when K = C is isomorphic to the representation obtained for K = Fq and large enough characteristic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indeed this can be proved by base change exactly like in [HLRV13, Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore from now on we assume: Assumption 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' K = Fq and the characteristic is large enough.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This assumption is very convenient as it allows to introduce Frobenius endo- morphism and use Grothendiek’s trace formula to compute the traces of the action obtained.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote by F the Frobenius endomorphism on gln raising the coefficients to the power q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The set of F-fixed points in gln is gln(Fq) and similarly for the group GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Assume that the Lj are subgroups of bock diagonal matrices, and that the P j are subgroups of block upper triangular matrices, then they are F-stable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The morphism F induces a Frobenius endomorphism on �Qreg L,P and on Breg also denoted by F, F � σ, (Ai, Bi)1≤i≤g , (Xj, gjLj)1≤j≤k � = � F(σ), (F(Ai), F(Bi))1≤i≤g , (F(Xj), F(gj)Lj)1≤j≤k � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 37 This Frobenius endomorphism can be twisted by an element w = (w1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , wk) in the Weyl group W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For σ ∈ Breg, define wF(σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk) := (w1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='F(σ1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , wk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='F(σk)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The set of points fixed by wF is (Breg)wF.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Similarly, the w-twisted Frobenius on �Qreg L,P is wF := w ◦ F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They are compatible, preg ◦ wF = wF ◦ preg so that for σ, τ ∈ Breg the following diagram commutes Hi c( �QL,P,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,F −1(σ);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,F −1(τ);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' F ∗ fσ,τ fF −1(σ),F −1(τ) F ∗ Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For τ ∈ (Breg)F and for σ ∈ (Breg)wF, the cardinal of the set of wF fixed points in �QL,P ,σ is ♯ �QwF L,P ,σ = � i tr � ρ2i(w), H2i c ( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) � qi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the commutative diagram Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,w−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,F (σ);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) Hi c( �QL,P,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' w∗ ρ(w−1) fw−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='τ,τ F ∗ fσ,τ w∗ fF (σ),τ F ∗ fσ,τ Apply Grothendieck trace formula to wF, ♯ �QwF L,P ,σ = � i (−1)i tr � (wF)∗, Hi c( �QL,P ,σ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) � = � i (−1)i tr � F ∗ ◦ ρi(w−1), Hi c( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The varieties QL,P,τ are pure and polynomial count (see Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='13 and [HLRV11, Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1]) and ρ(w−1) commutes with F so that ♯ �QwF L,P ,σ = � i tr � F ∗ ◦ ρ2i(w−1), H2i c ( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) � = � i tr � ρ2i(w−1), H2i c ( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) � qi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now as W is isomorphic to a product of symmetric groups, w is conjugated to its inverse w−1 and ♯ �QwF L,P ,σ = � i tr � ρ2i(w), H2i c ( �QL,P,τ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' κ) � qi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 38 Notations 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The j-th part of L is Lj ∼= GLcj 1 × · · · × GLcj 1 � �� � mj 1 × · · · × GLcj lj × · · · × GLcj lj � �� � mj lj , with cj r ̸= cj s for r ̸= s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the j-th part of the relative Weyl group W is WGLn(Lj) ∼= Smj 1 × · · · × Smj lj The symmetric group Smj r acts by permuting the blocks of size cj r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Take w = (w1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , wk) in W and choose σw = (σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk) in (Breg)wF.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The conjugacy class of the element wj is determined by a lj-tuple (ηj,1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , ηj,lj) with ηj,r ∈ Pmj r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let Oσj be the adjoint orbit of σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This orbit is semisimple, F-stable and of the following type (as defined in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3), � ηj,1 1 , 1cj 1 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � ηj,1 l(ηj,1), 1cj 1 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � η j,lj 1 , 1 cj lj � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' � η j,lj l(ηj,lj ), 1 cj lj � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define Ow := (Oσ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Oσk).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lemma 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' With the previous notations we have the following identity between cardinals, ♯ �QwF L,P ,σ = ♯QF Ow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (42) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As the orbits Ow = (Oσ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Oσk) are semisimple (hence they are closed), the map �QL,P ,σ → QOw is an isomorphism compatible with the Frobenius wF on the source and the Frobenius F on the target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier [Let11] computed the number of points of comet shaped quiver varieties, in particular of QF Ow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' With Notations 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15, the cardinal of QF Ow is given by ♯QF Ow = (−1)r(η)q dµ 2 � �hη, HHLV n (0, q 1 2) � , where �hη is a particular case of the generalized Schur function from [Let11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This symmetric function can be expressed in terms of complete symmetric functions hn, �hη := k � j=1 lj � r=1 l(ηj,i) � s=1 hcj r � Xηj,r s j � , and r(η) := k � j=1 lj � r=1 cj r l(ηj,i) � s=1 (ηj,r s − 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As the orbits Oσj are semisimple, the variety QOw is smooth so that the characteristic function of the intersection complex is constant with value 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The result follows from Letellier [Let11, Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1, Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 and Corollary 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 39 Corollary 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For σ ∈ B and η representing a conjugacy class in the Weyl group as described in Notations 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15, the η-twisted Poincaré polynomial of �QL,P ,σ is � i tr � η, Hi c( �QL,P ,σ, κ) � vi = (−1)r(η)vdµ � �hη, HHLV n (0, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The action comes from the W-equivariant structure of the constant sheaves Hiη!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='κ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore, up to isomorphism, the representation does not depend on the choice of σ ∈ B so that the twisted Poincaré polynomial can be computed for τ ∈ (Breg)F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then from Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14 and from (42), � i tr � ρ2i(η), H2i c � �QL,P ,τ, κ �� qi = (−1)r(η)q dµ 2 � �hη, HHLV n (0, q 1 2) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This equality remains true after substituting qn for q for n > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Thus it is an equality between two polynomials and the corollary is proved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='19 (Comparison between monodromic and Springer action).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let σ = (σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σk) ∈ B, as before Mj is the stabilizer of σj in GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The relative Weyl group is WM(L) = k � j=1 WMj(Lj) with WMj(Lj) = NMj(Lj)/Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then WM (L) is a subgroup of the Weyl group W studied in this section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The group WM (L) is exactly the subgroup of elements w ∈ W such that w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='σ = σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The monodromic Weyl group action from Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12 induces an action of WM (L) on Hi c � �QL,P ,σ, κ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Interestingly, this action comes only from the W-equivariant structure, it does not rely on the constant property of the sheaf: it is given explicitly by ρi(w) = (w−1)∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There is another action of WM (L) on Hi c � �QL,P ,σ, κ � , the Springer action con- structed by Letellier and mentioned in 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier computed the twisted Poincaré polynomial for this Springer action [Let11, Corollary 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3], it coincides with the Poincaré polynomial obtained from the monodromic action, therefore both action are isomorphic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It would be interesting to have a direct proof of this fact.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We proved it in the character variety setting for just one orbit, regular, with a unique eigenvalue [Bal21, Chapter 5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is also interesting to consider the monodromic action over the regular locus Breg as an action on the cohomology of a quiver variety with semisimple adjoint orbits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For σ ∈ Breg consider the associated generic k-tuple of semisimple adjoint orbits Oσ = (Oσ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , Oσk).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Weyl group WGLn(Lj) is the group of permutation of the distinct eigenvalues of Oσj with the same multiplicities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This provides another formulation of Corollary 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Corollary 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For η representing a conjugacy class in the Weyl group as described in Notations 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15, the η-twisted Poincaré polynomial of QOσ is � i tr � η, Hi c(QOσ, κ) � vi = (−1)r(η)vdη � �hη′, HHLV n (0, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 40 The interpretation over the regular locus in terms of semisimple quiver varieties together with Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='19 show the advantages of extending the W-equivariant structure from Breg to B (Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This provides a uniform description of the Springer action on the cohomology of some resolution �QL,P ,σ and the monodromic action on the cohomology of semisimple quiver varieties QOσ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is also interesting to study the action of a Weyl group relative to a particuar leg 1 ≤ j ≤ k, for instance relative to the first one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This will be used in 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 to describe some structure coefficients of the algebra spanned by Kostka polynomial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A particularly interesting case is when L1 is a maximal torus and M1 = GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the component of the Weyl group relative to the first leg is WM1(L1) ∼= Sn and WM(L) ∼= Sn × k � j=2 WMj(Lj).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' According to this decomposition, consider an element (w, 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , 1) ∈ WM(L) with w ∈ Sn an element of cycle type λ ∈ Pn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then �hη = pλ[X1]hµ′2[X2] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' hµ′k[Xk], and (−1)r(η) = ǫ(λ) is the sign of the permutation w with cycle type λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Corollary 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18 reads P η c � �QL,P ,σ, v � = vdµǫ(λ) � pλ[X1]hµ′2[X2] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' hµ′k[Xk], HHLV n (0, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This can be understood in terms of Frobenius characteristic, see Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the representation of Sn on the cohomology of �QL,P ,σ twisted by the sign, H•( �QL,P ,σ, κ) ⊗ ǫ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Its v-graded Frobenius characteristic is given by the following symmetric function in X1 vdµ � hµ′2[X2] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' hµ′k[Xk], HHLV n (0, v) � X2,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=',Xk .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notice that Vρ ⊗ ǫ ∼= Vρ′, then by Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11, the multiplicity of the irreducible representation Vρ in H•( �QL,P ,σ, κ) is given by vdµ � sρ′[X1]hµ′2[X2] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' hµ′k[Xk], HHLV n (0, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 7 Geometric interpretations in the algebra spanned by Kostka polynomials 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1 Description of the algebra In this section an algebra spanned by Kostka polynomials is studied and some struc- ture coefficients are related to traces of Weyl group action on the cohomology of quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Define a linear map ∆# : Sym[X] → Sym[X, Y ] such that on the basis of modified Macdonald polynomials, ∆# � ˜Hλ[X] � := ˜Hλ[X] ˜Hλ[Y ] for all λ ∈ P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 41 As in 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='20, the variables (q, t) are implicit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now as the Hall pairing is non-degenerate, there is a uniquely determined bilinear map .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' # .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' such that for all F, G and H in Sym[X], ⟨F[X]#G[X], H[X]⟩ = � F[X]G[Y ], ∆# (H[X]) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The product # defines an associative and commutative algebra structure on Sym[X].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For a k-tuple of partitions µ = � µ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , µk� ∈ Pk n and for λ ∈ Pn we denote by cλ µ the structure coefficients of the product # in the basis of Schur functions sµ1#sµ2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' #sµk = � |µ|=n cλ µsλ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (43) Lemma 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For µ = (µ, ν), the coefficients cλ µ,ν coincide with those defined in the introduction, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', the following relation is satisfied �Kµ,ρ �Kν,ρ = � λ cλ µ,ν �Kλ,ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (44) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First let � �Lη,λ � λ,η∈Pn be the inverse of the matrix of Kostka polynomials � �Kη,λ � λ,η∈Pn (see Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='19), then sλ = � η∈Pn �Lη,λ ˜Hη[X].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now the coefficient cλ µ,ν is defined by cλ µ,ν = ⟨sµ#sν, sλ⟩ = � sµ#sν, � η∈Pn �Lη,λ ˜Hη[X] � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' By definition of the product # and of the coproduct ∆#, cλ µ,ν = � η∈Pn �Lη,λ � sµ[X]sν[Y ], ˜Hη[X] ˜Hη[Y ] � , cλ µ,ν = � η∈Pn �Lη,λ �Kµ,η �Kν,η.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Multiply the last equation by �Kλ,ρ and sum over λ ∈ Pn, �Kµ,ρ �Kν,ρ = � λ cλ µ,ν �Kλ,ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This last relation is exactly the one used in the introduction to define the coefficients cλ µ,ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Example 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We computed some coefficients with the software SageMath c(2,1,1) (2,2),(2,1,1) = −q3t − q2t2 − qt3 − q2t − t2q + q2 + qt + t2, c(1,1,1,1) (2,2),(2,1,1) = q3 + q2t + qt2 + t3 + q2 + 2qt + t2 + q + t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 42 The next conjecture comes from unpublished notes by Rodriguez-Villegas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Conjecture 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The structure coefficients cλ µ lie in Z[q, t].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Some evidences supporting this conjecture will be provided.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following defi- nition and remark were suggested by François Bergeron.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Definition 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let F be a symmetric function, consider the operator F# .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' : Sym [X] → Sym [X] G �→ F#G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We denote ψF its adjoint with respect to the Hall pairing so that for any G, H ∈ Sym [X] ⟨F#G, H⟩ = ⟨G, ψF(H)⟩ (45) Those operators are diagonal in the basis of modified Macdonald polynomials ψF( ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t]) = � F, ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] � ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (46) Remark 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Applying the relation (46) with en, ψen � ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] � = qn(λ′)tn(λ) ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t], and we recognize the usual expression of the operator ∇ introduced by Bergeron- Garsia [BG98].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The higher (q, t)-Catalan sequence from Garsia–Haiman [GH96] (see also Haiman [Hai02, p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='95]) is defined by C(m) n (q, t) = ⟨en, ∇men⟩ , but ∇ = ψen is the adjoint of en# .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' and s1n = en so that C(m) n (q, t) = c1n 1n, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , 1n � �� � m+1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The higher (q, t)-Catalan sequences are particular cases of the coefficients c1n µ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We recall an important theorem which was first conjectured by Garsia–Haiman [GH96].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7 ([Hai02] theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The symmetric function ∇(en) is obtained as the Frobenius characteristic (see definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12) of a bigraded representation of Sn called the diagonal harmonics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In particular, ⟨∇(en), sλ⟩ ∈ Z≥0[q, t].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Corollary 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For any µ ∈ Pn, the structure coefficients c1n 1n,µ gives the multiplicity of the irreducible representation of type µ in the bigraded representation of Sn on diagonal harmonics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In particular c1n 1n,µ(q, t) ∈ Z≥0[q, t] so that the conjecture 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4 is true for those particular coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 43 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' According to Remark 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='6 and to the adjonction relation (45), ⟨sµ, ∇(en)⟩ = ⟨en#sµ, en⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (47) By definition of the structure coefficients cλ µ,ν and as en = s1n, we have en#sµ = � λ∈Pn cλ 1nµsλ, substituting in (47) we obtain c1n 1n,µ(q, t) = ⟨sµ, ∇(en)⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We conclude with the interpretation of ∇(en) as a Frobenius characteristic from Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The next theorem and the following corollary come from unpublished notes by Rodriguez-Villegas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' They relate particular structure coefficients c1n µ to the kernel HHLV n .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Consider the generating function from Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='16 for genus g = 0, k + 2 punctures and with variable z = q 1 2, w = t 1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' It is given by Ω0 k+2 := � λ∈P �k+2 i=1 ˜Hλ [Xi;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] aλ(q, t) s|λ|, with aλ(q, t) = � ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t], ˜Hλ[X;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' q, t] �q,t as in 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following relation holds, � p(n)[Xk+1]h(n−1,1)[Xk+2], Log � Ωg k+2 �� Xk+1,Xk+2 = � |λ|=n φλΠ′ λ aλ k � i=1 ˜Hλ[Xi]s|λ|, with φλ = � i,j∈λ qj−1ti−1, Π′ λ = � i,j∈λ\\(1,1) (1 − qj−1ti−1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' According to Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='23, to take the Hall pairing with h(n−1,1)[Xk+2] is equivalent to do plethystic substitution Xk+2 = 1 + u and to take the degree n coefficient in front of u.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' As the plethystic substitution and the plethystic logarithm commute, we can perform this substitution inside the plethystic logarithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We consider terms of order 1 in u using Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='22, Log � Ω0 k+2 � = Log � Ω0 k+1 + u � λ∈P∗ φλ aλ k+1 � i=1 ˜Hλ[Xi]s|λ| + O(u2) � = Log � Ω0 k+1 � 1 + u 1 Ω0 k+1 � λ∈P∗ φλ aλ k+1 � i=1 ˜Hλ[Xi]s|λ| + O(u2) �� = Log � Ω0 k+1 � + Log � 1 + u 1 Ω0 k+1 � λ∈P∗ φλ aλ k+1 � i=1 ˜Hλ[Xi]s|λ| + O(u2) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 44 We used that plethystic logarithm turns products into sums.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' From the definition of the plethystic logarithm, as pn[u] = un, we easily see the coefficient in front of u in the previous expression Log � Ω0 k+2 ��� u = 1 Ω0 k+1 � λ∈P∗ φλ aλ k+1 � i=1 ˜Hλ[Xi]s|λ|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Keeping the terms of degree n we obtain � h(n−1,1)[Xk+2], Log � Ω0 k+2 �� Xk+2 = 1 Ω0 k+1 � λ∈P∗ φλ aλ k+1 � i=1 ˜Hλ[Xi]s|λ| ����� sn .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Inverting Ω0 k+1 is licit, it is defined by 1 Ω0 k+1 = 1 1 + � Ω0 k+1 − 1 � = � k � 1 − Ω0 k+1 �k .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now we just have to take the Hall pairing with the power sum p(n) [Xk+1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This is equivalent to picking the coefficient in front of n−1p(n) [Xk+1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' But p(n) cannot be written as the product of two symmetric functions of degree strictly smaller than n so that the contribution of Ω0 k+1 in the denominator is irrelevant for the coefficient in front of n−1p(n) [Xk+1] and � p(n)[Xk+1]h(n−1,1)[Xk+2], Log � Ω0 k+2 �� Xk+1,Xk+2 = � p(n)[Xk+1], � λ∈P∗ φλ aλ k+1 � i=1 ˜Hλ[Xi]s|λ| � Xk+1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We conclude with Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='24 and (4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following corollary allows to obtain a geometric interpretation of the coeffi- cients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indeed, it relates the coefficient c(1n) µ to the generating serie Ω0 k+2 known to encode cohomological information about comet-shaped quiver varieties and charac- ter varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Corollary 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' With the notations of the previous theorem and Definition 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='1, (−1)n−1c(1n) µ = (q−1)(1−t) � k � j=1 sµj[Xj]p(n)[Xk+1]h(n−1,1)[Xk+2], Log � Ω0 k+2 � � X1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=',Xk+2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (48) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We apply Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='9 to express the right hand side of (48) as (q − 1)(1 − t) � sµ1[X1] .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' sµk[Xk], � |λ|=n φλΠ′ λ aλ k � i=1 ˜Hλ[Xi] � X1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=',Xk .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' By definition of the product #, (q − 1)(1 − t) � sµ1# .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' #sµk[X], � |λ|=n φλΠ′ aλ ˜Hλ[X] � X .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 45 Here we recognize the expression of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='25 � sµ1# .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' #sµk[X], (−1)n−1s(1n) � X so that if we write sµ1# .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' #sµk[X] = � λ cλ µsλ[X] the result follows from orthonormality of Schur functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2 Interpretation of certain coefficients as traces of Weyl group actions on the intersection cohomology of quiver varieties In this section a cohomological interpretation is given for the coefficients c1n µ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In order to lighten the notations, the description is only given for the coefficient c1n µ,ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The generalization to any µ is straightforward.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First let us detail the data to describe the relevant variety �QL,P ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Levi sub- groups are torus of diagonal matrices Lj = T for 1 ≤ j ≤ 3 and L4 = GL1 × GLn−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The semisimple part σ = (σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σ4) is such that: σ1 = ζ1 Id is central, σ2 = ζ2 Id is central, σ3 = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed α1 α2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' αn \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 with αr ̸= αs for r ̸= s, σ4 = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed β γ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' γ \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 has two eigenvalues β ̸= γ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The multiplicity of β is one and the multiplicity of γ is n − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notice that such a choice can be made in the generic locus, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', with σ ∈ B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First we consider Letellier’s construction of the action à la Springer in order to compute isotypical components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let M = M1 × · · · × M4 with Mj the stabilizer in GLn of σj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then WM(L) ∼= S2 n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier’s construction (Remark 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14) provides an action of WM(L) on the cohomology of �QL,P ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Moreover for Vµ′, respectively Vν′, the irreducible representation of Sn associated to the transpose of some partition µ, respectively ν, HomWM(L) � Vµ′ ⊗ Vν′, H i+d � QL,P ,σ c � �QL,P ,σ, κ �� = H i+dQO c (QO, κ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' (49) With O = (O1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , O4) the 4-tuple of generic adjoint orbits defined by, O1 has Jordan type µ′ and eigenvalue ζ1, O2 has Jordan type ν′ and eigenvalue ζ2, 46 O3 is the orbit of σ3, O4 is the orbit of σ4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now with the construction from Theorem 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12, there is an action of the whole group W ∼= S3 n on the cohomology of �QL,P ,σ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The restriction of this W-action to WM(L) ∼= S2 n is isomorphic to the Springer action (see Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='19).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First take the Vµ′ ⊗ Vν′ isotypical component with respect to the S2 n-action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There remains an action of the Weyl group Sn relative to the third leg on the intersection cohomology IHi c (QO, κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let w be a n-cycle in the Weyl group relative to the third leg (this terminology comes from the comet-shaped quiver).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The coefficient c1n µ,ν, after special- ization q = 0, is given by the w-twisted Poincaré polynomial of QO, namely c1n µ,ν(0, t) = t− dO 2 � i tr � w, IH2i c (QO, κ) � ti.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Combining (49), Corollary 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18 and Remark 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='21, � i tr � w, IHi c (QO, κ) � vi = (−1)n−1vdO � sµ[X1]sν[X2]p(n)[X3]h(n−1,1)[X4], HHLV n (0, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The theorem now follows from Corollary 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='3 Cohomological interpretation in the multiplicative case There are similar interpretations in the multiplicative case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A conjectural one in- volving c1n µ,ν(q, t) which is a theorem after specializing to c1n µ,ν(1, t).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Unfortunately in the multiplicative case the monodromic action is not defined in the general case so that we have to rely only on the Springer action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Therefore the statements involve partial resolutions of character varieties instead of actual character varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First introduce the relevant parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Levi subgroups are torus of diagonal matrices Lj = T for 1 ≤ j ≤ 3 and L4 = GL1 × GLn−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The semisimple part σ = (σ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , σ4) is such that: σ1 = ζ1 Id is central, σ2 = ζ2 Id is central, σ3 = ζ3 Id is central, σ4 = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed β γ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' γ \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 has two eigenvalues β ̸= γ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This 4-tuple is chosen to be generic (in the multiplicative sense of Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='10).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This is the case for instance if ζ1ζ2ζ3 = 1 and γn−1 = β−1 ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The relative Weyl group is WM(L) ∼= S3 n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Now consider the following conjugacy classes C1 has Jordan type µ′ and eigenvalue ζ1, 47 C2 has Jordan type ν′ and eigenvalue ζ2, C3 has one Jordan block of size n with eigenvalue ζ3, C4 is the conjugacy class of σ4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then � ML,P ,σ is the resolution of MC with C = (C1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , C4) (see Definition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' An intermediate in between � ML,P ,σ and MC is given by the variety Mµ,ν = � (X1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' , X4) ∈ C1 × · · · × C4, gB ∈ GLn /B ��g−1X3g ∈ ζ3U X1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' X4 = Id} // PGLn, with B the Borel subgroup of upper triangular matrices in GLn and U its unipotent radical.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Then the resolution � ML,P ,σ → MC factors through Mµ,ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This is a particular case of the partial resolutions of character varieties studied by Letellier [Let13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The result we recalled about Springer theory for resolutions of character varieties (16) admit a more general version for partial resolutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In particular considering the action of S2 n with respect to the first two punctures and taking the Vµ′ ⊗ Vν′ isotypical component of the cohomology H• c � � ML,P ,σ, κ � we obtain, HomWM(L) � Vµ′ ⊗ Vν′, H i+d � ML,P ,σ c � � ML,P ,σ, κ �� = H i+dMµ,ν c (Mµ,ν, κ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' There remains an action of the Weyl group Sn relative to the third puncture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' For w in this Weyl group define the w-twisted mixed Hodge polynomial by IHw c (Mµ,ν;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' u, v) := � i,r urvi tr � w, IHr,r,i c (Mµ,ν, κ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Conjecture 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18 admits a generalization describing the Weyl group action on the intersection cohomology of partial resolutions of character varieties (Letellier [Let13, Conjecture 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='5]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In particular this conjecture predicts the following formula for the w-twisted mixed Hodge polynomial for w a n-cycle, IHw c (Mµ,ν;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' u, v) = (−1)n−1 � v√u �dim Mµ,ν � sµ[X1]sν[X2]p(n)[X3]h(n−1,1)[X4], HHLV n � −1 √u, v√u �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The next conjecture follows from this conjectural formula, just like Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='11 is deduced from Corollary 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Conjecture 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let w be a n-cycle in the Weyl group relative to the third punc- ture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The coefficient c1n µ,ν relates to the w-twisted mixed Hodge polynomial of Mµ′,ν′ by c1n µ,ν(q, t) = t − dim Mµ,ν 2 IHw c � Mµ,ν, 1 q, √qt � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' We will prove that the right handside of this conjecture is indeed a polynomial in q, t thus supporting Conjecture 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 48 Lemma 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Poincaré polynomial (for compactly supported intersection co- homology) of the variety Mµ,ν is � i vi dim IHi c (Mµ,ν, κ) = vd � sµ[X1]sν[X2]h1n[X3]h(n−1,1)[X4], HHLV n (−1, v) � , (50) and for w a n-cycle in the Weyl group relative to the third puncture the w-twisted Poincaré polynomial is � i vi dim tr � w, IHi c (Mµ,ν, κ) � = vd � sµ[X1]sν[X2]p(n)[X3]h(n−1,1)[X4], HHLV n (−1, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Letellier [Let13, Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='4] studied Springer theory for partial resolutions of character varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' This allows to describe the intersection cohomology of the partial resolutions, together with its Weyl group action, in terms of intersection cohomology of character varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Poincaré polynomials for intersection coho- mology of character varieties are computed in [Bal22].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The formula for (twisted) Poincaré polynomials of the resolutions are therefore a consequence of the Poincaré polynomial specialization of [Let13, Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proposition 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The following expression t − dim Mµ,ν 2 IHw c � Mµ,ν, 1 q, √qt � , which is the value of c1n µ,ν(q, t) according to Conjecture 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12, is a polynomial in q, t with integer coefficients.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' First note that only integer powers of q and t appear because of (50) and the fact that in genus g = 0 the kernel HHLV n (z, w) contains only terms in z2, w2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let S be a regular semisimple conjugacy class such that the 4-tuple C′ = (C1, C2, S, C4) is generic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' By [Bal22] and (50), the Poincaré polynomial of the char- acter variety MC ′ is the same as the Poincaré polynomial of the variety Mµ,ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' But the variety MC ′ is affine, hence its compactly supported intersection cohomology vanishes in degree strictly smaller than its dimension and only positive power of t appear in the expression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Only positive power of q appear because the weight on the compactly supported intersection cohomology is smaller than the cohomological degree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The second equality in Lemma 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='13 implies the Poincaré polynomial specialisa- tion of Conjecture 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Theorem 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Let w be a n-cycle in the Weyl group relative to the third puncture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The coefficient c1n µ,ν relates to the w-twisted Poincaré polynomial of Mµ,ν: c1n µ,ν(1, t) = t − dim Mµ,ν 2 � i t i 2 tr � w, IHi c (Mµ,ν, κ) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' References [Bal20] Mathieu Ballandras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Trivializations of moment maps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='08294, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 49 [Bal21] Mathieu Ballandras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Weyl group actions on the cohomology of character and quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' PhD thesis, Université de Paris – Scuola Inter- nazionale Superiore di Studi Avanzati, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Bal22] Mathieu Ballandras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Intersection cohomology of character varieties for punctured riemann surfaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='08795, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [BBDG18] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Beilinson, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Bernstein, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Deligne, and O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Gabber.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Faisceaux pervers, volume 100 of Astérisque.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Société mathématique de France, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [BG98] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Bergeron and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Garsia.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Science fiction and Macdonald’s polynomials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' CRM Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lecture Notes, 22, 10 1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [BM83] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Borho and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' MacPherson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Partial resolutions of nilpotent vari- eties, Analysis and topology on singular spaces, II, III, volume 101 of Astérisque.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Société mathématique de France, 1983.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [CB01] William Crawley-Boevey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Geometry of the moment map for representa- tions of quivers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Compositio Mathematica, 126(3):257–293, 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [CB03a] William Crawley-Boevey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Indecomposable parabolic bundles and the existence of matrices in prescribed conjugacy class closures with product equal to the identity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Publications mathématiques, 100, 08 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [CB03b] William Crawley-Boevey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' On matrices in prescribed conjugacy classes with no common invariant subspace and sum zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Duke Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 118(2):339–352, 06 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [CB06] William Crawley-Boevey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Quiver algebras, weighted projective lines, and the Deligne-Simpson problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' International Congress of Mathe- maticians, ICM 2006, 2, 05 2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [GH93] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Garsia and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Haiman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A graded representation model for Mac- donald’s polynomials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proceedings of the National Academy of Sciences, 90(8):3607–3610, 1993.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [GH96] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Garsia and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Haiman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A remarkable q,t-catalan sequence and q-lagrange inversion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Journal of Algebraic Combinatorics, 5:191–244, 1996.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Hai01] Mark Haiman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Hilbert schemes, polygraphs and the Macdonald positiv- ity conjecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Journal of the American mathematical society, 14(4):941– 1006, 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Hai02] Mark Haiman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Combinatorics, symmetric functions, and Hilbert schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Current Developments in Mathematics, 2002:39–111, 2002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [HLRV11] Tamás Hausel, Emmanuel Letellier, and Fernando Rodriguez-Villegas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Arithmetic harmonic analysis on character and quiver varieties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Duke Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 160(2):323–400, 11 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 50 [HLRV13] Tamás Hausel, Emmanuel Letellier, and Fernando Rodriguez-Villegas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Positivity for Kac polynomials and DT-invariants of quivers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Ann.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' of Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 177(3):1147–1168, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Kos04] Vladimir Petrov Kostov.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Deligne–Simpson problem—a survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Jour- nal of Algebra, 281(1):83–108, 2004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [KP81] Hanspeter Kraft and Claudio Procesi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Minimal singularities in GLn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Inventiones mathematicae, 62:503–515, 1980/81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [KW01] Reinhardt Kiehl and Rainer Weissauer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Weil Conjectures, Perverse Sheaves and l’adic Fourier Transform.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Springer, 01 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Let05] Emmanuel Letellier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Fourier Transforms of Invariant Functions on Fi- nite Reductive Lie Algebras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lecture Notes in Mathematics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Springer Berlin Heidelberg, Berlin, Heidelberg, 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Let11] Emmanuel Letellier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Quiver varieties and the character ring of general linear groups over finite fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Journal of the European Mathematical Society, 15, 03 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Let12] Emmanuel Letellier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Tensor products of unipotent characters of general linear groups over finite fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Transformation Groups, 18, 04 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Let13] Emmanuel Letellier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Character varieties with Zariski closures of GLn conjugacy classes at punctures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Selecta Mathematica, 21, 09 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [LL19] Gérard Laumon and Emmanuel Letellier.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Notes on a conjecture of Braverman-Kazhdan, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Lus84] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lusztig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Intersection cohomology complexes on a reductive group.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Inventiones mathematicae, 75:205–272, 1984.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Lus85] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lusztig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Character sheaves I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Advances in Mathematics, 56(3):193– 237, June 1985.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Lus86] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lusztig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' On the character values of finite Chevalley groups at unipo- tent elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Algebra, 104:146–194, 1986.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Lus00] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Lusztig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Quiver varieties and Weyl group actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Annales de l’Institut Fourier, 50(2):461–489, 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Mac88] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Macdonald.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A new class of symmetric functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Séminaire Lotharingien de Combinatoire [electronic only], 20:B20a, 41 p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content='–B20a, 41 p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 1988.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Mac15] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Macdonald.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Symmetric functions and Hall polynomials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Oxford Classic Texts in the Physical Sciences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' The Clarendon Press, Oxford University Press, New York, second edition, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' With contribution by A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Zelevinsky and a foreword by Richard Stanley, Reprint of the 2008 paperback edition [ MR1354144].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 51 [Maf02] Andrea Maffei.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A remark on quiver varieties and Weyl groups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' An- nali della Scuola Normale Superiore di Pisa - Classe di Scienze, Ser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 5, 1(3):649–686, 2002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Mel18] Anton Mellit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Integrality of Hausel–Letellier–Villegas kernels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Duke Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 167(17):3171–3205, 11 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Mel19] Anton Mellit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Cell decompositions of character varieties, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Mel20] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Mellit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Poincaré polynomials of character varieties, Macdonald poly- nomials and affine Springer fibers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Annals of Mathematics, 192(1):165 – 228, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Nak94] Hiraku Nakajima.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Instantons on ALE spaces, quiver varieties, and Kac- Moody algebras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Duke Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 76(2):365–416, 11 1994.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Nak98] Hiraku Nakajima.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Quiver varieties and Kac-Moody algebras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Duke Math- ematical Journal, 91(3):515 – 560, 1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Nak00] Hiraku Nakajima.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Reflection functors for quiver varieties and Weyl group actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Mathematische Annalen, 327, 08 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Nak01] Hiraku Nakajima.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Quiver varieties and finite dimensional representa- tions of quantum affine algebras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Journal of the American Mathematical Society, 14(1):145–238, 2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Sai86] Morihiko Saito.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Mixed Hodge modules.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Japan Acad.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Ser.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Sci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=', 62(9):360–363, 1986.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Shm09] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Shmelkin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Some remarks on Nakajima’s quiver varieties of type A, 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Sho88] Toshiaki Shoji.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Geometry of orbits and springer correspondence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' In Orbites unipotentes et représentations - I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Groupes finis et Algèbres de Hecke, number 168 in Astérisque.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Société mathématique de France, 1988.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' [Spr76] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Trigonometric sums, Green functions of finite groups and representations of Weyl groups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' Inventiones Mathematicae, 36:173–207, 1976.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
+page_content=' 52' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/fNE1T4oBgHgl3EQfygV0/content/2301.03434v1.pdf'}
diff --git a/g9E0T4oBgHgl3EQfXgDe/content/2301.02295v1.pdf b/g9E0T4oBgHgl3EQfXgDe/content/2301.02295v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..be042b78c89deecbff5f634eb6ba902b1172c649
--- /dev/null
+++ b/g9E0T4oBgHgl3EQfXgDe/content/2301.02295v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a94f180cfc1d469660128d862bbad61dbe2c040ce55188a47f2d9a3b16151e6
+size 153626
diff --git a/g9E0T4oBgHgl3EQfXgDe/vector_store/index.pkl b/g9E0T4oBgHgl3EQfXgDe/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..0efe9109c6b124738dd11226bcd38fb6bff6cc26
--- /dev/null
+++ b/g9E0T4oBgHgl3EQfXgDe/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e80fd0a0c2f20203811c4b6111d154375d061b86d61b4ba82dd3cb24beffe71
+size 30802
diff --git a/h9FAT4oBgHgl3EQf9h7F/content/2301.08757v1.pdf b/h9FAT4oBgHgl3EQf9h7F/content/2301.08757v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..467006f65551294bf20452f595df4dee0b5e3f18
--- /dev/null
+++ b/h9FAT4oBgHgl3EQf9h7F/content/2301.08757v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5618b97e22822e5bc9d0432abb80a0e712edf23def9c78826685855be603b52a
+size 172846
diff --git a/h9FAT4oBgHgl3EQf9h7F/vector_store/index.faiss b/h9FAT4oBgHgl3EQf9h7F/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..7149b89a7113e160e4a21d80f830a4ae0688ba4f
--- /dev/null
+++ b/h9FAT4oBgHgl3EQf9h7F/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf3e8a48b5ea67c1db8c0d29ad2e4e412419f49a2d7a695b8c75e5174624a804
+size 1638445
diff --git a/h9FAT4oBgHgl3EQf9h7F/vector_store/index.pkl b/h9FAT4oBgHgl3EQf9h7F/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..dfe6cb5f5a91aa8d3460e0b5995073fcdc764445
--- /dev/null
+++ b/h9FAT4oBgHgl3EQf9h7F/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ff06c3e62dced47cff99620cb110102e6f5a9cd4f1bb89c2537d0dcdf55ebe6
+size 62688
diff --git a/hNE0T4oBgHgl3EQfpQGK/content/tmp_files/2301.02537v1.pdf.txt b/hNE0T4oBgHgl3EQfpQGK/content/tmp_files/2301.02537v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3585157105e31bdecbc9207f6db06c91676c96de
--- /dev/null
+++ b/hNE0T4oBgHgl3EQfpQGK/content/tmp_files/2301.02537v1.pdf.txt
@@ -0,0 +1,5876 @@
+SPLITTING SCHEMES FOR SECOND ORDER APPROXIMATIONS OF
+PIECEWISE-DETERMINISTIC MARKOV PROCESSES
+ANDREA BERTAZZI1, PAUL DOBSON2, AND PIERRE MONMARCH´E3
+Abstract. Numerical approximations of piecewise-deterministic Markov processes based on splitting
+schemes are introduced, together with their Metropolis-adjusted versions. The unadjusted schemes
+are shown to have a weak error of order two in the step size in a general framework. Focusing then
+on unadjusted schemes based on the Bouncy Particle and Zig-Zag samplers, we provide conditions
+ensuring ergodicity and consider the expansion of the invariant measure in terms of the step size. The
+dependency of the leading term in this expansion in terms of the refreshment rate, depending of the
+splitting order, is analyzed. Finally, we present numerical experiments on Gaussian targets, a Bayesian
+imaging inverse problem and a system of interacting particles.
+1. Introduction
+Piecewise deterministic Markov processes (PDMP) are non-diffusive Markov processes combining
+a deterministic motion and random jumps.
+They appear in a wide range of modelling problems
+[13, 29, 31] and, over the last decade, have gained considerable interest as Markov Chain Monte Carlo
+(MCMC) methods [39, 34, 7, 10, 19, 43].
+Their dynamics can be described by their infinitesimal
+generator, which is of the form
+Lf(z) = ⟨Φ(z), ∇zf(z)⟩ + λ(z)
+�
+E
+(f(y) − f(z))Q(z, dy) ,
+(1)
+where E is the state space and, in this work, Φ is a smooth and globally Lipschitz vector field,
+λ : E → [0, ∞) is a continuous function and Q is a probability kernel.
+The associated process
+follows the ordinary differential equation (ODE) ˙z = Φ(z) and, at rate λ(z), jumps to a new position
+distributed according to Q(z, ·). We refer to [15, 20] for general considerations on PDMP. Denoting
+by ϕt the integral curve of Φ, i.e. the solution to
+d
+dtϕt(z) = Φ(ϕt(z)),
+ϕ0(z) = z,
+for all t ≥ 0, z ∈ E,
+which exists since Φ is globally Lipschitz, we assume that ϕt leaves E invariant. For T ∼ Exp(1), the
+random time of the next jump, τ, is given by
+τ = inf
+�
+t > 0 :
+� t
+0
+λ(ϕs(z))ds ≥ T
+�
+.
+(2)
+This work addresses the question of the simulation of a PDMP with generator (1). The classical
+method is to use a Poisson thinning procedure [30, 28] to sample the jump times, and then to solve
+the ODE exactly if possible, or otherwise by a standard numerical scheme. Similarly to rejection
+sampling which requires a good reference measure, an efficient Poisson thinning algorithm requires
+the knowledge of good bounds for the jump rate λ along the trajectory of the ODE. In this work, we
+1 Delft Institute of Applied Mathematics, TU Delft
+2 University of Edinburgh
+3 Sorbonne Universit´e
+E-mail addresses: a.bertazzi@tudelft.nl, pdobson@ed.ac.uk, pierre.monmarche@sorbonne-universite.fr.
+Date: January 9, 2023.
+1
+arXiv:2301.02537v1 [math.PR] 6 Jan 2023
+
+2
+SPLITTING SCHEMES FOR PDMPS
+focus on the case in which such bounds are not available, or are so crude that thinning would not be
+numerically efficient. In that case, the random event times have to be approximated even if the ODE
+can be solved exactly. This question has recently been addressed in [3, 38, 14] with three different
+schemes. Here, rather than designing an ad hoc numerical schemes, we work in the general framework
+of splitting schemes, which are widely used for e.g. Hamiltonian or underdamped Langevin processes
+[26, 27, 36]. One of the main interests is that, by design, such schemes have a numerical error which
+is of order 2 in the step-size, without the need of an approximation of the jump rate along the ODE.
+Moreover, it is a flexible framework and thus such schemes can be easily combined with multi-time-
+step or factorization methods [25] or integrated in hybrid PDMP/diffusion schemes [37, 35]. Note
+that, by using a numerical approximation, we lose one of the interest of PDMP for MCMC purpose,
+which is the exact simulation by thinning, while in our case the invariant measure of the scheme
+will have a deterministic bias with respect to the true target measure.
+However, we still benefit
+from the good long-time convergence properties of the ballistic non-reversible process and, contrary
+to Hamiltonian-based dynamics, it is still possible to factorize the target measure and define efficient
+schemes in terms of number of computations of forces (see [37, 35] and Section 5.3). We shall also
+show how the correct stationary distribution can be recovered by means of a non-reversible Metropolis-
+Hastings acceptance/rejection step (see Section 1.2). Moreover, for classical velocity jump processes
+used in MCMC, since the norm of the velocity is constant (between possible refreshments which are
+independent of the potential), these schemes are numerically stable (see the numerical experiments in
+Section 5 where the step-size of PDMP schemes can be taken larger than for the classical ULA), even
+for non-globally Lipschitz potentials.
+The core idea of splitting schemes is first to split the generator in several parts such that a process
+associated to each part can be simulated exactly. For instance, when the ODE can be solved exactly,
+one can write L = LD + LJ with
+LDf(z) = ⟨Φ(z), ∇zf(z)⟩,
+LJf(z) = λ(z)
+�
+E
+(f(y) − f(z))Q(z, dy) ,
+in which case the process associated to LD is simply the solution of the ODE, hence D stands for
+drift, while the process associated to LJ is a continuous-time Markov chain, for which the jump rate is
+constant between two jumps (so that the jump times are simple exponential random variables), hence
+J stands for jumps. Then, one approximates the semigroup of the true process by a Strang splitting
+Pδ = eδ(LD+LJ) ≈ e
+δ
+2 LDeδLJe
+δ
+2 LD
+(3)
+for a small step size δ > 0. Therefore, over one time step the approximation follows LD for time
+δ/2, then LJ for time δ and finally LD again for time δ/2. Given a step size δ, now we illustrate
+how the (n + 1)-th iteration works. Starting at time tn = nδ at state Ztn the process first moves
+deterministically for a half step:
+Ztn+δ/2 = ϕδ/2(Ztn).
+Then we simulate the pure jump part of the process: we generate an event time τ1 ∼ Exp(λ(Ztn+δ/2))
+and, if τ1 < δ, we set Ztn+δ/2 ∼ Q(Ztn+δ/2, ·). Then we repeat this step as long as �
+i τi < δ, though,
+since we are interested in second order schemes, it is enough to limit ourselves to two jumps per time
+step. Note that the rate is updated after every jump and is constant between jumps. We conclude the
+iteration by a final half step of deterministic motion:
+Ztn+1 = ϕδ/2(Ztn+δ/2).
+We refer to this scheme as the splitting scheme DJD, where consistently with above D stands for
+drift and J for jumps. When the ODE cannot be solved exactly, any second-order numerical scheme
+
+SPLITTING SCHEMES FOR PDMPS
+3
+can be used instead of ϕt. Moreover, in some cases (typically for the Hamiltonian dynamics) the
+generator LD can be further divided in several ODEs. Similarly, for computational purpose, it can
+be interesting in some cases to split the jump part LJ in several operators. It is also possible to keep
+in LD a combination of ODE and jump, simulated e.g. by thinning, while some parts of the jump
+are treated separately in LJ (it could make sense for instance in the context of [37]). When there are
+more than two parts in the splitting of L, a scheme is obtained by starting from (3) and using e.g.
+eδLJ ≈ e
+δ
+2 LAeδLBe
+δ
+2 LA if LJ = LA + LB, etc.
+Such splitting schemes can be used to simulate any PDMP. For some modelling problems, it can
+be interesting to have estimates on the trajectorial error between the approximated process and the
+two process, for instance when dynamical properties (like mean squared displacement or transition
+rates) are of interest. However, in this work, we have mainly in mind the PDMPs which are used for
+MCMC methods, in particular our recurrent examples will be the Zig-Zag sampler (ZZS) [7, 5] and
+the Bouncy Particle sampler (BPS) [39, 34, 10]. As a consequence, we will not discuss trajectorial
+errors but rather focus on what is relevant for MCMC purpose, namely the long-time convergence of
+the Markov chain (which should scale properly as the step size vanishes) and the numerical bias on
+the invariant measure and on empirical averages of the chain.
+Organisation of the paper. The article is organized as follows.
+We conclude this introduction by
+presenting the algorithms we focus on in this paper. In Section 1.1 we discuss our two main examples
+and their approximation with splitting schemes. In Sections 1.2 and 1.3 we discuss respectively how we
+can Metropolis-adjust our schemes in a non-reversible fashion and how we can modify the algorithms
+to do subsampling. We conclude our introduction with Section 1.4, where we describe how boundaries
+can be treated with our splitting schemes. Section 2 is devoted to the analysis of the weak error for
+the finite-time empirical averages of the scheme DJD. The main result, Theorem 2.6, states that for
+this scheme the weak error is of order 2 in the step-size. The geometric ergodicity of splitting schemes
+based on our main examples is established in Section 3, with a consistent dependency of the estimates
+on the step-size. In Section 4, we provide a formal expansion (in terms of the step-size) of the invariant
+measure of the scheme based on the so-called Bouncy Particle Sampler depending on the choice of the
+splitting, in the spirit of [26], with a particular focus in Section 4.2 on three one-dimensional examples
+where everything can be made explicit. Numerical experiments are provided in Section 5. Finally,
+technical proofs are gathered in an Appendix.
+Comparison to related works. The work in this paper can be seen as a continuation of the work that
+two of the authors started with their coauthors in [3], in which a general framework to approximate
+PDMPs is introduced and studied.
+In this previous work, the focus is not a specific scheme and
+thus the results are mostly general and not tailored for particular processes or schemes, though the
+ZZS and BPS are considered as recurrent examples.
+In particular, the schemes introduced in [3]
+leave considerable freedom to the user in the choice of some crucial components of the algorithm,
+namely an approximation of the switching rates or a numerical integrator in place of the exact flow
+map. On the other hand, in this paper we follow the philosophy of splitting schemes to describe a
+simple recipe to approximate PDMPs. Note that splitting schemes are not considered in [3]. The
+main advantage of splitting schemes is the second order of accuracy with one gradient evaluation per
+iteration, whereas second order algorithms considered in [3] relied on approximations of second order
+of the switching rates, which can be usually obtained with the expensive computation of the Hessian
+of the negative log-target. Moreover, in this work we describe how to remove the bias introduced by
+our approximation with a non-reversible Metropolis-Hastings step. Two other works ([38] and [14])
+focus on approximate simulation of the Zig-Zag sampler, which is one of our two main examples. In
+[38] the authors suggest to approximate event times by using numerical approximations of the integral
+of the rates along the dynamics (2), as well as a root finding algorithm. In [14], the authors suggest
+
+4
+SPLITTING SCHEMES FOR PDMPS
+Algorithm 1: Splitting scheme DBD for ZZS
+Input
+: Number of iterations N, initial condition (x, v), step size δ.
+Output: Chain (Xtn, V tn)N
+n=0.
+Set n = 0, (X0, V 0) = (x, v);
+while n < N do
+Set Xtn+1 = Xtn + δ
+2V tn;
+Set V tn+1 = V tn;
+for i = 1 . . . , d do
+With probability (1 − exp(δλi(Xtn+1, V tn+1))) set V tn+1 = RiV tn+1;
+end
+Set Xtn+1 = Xtn+1 + δ
+2V tn+1;
+Set n = n + 1;
+end
+using a numerical optimisation algorithm at each iteration to obtain a suitable bound that enables
+the use of Poisson thinning. The first difference is that we mainly consider our approximations as
+discrete time Markov chains, whereas the processes of [38] and [14] are interpreted in continuous time,
+although neither resulting process is a Markov process due to the nature of the numerical algorithms
+that are used.
+Naturally, one could interpret our algorithms as continuous time processes, which
+again would not be Markov processes. Secondly, without assuming any properties that we do not
+verify, we derive theoretical justifications of our proposed algorithms, such as bounds on the weak
+error and existence, uniqueness, and geometric convergence to a stationary distribution under simple
+conditions. Moreover, we introduce Metropolis adjusted algorithms to eliminate the error introduced
+by the numerical approximations, while this aspect is not studied in previous works and thus we
+introduce the first exact PDMP based samplers that can be simulated with only access to the gradient
+of the negative logarithm of the target distribution.
+1.1. Main examples. Let us now introduce two examples from the computational statistics liter-
+ature.
+In this setting we have a target probability measure with density π(x) ∝ exp(−ψ(x)) for
+x ∈ Rd.
+Example 1.1 (Zig-Zag sampler [5]). Let E = Rd × {+1, −1}d. For any z ∈ E, we write z = (x, v)
+for x ∈ Rd, v ∈ {+1, −1}d, where x is interpreted as the position of the particle and v denotes
+the corresponding velocity. The deterministic motion of ZZS is determined by Φ(x, v) = (v, 0)T , i.e.
+the particle travels with constant velocity v. For i = 1, . . . , d we define the jump rates λi(x, v) :=
+(vi∂iψ(x))++γi(x, v), where γi(x) can be any non-negative function and is often chosen to be zero. The
+corresponding (deterministic) jump kernels are given by Qi((x, v), (dy, dw)) = δ(x,Riv)(dy, dw), where
+δz denotes the Dirac delta measure and Ri is the operator that flips the sign of the i-th component of
+the vector it is applied to, that is
+Riv = (v1 . . . , vi−1, −vi, vi+1, . . . , vd).
+Hence the i-th component of the velocity is flipped with rate λi. The ZZS is described by its generator
+Lf(x, v) = ⟨v, ∇xf(x, v)⟩ +
+d
+�
+i=1
+λi(x, v)[f(x, Riv) − f(x, v)].
+(4)
+Simulating the event times with rates of this form is in general a very challenging problem.
+
+SPLITTING SCHEMES FOR PDMPS
+5
+Algorithm 2: Splitting scheme RDBDR for BPS
+Input
+: Number of iterations N, initial condition (x, v), step size δ.
+Output: Chain (Xtn, V tn)N
+n=0.
+Set n = 0, (X0, V 0) = (x, v);
+while n < N do
+Set V tn+1 = V tn ;
+With probability (1 − exp(−λr δ
+2)) draw V tn+1 ∼ Unif(Sd−1) ;
+Set Xtn+1 = Xtn + δ
+2V tn+1;
+With probability (1 − exp(δλ1(Xtn+1, V tn+1))) set V tn+1 = R(Xtn+1)V tn+1;
+Set Xtn+1 = Xtn+1 + δ
+2V tn+1;
+With probability (1 − exp(−λr δ
+2)) set V tn+1 ∼ Unif(Sd−1) ;
+Set n = n + 1;
+end
+We can apply the splitting scheme above as follows. Assume the process has canonical rates, i.e.
+γi = 0 for all i. Then we can split the generator as
+LDf(x, v) = ⟨v, ∇xf(x)⟩,
+LBf(x, v) =
+d
+�
+i=1
+λi(x, v)[f(x, Riv) − f(x, v)].
+Here we define the scheme DBD, where B stands for bounces. Given (Xtn, V tn), we start by a half
+step of deterministic motion:
+Xtn+ δ
+2 = Xtn + δ
+2V tn.
+Then for i = 1, . . . , d we draw τi
+iid
+∼ Exp(λi(Xtn+δ/2, V tn)), which are homogeneous exponential random
+variables. Then let τ(1) = min τi and set
+V tn+1 =
+�
+V tn
+if τ(1) > δ
+RIV tn
+if τ(1) ≤ δ
+where RI = �
+i∈I Ri and I is the set of indices i for which τi ≤ δ. Alternatively to have a second order
+scheme it is sufficient to flip only the two components with the smallest switching time τi, given that
+it is before time δ. Observe that for canonical rates flipping the sign of a component does not affect
+the other switching rates, and thus it is not possible to have two flips in the same component when
+γi = 0. Finally, set
+Xtn+1 = Xtn+ δ
+2 + δ
+2V tn+1,
+which concludes the iteration. The procedure is described in pseudo code form in Algorithm 1. An
+interesting feature of the algorithm is that the jump part of the chain can be computed in parallel,
+since in that stage a velocity flip in one component does not affect the other components of the process.
+Example 1.2 (Bouncy Particle Sampler [10]). Let E = Rd×Rd, and for any z ∈ E we write z = (x, v) for
+x ∈ Rd, v ∈ Rd. The deterministic motion is the same of ZZS: Φ(x, v) = (v, 0)T . The BPS has two types
+of random events: reflections and refreshments. These respectively have rates λ1(x, v) = (vT ∇xψ(x))+
+and λ2(x, v) = λr for λr > 0, and corresponding jump kernels
+Q1((x, v), (dy, dw)) = δ(x,R(x)v)(y, w),
+Q2((x, v), (dy, dw)) = δx(dy)ν(dw),
+
+6
+SPLITTING SCHEMES FOR PDMPS
+where ν is a rotation-invariant probability measure on Rd (typically the standard Gaussian measure
+or the uniform measure on Sd−1), and
+R(x)v = v − 2⟨v, ∇xψ(x)⟩
+|∇xψ(x)|2 ∇xψ(x).
+The operator R reflects the velocity v off the hyperplane that is tangent to the contour line of ψ
+passing though point x. Importantly, the norm of the velocity is unchanged by the application of R,
+and this corresponds to an elastic collision of the particle on the hyperplane. The BPS has generator
+Lf(x, v)=⟨v, ∇xf(x)⟩ + λ1(x, v)[f(x, R(x)v) − f(x, v)] + λ2
+� �
+f(x, w) − f(x, v)
+�
+ν(dw).
+In this case we split the generator in three parts:
+LDf(x, v) = ⟨v, ∇xf(x)⟩,
+LBf(x, v) = λ1(x, v)[f(x, R(x)v) − f(x, v)],
+LRf(x, v) = λ2
+� �
+f(x, w) − f(x, v)
+�
+ν(dw),
+We then define the scheme RDBDR, where R stands for refreshments. Starting at time tn = nδ at
+state (Xtn, V tn) we begin by drawing τ1 ∼ Exp(λr) and setting
+˜Vtn+ δ
+2 =
+�
+V tn
+if τ1 > δ/2
+W1
+if τ1 ≤ δ/2
+for W1 ∼ ν. Then the process evolves deterministically for time δ/2:
+Xtn+ δ
+2 = Xtn + δ
+2
+˜Vtn+ δ
+2 .
+At this point, we check if a reflection takes place by drawing τ2 ∼ Exp(λ1(Xtn+ δ
+2 , ˜Vtn+ δ
+2 )) and set
+V tn+ δ
+2 =
+� ˜Vtn+ δ
+2
+if τ2 > δ
+R(Xtn+ δ
+2 ) ˜Vtn+ δ
+2
+if τ2 ≤ δ
+Importantly, λ1(Xtn+ δ
+2 , V tn+ δ
+2 ) = 0 if a reflection takes place and thus at most one reflection can
+happen. This is a consequence of the fact that ⟨R(x)v, ∇ψ(x)⟩ = −⟨v, ∇ψ(x)⟩ by definition of the
+reflection operator. After this we set
+Xtn+1 = Xtn+ δ
+2 + δ
+2,
+and finally conclude the iteration drawing τ3 ∼ Exp(λr) and letting
+˜Vtn+1 =
+�
+V tn+ δ
+2
+if τ3 > δ/2
+W2
+if τ3 ≤ δ/2
+where W2 ∼ ν. The pseudo code can be found in Algorithm 2.
+1.2. Metropolis adjusted algorithms. Naturally, the use of splitting schemes to approximate a
+PDMP introduces a discretisation error. In the context of Bayesian statistics, this means that a bias
+term is introduced in the estimators for statistics of interest. In this section we discuss how to eliminate
+this bias with the addition of a Metropolis-Hastings (MH) acceptance-rejection step. In Section 1.2.1
+we describe the general procedure, which is a non-reversible MH algorithm, and then apply this to
+ZZS and BPS. Similarly this can be applied to other kinetic PDMPs used in MCMC.
+
+SPLITTING SCHEMES FOR PDMPS
+7
+1.2.1. Non-reversible Metropolis-Hastings. The classical MH algorithm gives a simple procedure to
+construct a µ invariant Markov chain P by satisfying detailed balance (DB): for all x, y it holds that
+µ(dx)P(x, dy) = µ(y)P(y, dx). Integrating DB with respect to x one shows that P is µ-invariant, that
+is
+�
+µ(dx)P(x, dy) = µ(dy). Given a proposal mechanism Q(x, ·) with density q with respect to some
+measure ν, MH constructs P by accepting the state y proposed by Q with probability
+a(x, y) = 1 ∧ µ(y)q(y, x)
+µ(x)q(x, y).
+(5)
+The resulting chain P is reversible as it satisfies DB. PDMPs such as BPS and ZZS break DB and are
+said to be non-reversible. Since there is evidence that this property can lead to a faster converging
+process (see e.g. [18]), it is reasonable here to Metropolise our splitting schemes in a non-reversible
+fashion. Moreover, as we shall see below, for our chains based on splitting schemes of PDMPs it is
+not possible to use the standard MH framework, as typically q(x, y) > 0 implies q(y, x) = 0. The
+reasoning below is an extension of the idea of lifting (for more background on lifting we refer the
+reader to [42, 45, 24, 33]). We shall now consider a chain for which the state can be decomposed as
+z = (x, v), as is the case with the position and velocity parts of the PDMPs we consider. In this
+context, a sufficient condition other than DB ensuring stationarity is skew detailed balance: for all
+x, v, y, w
+µ(dx, dv)P((x, v), (dy, dw)) = µ(dy, dw)P((y, −w), (dx, −dv)).
+(6)
+Indeed, integrating both sides wrt x and v it follows
+�
+µ(dx, dv)P((x, v), (dy, dw)) = µ(dy, dw), that
+is µ is a stationary measure for the chain P. As in the context of the standard MH algorithm, assume
+now we wish to construct our P by accepting or rejecting proposals from a kernel Q.
+Denote as
+a((x, v), (y, w)) the probability of accepting proposal (y, w). For x ̸= y in order for (6) to hold it
+should be that
+a((x, v), (y, w))
+a((y, −v), (x, −v)) = µ(dy, dw)Q((y, −w), (dx, −dv))
+µ(dx, dv)Q((x, v), (dy, dw))
+.
+Hence the acceptance probability can be taken to be
+a((x, v), (y, w)) = 1 ∧ µ(dy, dw)Q((y, −w), (dx, −dv))
+µ(dx, dv)Q((x, v), (dy, dw))
+.
+(7)
+If the proposal (y, w) is rejected, the new state of the chain becomes (x, −v), in which case (6) is
+trivially satisfied.
+Note 1.3. This algorithm can be also obtained with a different reasoning.
+Suppose µ(dx, dv) =
+µ(dx, −dv). At every iteration, the algorithm generates (y, w) ∼ Q((x, v), ·) and successively flips
+the sign of the velocity w. Then the proposal (y, −w) is accepted or rejected with the classical MH
+step, that is (5). Afterwards, the sign of the velocity is flipped again, hence in case of acceptance the
+next state is (y, w), while in case of rejection it is (x, −v). This procedure can be justified noting that
+the MH step ensures that µ is stationary, while changing the sign of velocity preserves µ. We remark
+that, although both the MH step and the velocity reflection are reversible with respect to µ, their
+composition is not.
+1.2.2. Non-reversible Metropolis adjusted ZZS. Consider the splitting DBD of ZZS with initial con-
+dition (x, v). Let δ > 0 be the step size and x1/2(x, v) = x + vδ/2 (we will drop the dependence on
+(x, v) when clear). As explained in Example 1.1, after one iteration the algorithm has state
+( ˜X, ˜V ) = (x1/2 + δ
+2RIv, RIv)
+
+8
+SPLITTING SCHEMES FOR PDMPS
+Algorithm 3: Non-reversible Metropolis adjusted ZZS
+Input
+: Number of iterations N, initial condition (x, v), step size δ.
+Output: Chain (Xtn, V tn)N
+n=0.
+Set n = 0, (X0, V 0) = (x, v);
+while n < N do
+Set Xtn+δ/2 = Xtn + δ
+2V tn;
+Set ˜V = V tn;
+for i = 1 . . . , d do
+With probability (1 − exp(δλi(Xtn+δ/2, ˜V ))) set ˜V = Ri ˜V ;
+end
+Set ˜X = Xtn+δ/2 + δ
+2 ˜V ;
+Set (Xtn+1, V tn+1) = ( ˜X, ˜V ) with probability
+1 ∧ π( ˜X)
+π(Xtn) exp
+�
+δ
+d
+�
+j=1
+�
+λj(Xtn+δ/2, V tn) − λj(Xtn+δ/2, − ˜V )
+� �
+else set (Xtn+1, V tn+1) = (Xtn, −V tn);
+Set n = n + 1;
+end
+with corresponding probability
+exp
+�
+−δ
+�
+i/∈I
+λi(x1/2, v)
+� �
+i∈I
+(1 − exp(−δλi(x1/2, v)).
+(8)
+We now want to accept or reject the proposed state with suitable probability to ensure µ-stationarity.
+Note the classical MH scheme (5) is not directly applicable, as typically there is a 0 probability that
+the process goes from ( ˜X, ˜V ) to (x, v). Hence we use the non-reversible MH acceptance probability
+(7). For this we need to compute the probability of going from ( ˜X, − ˜V ) to (x, −v) according the
+transition kernel of DBD. This can only be achieved by following the same path of (x, v) → ( ˜X, ˜V )
+with reversed time. Hence the sign of the velocity of the components in α needs to be flipped. Noticing
+that x1/2(x, v) = x1/2( ˜X, − ˜V ), we find that the probability of this path is
+exp
+�
+−δ
+�
+i/∈I
+λi(x1/2, − ˜V )
+� �
+i∈I
+(1 − exp(−δλi(x1/2, − ˜V )),
+where I is the same set of indices of (8). Observe that for i ∈ I it holds that ˜Vi = −vi and thus
+λi(x1/2, v) = λi(x1/2, − ˜V ), while for i /∈ I we have ˜Vi = vi and hence λi(x1/2, v) − λi(x1/2, − ˜V ) =
+vi∂ψ(x1/2). Therefore the acceptance probability (7) simplifies to
+1 ∧ π( ˜X) × exp(−δ �
+i/∈I λi(x1/2, − ˜V ))
+π(x) × exp(−δ �
+i/∈I λi(x1/2, v))
+= 1 ∧ exp
+�
+ψ(x) − ψ( ˜X) + δ
+�
+i/∈I
+vi∂iψ(x1/2)
+�
+.
+(9)
+In case of rejection, the state is set to (x, −v). The procedure is written as pseudo-code in Algorithm
+3.
+
+SPLITTING SCHEMES FOR PDMPS
+9
+Algorithm 4: Non-reversible Metropolis adjusted BPS
+Input
+: Number of iterations N, initial condition (x, v), step size δ.
+Output: Chain (Xtn, V tn)N
+n=0.
+Set n = 0, (X0, V 0) = (x, v);
+while n < N do
+Set V tn+δ/2 = V tn ;
+With probability (1 − exp(−λrδ/2)) draw V tn+δ/2 ∼ Unif(Sd−1) ;
+Set Xtn+δ/2 = Xtn + δ
+2V tn+δ/2;
+Set ˜V = V tn+δ/2;
+With probability (1 − exp(δλ1(Xtn+δ/2, ˜V ))) set ˜V = R(Xtn+δ/2) ˜V ;
+Set ˜X = ˜X + δ
+2 ˜V ;
+Set (Xtn+1, V tn+1) = ( ˜X, ˜V ) with probability
+1 ∧ π( ˜X) × exp(−δλ(Xtn+δ/2, − ˜V ))
+π(Xtn) × exp(−δλ(Xtn+δ/2, V tn))
+else set (Xtn+1, V tn+1) = (Xtn, −V tn+δ/2) ;
+With probability (1 − exp(−λrδ/2)) set V tn+1 ∼ Unif(Sd−1) ;
+Set n = n + 1;
+end
+Note 1.4. Assuming ψ ∈ C2, the terms ψ(x) and ψ( ˜X) = ψ(x1/2 + RIv δ/2) can be expanded by
+Taylor’s theorem around x1/2. This gives that the acceptance probability (9) is
+1 ∧ exp
+�δ2
+8
+�
+⟨v, ∇2ψ(x1)v⟩ − ⟨RIv, ∇2ψ(x2)RIv⟩
+��
+,
+(10)
+where x1 ∈ (x, x1/2) and x2 ∈ (x1/2, x1/2 +RIvδ/2). This result shows that the probability of rejecting
+the proposed state is of order δ2 and gives first evidence that the splitting scheme introduces an error of
+second order in the invariant measure. Moreover, it is clear from (10) that the acceptance probability
+equals 1 for instance when ∇2ψ is a constant diagonal matrix, as is the case in a d-dimensional
+independent Gaussian vector. In this setting the splitting scheme DBD has the correct stationary
+distribution µ and does not need a Metropolis correction. Finally, observe that the rejection probability
+is of order δ3 if ∇2ψ is diagonal but not constant.
+1.2.3. Non-reversible Metropolis adjusted BPS. Here we consider scheme RDBDR of BPS. The re-
+freshment does not alter the stationary distribution of the process, thus we focus first on the DBD
+part. Denote x1/2(x, v) = x + δv/2. According to DBD, the process moves from an initial condition
+(x, v) to
+( ˜X, ˜V ) =
+�
+(x1/2 + δ
+2R(x1/2)v, R(x1/2)v)
+with probability 1 − exp(−δλ(x1/2, v)),
+(x + δv, v)
+with probability exp(−δλ(x1/2, v)).
+(11)
+Observe that for both states in (11) it holds x1/2(x, v) = x1/2( ˜X, − ˜V ) = x1/2. We now focus on
+computing the acceptance probability (7) in the two cases in (11).
+Consider first the case in which a reflection took place, which corresponds to the first line of (11).
+Then we need to compute the probability that the process goes from ( ˜X, − ˜V ) back to (x, −v) using
+scheme DBD, which is equal to the probability that the process has a reflection at x1/2. By definition
+
+10
+SPLITTING SCHEMES FOR PDMPS
+of the reflection rate λ, it holds that λ(x1/2, v) = λ(x1/2, −R(x1/2)v).
+Therefore in this case the
+probability that the process goes from ( ˜X, − ˜V ) to (x, −v) is the same as that of going from (x, v) to
+( ˜X, ˜V ) and thus the acceptance probability (7) is
+1 ∧ π(x1/2 + δ
+2R(x1/2)v)
+π(x)
+= 1 ∧ exp
+�
+ψ(x) − ψ
+�
+x1/2 + δR(x1/2)v/2
+� �
+.
+(12)
+Observe that moves that decrease ψ are accepted with probability 1.
+Consider now the second case in (11). The probability that the process goes from (x + δv, −v) to
+(x, −v) is exp(−δλ(x1/2, −v)), while the probability of going from (x, v) to (x+δv, v) is exp(−δλ(x1/2, v)).
+Observing that λ(x1/2, v) − λ(x1/2, −v) = ⟨v, ∇ψ(x1/2)⟩ we find that in this case the MH acceptance
+probability is
+1 ∧ π(x + δv) × exp(−δλ(x1/2, −v))
+π(x) × exp(−δλ(x1/2, v))
+= 1 ∧ exp
+�
+ψ(x) − ψ(x + vδ) + δ⟨v, ∇ψ(x1/2)⟩
+�
+.
+(13)
+Hence we have shown that the acceptance probability can in general be written as
+1 ∧ π( ˜X) × exp(−δλ(x1/2, − ˜V ))
+π(x) × exp(−δλ(x1/2, v))
+= 1 ∧ exp
+�
+ψ(x) − ψ( ˜X) + δ(λ(x1/2, v) − λ(x1/2, − ˜V ))
+�
+.
+In case of rejection, the state is set to (x, −v). Two refreshments half-steps, to be executed before and
+after the scheme DBD, are necessary to ensure irreducibility of the Markov chain.
+The described procedure is written in pseudo code form in Algorithm 4.
+Note 1.5. Let us Taylor expand the acceptance probabilities similarly to Note 1.4. Indeed for (12) we
+expand both terms around x1/2 and for x1 ∈ (x, x1/2) and x2 ∈ (x1/2, x + vδ) we obtain
+ψ(x) − ψ
+�
+x1/2 + δ
+2R(x1/2)v
+�
+= −δ
+2
+�
+⟨v, ∇ψ(x1/2)⟩ + ⟨R(x1/2)v, ∇ψ(x1/2)⟩
+�
++ δ2
+8
+�
+⟨v, ∇2ψ(x1)v⟩ − ⟨R(x1/2)v, ∇2ψ(x2)R(x1/2)v⟩
+�
+= δ2
+8
+�
+⟨v, ∇2ψ(x1)v⟩ − ⟨R(x1/2)v, ∇2ψ(x2)R(x1/2)v⟩
+�
+.
+In the last line we used that ⟨R(x1/2)v, ∇ψ(x1/2)⟩ = −⟨v, ∇ψ(x1/2)⟩. Similarly, in (13) we expand
+terms ψ(x) and ψ(x + vδ) around x1/2 to find
+ψ(x) − ψ(x + vδ) + δ⟨v, ∇ψ(x1/2)⟩ =δ2
+8
+�
+⟨v, ∇2ψ(x1)v⟩ − ⟨v, ∇2ψ(x2)v⟩
+�
+for x1 ∈ (x, x1/2) and x2 ∈ (x1/2, x + vδ). If the Hessian of ψ is constant, as for instance in the
+Gaussian case, then proposals of this type are accepted with probability 1.
+Overall, this means the acceptance probability has the form
+1 ∧ exp
+�δ2
+8
+�
+⟨v, ∇2ψ(x1)v⟩ − ⟨ ˜V , ∇2ψ(x2) ˜V ⟩
+��
+,
+which means that the probability of rejecting the proposed state is of order δ2. In particular if π is a
+d-dimensional Gaussian with covariance Σ = cId, then the probability of accepting the proposed state
+in the MH step is equal to 1, as ∥ ˜V ∥ = ∥v∥ = 1. Hence in this case the splitting scheme RDBDR
+has the correct stationary distribution.
+
+SPLITTING SCHEMES FOR PDMPS
+11
+Algorithm 5: Splitting scheme DBD for ZZS with subsampling
+Input
+: Number of iterations N, initial condition (x, v), step size δ.
+Output: Chain (Xtn, V tn)N
+n=0.
+Set n = 0, (X0, V 0) = (x, v);
+while n < N do
+Set Xtn+1 = Xtn + δ
+2V tn;
+Draw J ∼ Unif({1, . . . , N});
+for i = 1 . . . , d do
+Obtain (V tn+1)i by simulating a pure jump process with kernel Ri and rate
+v �→ (v∂iψj(Xtn+1))+ with initial velocity (V tn)i and time horizon δ ;
+end
+Set Xtn+1 = Xtn+1 + δ
+2V tn+1;
+Set n = n + 1;
+end
+1.3. Algorithms with subsampling. One of the attractive features of ZZS and BPS is exact sub-
+sampling, i.e. the possibility when the potential is of the form ψ(x) =
+1
+N
+�N
+j=1 ψj(x) of using only
+a randomly chosen ψj to simulate the next event time. The clearest application of this technique is
+Bayesian statistics, where ψ(x) is the posterior distribution, x is the parameter of the chosen statistical
+model and, when the data points are independent realisations, ψj can be chosen to depend only on
+the j-th batch of data points and not on the rest of the dataset. Therefore, this technique can greatly
+reduce the computational cost per event time.
+Naturally, Bayesian statistics is not the only area
+where this structure of ψ arises. An example from molecular dynamics whit this type of potential is
+considered in Section 5.3. Here we define a splitting scheme of ZZS with this feature. With the same
+ideas it is possible to define a splitting scheme with subsampling based on BPS, but we do not give
+the details here for the sake of brevity.
+Let us briefly explain the basic idea in the case of ZZS, as given in [5]. Assume the target distri-
+bution is of the form ψ(x) =
+1
+N
+�N
+j=1 ψj(x) and define the switching rates λj
+i(x, v) = (vi∂iψj(x))+
+for i = 1, . . . , d and j = 1, . . . , N. Assuming we have a tractable M such that λj
+i(x + vt, v) ≤ M(t)
+for all j = 1, . . . , N, one can use Poisson thinning to obtain a proposal τ for the next event time
+distributed as Exp(M(t)). This proposal is then accepted with probability λJ
+i (x + vτ, v)/M(τ), where
+J ∼ Unif({1, . . . , N}) independently of the rest. This procedure defines a ZZS with switching rates
+λi(x, v) = 1
+N
+�N
+j=1 λj
+i(x, v), which are larger than the canonical rates, but keep π stationary.
+Clearly the bottleneck of this procedure is that a sharp bound M needs to be available. Algorithm 5
+defines an approximation of this process with a similar idea as [3]. At each iteration the algorithm
+draws J ∼ Unif({1, . . . , N}) independently of the rest and uses the corresponding ψJ to update the
+process. Since the rates are now larger than the canonical rates, that is λi(x, v) > (vi∂iψ(x))+, there
+can be more than one jump per component at each iteration. Nonetheless, the algorithm requires
+only one gradient computation per iteration since the position is not updated during the jump part.
+Moreover, in this case obtaining the gradient ∇ψJ is an order 1 computation as opposed to the usual
+order N needed to compute the full gradient ∇ψ.
+1.4. PDMPs with boundaries. Another interesting feature of PDMPs such as BPS and ZZS is that,
+thanks to the simple deterministic dynamics, boundary conditions can be included and hitting times
+of the boundary can be easily computed (see [16] or [12] for a discussion of PDMPs with boundaries).
+Here we illustrate how to simply adapt splitting schemes to these settings by adding the boundary
+behaviour to the D part of the scheme.
+
+12
+SPLITTING SCHEMES FOR PDMPS
+Boundary terms appear for instance when the target distribution π is defined on a restricted domain,
+in which case a boundary jump kernel can be introduced as considered in [4]. In this case, Algorithms
+1 and 2 can be easily modified by incorporating the boundary term in part D of the splitting scheme,
+as typically the boundary can be hit only if there is deterministic motion. Hence, the continuous
+deterministic dynamics are applied as in the exact process, while other jumps are performed in the B
+steps.
+Another example of this setting is when π is a mixture of a continuous density and a discrete
+distribution on finitely many states, as in Bayesian variable selection when a spike and slab prior is
+chosen. Sticky PDMPs were introduced in [6] to target a distribution of the form
+µ(dx) ∝ exp(−ψ(x))
+d
+�
+i=1
+(dxi + 1
+ci
+δ0(dxi)),
+which assigns strictly positive mass to events {xi = 0}. The sticky ZZS of [6] is obtained following the
+usual dynamics of the standard ZZS and in addition freezing the i-th component for a time τ ∼ Exp(ci)
+when xi hits zero. The simulation of this process is challenging for the same reasons of the standard
+ZZS, since the two processes have the same switching rates λi for i = 1, . . . , d. The i-th component is
+either frozen, which is denoted by (xi, vi) ∈ Ai, or it evolves as given by the usual dynamics of ZZS.
+The generator can then be decomposed as L = LD + LB where LD = �d
+i=1 LD,i and LB = �d
+i=1 LB,i,
+LD,if(x, v) = vi
+∂
+∂xi
+f(x, v)1AC
+i (xi, vi) + ci(f(Ti(x, v)) − f(x, v))1Ai(xi, vi),
+LB,if(x, v) = λi(x, v)[f(x, Riv) − f(x, v)]1AC
+i (xi, vi),
+and Ti(x, v) corresponds to unfreezing the i-th component (we refer to [6] for a detailed description).
+An iteration of the scheme DBD in this case proceeds by a first half step of D, which is identical to the
+continuous sticky ZZS but with λi temporarily set to 0. Hence frozen components are unfrozen with
+rate ci and then start moving again, or unfrozen components move with their corresponding velocity
+vi and become frozen for a random time with rate ci if they hit xi = 0. Then a full step of the usual
+bounce kernel B is done for the components which are not frozen, while for the frozen components, that
+is (xi, vi) ∈ Ai, the generator LB,i does nothing and so the velocity cannot be flipped. So unfreezing
+is not possible in this step. The iteration ends with another half step of D in a similar fashion to the
+previous one.
+These ideas are more general than the two specific examples we considered and do not introduce
+further difficulties for our schemes. We observe that in these cases it might be useful to consider
+the process obtained with the splitting schemes as continuous time processes. Finally, notice that
+a Metropolis correction can be added following Section 1.2, and subsampling is possible following
+Section 1.3.
+2. Convergence of the splitting scheme
+In this section we prove that under suitable conditions the splitting scheme DJD described in
+Section 1 is indeed a second order approximation of the original PDMP (1).
+Note that in this section we have a PDMP defined on some arbitrary space E therefore it is not clear
+what it means to have a derivative, indeed we will typically be interested in the setting E = Rd × V
+for some set V which may be a discrete set. Instead of working with a full derivative we will define
+the directional derivative, DΦ, in the direction Φ as
+DΦg(z) = lim
+t→0
+d
+dtg(ϕt(z))
+
+SPLITTING SCHEMES FOR PDMPS
+13
+for any g ∈ C(E) for which t �→ g(ϕt(z)) is continuously differentiable in t for every z. Note if E is a
+subset of Rd for some d and g is continuously differentiable then
+DΦg(z) = Φ(z)T ∇g(z).
+We extend this definition to multi-dimensional valued functions G : E → Rm by defining DΦG(z) =
+(DΦGi(z))m
+i=1. We define the space Ck,m
+Φ
+to be the set of all functions g : E → R which are k times
+continuously differentiable in the direction Φ with all derivatives Dℓ
+Φg(z) up to order k bounded by a
+polynomial of order m. We endow this space with the norm
+∥g∥Ck,m
+Φ
+:= sup
+z∈E
+|g(z)| + �k
+ℓ=1|Dℓ
+Φg(z)|
+1 + |z|m
+.
+Let us make the following assumptions.
+Assumption 2.1. Let Φ be a globally Lipschitz vector field defined on E.
+We assume that the
+directional derivative in the direction Φ is well-defined and that Φ be continuous.
+Assumption 2.2. The switching rate λ : E → [0, ∞) is twice continuously differentiable in the
+direction Φ and λ, DΦλ, D2
+Φλ grow at most polynomially. We denote by mλ a constant such that
+∥λ∥C
+2,mλ
+Φ
+< ∞.
+Assumption 2.3. Let Q be a probability kernel defined on E.
+We shall consider the operator
+Q : Cb(E) → Cb(E) defined by
+Qg(z) =
+�
+g(˜z)Q(z, d˜z),
+for any g ∈ Cb(E).
+(14)
+Moreover we assume that Q has moments of all orders and Qg has at most polynomial growth of order
+m whenever g has at most polynomial growth of order m. For any m ∈ N, and g ∈ C1,m
+Φ
+we assume
+the following distribution is well-defined:
+(DΦQ)g(z) = DΦ(Qg)(z).
+(15)
+As an abuse of notation we shall write DΦQ also as a kernel. We assume for any m ∈ N, and g ∈ C1,m
+Φ
+|Qg(ϕs(z)) − Qg(z)| ≤ Cs(1 + |z|m)∥g∥C1,m
+Φ ,
+(16)
+and also that there exists a constant C such that for any g ∈ C2,m
+Φ
+|Qg(ϕs(z)) − Qg(z) − sDΦQg(z)| ≤ Cs2(1 + |z|m)∥g∥C2,m
+Φ .
+(17)
+Assumption 2.4. The closure (L, D(L)) of the operator (L, C1
+c (E)) in L2
+µ generates a C0-semigroup
+Pt. If g ∈ C2,0
+Φ
+then we assume that Ptg is also twice continuously differentiable in the direction Φ and
+LPtg is continuously differentiable in the direction Φ. Moreover we assume D2
+ΦPtg and DΦLPtg are
+both polynomially bounded for finite t and for some C > 0, R ∈ R, mP ∈ N
+|DΦPtg(z)| + |D2
+ΦPtg(z)| + |DΦLPtg(z)| ≤ C(1 + |z|mP)eRt∥g∥C2,0
+Φ .
+Assumption 2.5. Let Ztk denote the approximation obtained by the splitting scheme DJD. Assume
+that for each k, Ztk has moments of all orders and moreover for every M ∈ N there exists some GM
+such that
+sup
+m≤M
+Ez[|Ztk|m] ≤ GM(z).
+
+14
+SPLITTING SCHEMES FOR PDMPS
+Theorem 2.6. Let Zt be a PDMP corresponding to the generator (1). Assume that Assumption 2.1
+to Assumption 2.5 hold. Then there exist constants C, R such that for any g ∈ C2,0
+Φ ∩ D(L) we have
+for some M ∈ N
+sup
+k≤n
+|E[g(Ztk)] − E[g(Ztk)]| ≤ CeRtnGM(z)δ3n∥g∥C2,0
+Φ .
+Proof. The proof is adapted from [3, Theorem 4.24] and can be found in Appendix A.1.
+□
+Example 2.7 (ZZS continued). Recall the Zig-Zag sampler from Example 1.1 let us verify the Assump-
+tion 2.1 to 2.5 in this case. In order to have a smooth switching rate we replace λi(x, v) by
+λi(x, v) = log (1 + exp(vi∂iψ(x))) .
+This is shown to be a valid switching rate in [1]. We will assume that ψ ∈ C2 with bounded second
+and third derivatives. Let us now consider each assumption in turn.
+Assumption 2.1: In this case Φ(x, v) = (v, 0)T which is clearly smooth and globally Lipschitz.
+Assumption 2.2: Since λi is the composition of smooth maps and ψ we have that λi has the same
+smoothness in x as ψ and hence x �→ λi(x, v) is C2. As s �→ log(1+es) grows at most linearly, has first
+and second derivatives bounded by 1 we have that λi, ∇xλi and ∇2
+xλi are all polynomially bounded.
+Assumption 2.3: The proof of this can be found in Section A.2.
+Assumption 2.4: By [1] we have that Pt is a strongly continuous semigroup on L2
+µ with generator
+(L, D(L)) given as the closure of (L, C1
+c (E)). Moreover we have that the assumptions of [20, Theorem
+17] are satisfied and hence Ptg(x, v) is differentiable in x. Following the proof of [20, Theorem 17] one
+also has
+|∇xPtg| ≤ C(1 + |x|m)eRt∥g∥C1,0
+Φ .
+Note here since DΦg(x, v) = vT ∇xg we have that Ck,0
+Φ
+coincides with the space of continuous functions
+which are k-times continuously differentiable in the variable x. By the same arguments one can also
+obtain
+|∇2
+xPtg| ≤ C(1 + |x|m)eRt∥g∥C2,0
+Φ .
+Assumption 2.5: This will be established in Theorem 3.6.
+3. Ergodicity of splitting schemes of BPS and ZZS
+We shall now focus on results on ergodicity of splitting schemes of BPS and ZZS. In particular
+we show existence of an invariant distribution, characterise the set of all invariant distributions, and
+establish convergence of the law of the process to such distributions with geometric rate. In order to
+prove this we rely on the following classical result, due to Meyn and Tweedie [32] (here the specific
+statement is based on [23, Theorem 1.2], see also [19, Theorem S.7] for the explicit constants). Recall
+the definition of V -norm: ∥µ∥V := sup|g|≤V |µ(g)|.
+Theorem 3.1. Consider a Markov chain with transition kernel P on a set E. Suppose that there
+exist constants ρ ∈ [0, 1), C, α > 0, a function V : E → [1, +∞) and a probability measure ν on E
+such that the two following conditions are verified:
+(1) Drift condition: for all x ∈ E,
+PV (x) ≤ ρV (x) + C.
+(18)
+(2) Local Dobelin condition: for all x ∈ E with V (x) ⩽ 4C/(1 − ρ),
+δxP ⩾ αν .
+
+SPLITTING SCHEMES FOR PDMPS
+15
+Then, for all probability measures µ, µ′ on E and all n ∈ N,
+∥µP n − µ′P n∥V ⩽ C
+α κn∥µ − µ′∥V
+(19)
+where κ = max(1 − α/2, (3 + ρ)/4). Moreover P admits a unique stationary distribution µ∗ satisfying
+µ∗(V ) < ∞.
+Note 3.2. Under the Drift condition (18) alone, following the proof of [23, Theorem 1.2] in the case
+α = 0 we get that for all probability measures µ, µ′ on E,
+∥µP − µ′P∥V ⩽ (ρ + 2C)∥µ − µ′∥V .
+We shall now consider our splitting schemes and prove geometric ergodicity under suitable conditions
+by showing that the assumptions of Theorem 3.1 are satisfied. The splitting schemes of BPS and ZZS
+are respectively addressed in Theorems 3.4 and 3.6 below and, in both cases, the dependence of all
+constants in (19) on the step size is made explicit (statements with more details are postponed to
+Appendixes B and C). More precisely, in both cases, we obtain a local Doeblin (or minorisation)
+condition with constant α after n∗ = ⌈t∗/δ⌉ steps, where t∗ > 0 plays the role of physical time and
+n∗ is the number of steps needed to travel for an equivalent time. Here t∗, α are independent of δ.
+On the other hand, we show that the drift condition holds for one step of the kernel with constants
+ρ = 1 − bδ and C = Dδ, where b, D and the Lyapunov function V are independent of δ. This implies
+that for any s > 0 and any δ ∈ (0, δ0]
+(Pδ)⌈s/δ⌉V
+⩽
+(1 − bδ)⌈s/δ⌉ V + Dδ
+⌈s/δ⌉−1
+�
+k=0
+(1 − bδ)k
+⩽
+e−bsV + D
+b .
+Applying Theorem 3.1, we get for P n∗
+δ
+a long-time convergence estimate which is uniform over δ ∈
+(0, δ0], that is for all δ ∈ (0, δ0] and n ≥ 1 we find
+∥µ(P n∗
+δ )n − µ′(P n∗
+δ )n∥V ⩽ C′
+α κn ∥µ − µ′∥V ,
+where C′ = D/b and κ = max(1 − α/2, (3 + e−bt∗)/4). Observe that the rhs does not depend on δ.
+Using the observation in Note 3.2, we can get convergence in V -norm for P n. Indeed for n = mn∗ + r
+with r < n∗ we have
+∥µP n
+δ − µ′P n
+δ ∥V
+=
+∥µP mn∗+r
+δ
+− µ′P mn∗+r
+δ
+∥V
+⩽
+C′
+α κm∥µP r
+δ − µ′P r
+δ ∥V
+⩽
+C′
+α
+�
+1 + 2C′�
+κm∥µ − µ′∥V
+⩽
+C′′˜κnδ∥µ − µ′∥V ,
+(20)
+where ˜κ = κ1/(t∗+δ0) ∈ (0, 1) and C′′ = C′ (1 + 2C′) /(ακ) are independent from δ. Here we used that
+with computations identical to above we get the drift condition P r
+δ V ≤ (1−bδ)V +D(1−(1−bδ)r)/b ≤
+V +C′, which is enough for the current purpose. As a conclusion, the estimates given in Theorems 3.4
+and 3.6 below (or in Appendixes B and C for more details) give the expected dependency in δ for the
+convergence rate of the process toward equilibrium.
+For splitting schemes of the BPS, we work under the following condition.
+
+16
+SPLITTING SCHEMES FOR PDMPS
+Assumption 3.3. The dimension is d ≥ 2, the velocity equilibrium ν is the uniform measure on Sd−1.
+There exists C > 0 such that
+1
+C |x|2 − C ⩽ ψ(x) ⩽ C|x|2 + C ,
+1
+C |x| − C ⩽ |∇ψ(x)| ⩽ C|x| + C
+for all x ∈ Rd. Moreover, ∥∇2ψ∥∞ < ∞ and, without loss of generality, inf ψ = 1.
+Notice that, when d = 1, the BPS and the ZZS coincide, in which case we refer to Theorem 3.6
+below. Our result of ergodicity for splitting schemes of the BPS is the following.
+Theorem 3.4. Consider any scheme of the BPS based on the decomposition D,R,B. Under Assump-
+tion 3.3, there exist δ0, a, C′′ > 0, ˜κ ∈ (0, 1) and V : Rd × Sd−1 → [1, +∞) satisfying
+for all x ∈ Rd, v ∈ Sd−1,
+e|x|/a/a ⩽ V (x, v) ⩽ aea|x|
+such that, for all δ ∈ (0, δ0], Theorem 3.1 is applicable and (20) holds with these C′′, ˜κ, V .
+Proof. The proof can be found in Appendix B.
+□
+More care is required for the DBD scheme of the ZZS since this Markov chain has periodicity and
+is not irreducible, which is reminiscent of the discrete-space Zig-Zag chain studied in [35]. Let us
+illustrate this behaviour by considering the one dimensional setting. Let (x, v) be the initial condition
+of the process. Since v has magnitude 1, the position component x can only vary by multiples of the
+step size δ. Thus for a fixed initial condition (x, v) the process remains on a grid (x + δZ) × {−1, 1}.
+Moreover, after a single step of the scheme there are two possible outcomes: either the velocity does
+not change, in which case x moves to x + δv, or the velocity is flipped and the position remains the
+same. This means that the change in the position (by amounts of δ) plus half the difference in the
+velocity always changes by ±1 each step and hence is equal to the number of steps in the scheme up
+to multiples of two, i.e.
+Xnδ − x
+δ
++ 1
+2(V nδ − v) ∈ n + 2Z.
+As a consequence, the chain lives on two disjoint sets depending on whether n is even or odd, which
+means that it is periodic. To overcome this issue, we consider the chain with one step transition
+kernel given by P 2
+δ = PδPδ, i.e. we restrict to the case of an even number of steps. The Markov chain
+with kernel P 2
+δ is aperiodic, but it is not irreducible on Rd and hence has (infinitely) many invariant
+measures. In order to characterise the invariant measures we restrict to the set in which the Markov
+kernel P 2
+δ is irreducible. For fixed (x, v) ∈ Rd × {−1, 1}d we construct the grid which contains (x, v)
+as follows:
+D(x, v) := {(y, w) ∈ C × {±1}d : (yi, wi) ∈ D1(xi, vi) for all i = 1, . . . , d},
+(21)
+where D1(xi, vi) := D+(xi, vi) ∪ D−(xi, vi), with
+D+(xi, vi) := {(yi, wi) : wi = vi, yi = xi + mδ, m ∈ 2Z},
+D−(xi, vi) := {(yi, wi) : wi = −vi, y = xi + mδ, m ∈ 2Z + 1}.
+In this case we show in Theorem 3.6 that the Markov chain with transition kernel P 2
+δ is irreducible
+on D(x, v), has a unique invariant measure, πx,v
+δ , and is geometrically ergodic. Now we can characterise
+all the invariant measures of the Markov chain with transition kernel P 2
+δ defined on Rd × {−1, 1}d as
+the closed convex hull of the set {πx,v
+δ
+: x ∈ Rd, v ∈ {−1, 1}d}. Now consider the Markov chain with
+transition kernel P 2
+δ on Rd × {−1, 1}d. For any initial distribution µ we have convergence of µP 2n
+δ
+to
+some measure πµ
+δ as n tends to ∞ and πµ
+δ is given by
+πµ
+δ (ϕ) = (µπx,v
+δ )(ϕ) :=
+�
+Rd×{−1,1}d
+�
+Rd×{−1,1}d ϕ(y, w)πx,v
+δ (dy, dw)µ(dx, dv).
+(22)
+
+SPLITTING SCHEMES FOR PDMPS
+17
+We use the next assumption to verify that Theorem 3.1 applies for initial conditions drawn from
+probability distributions with support on D(x, v).
+Assumption 3.5. Consider switching rates λi(x, v) = (vi∂iψ(x))++γi(x) for i = 1, . . . , d. ψ ∈ C2(Rd)
+and the following conditions hold:
+(a) The switching rates λi(x, v) are such that there exist x0 ≥ 0 such that for all x1 > x0
+λ(x1) :=
+min
+i=1,...,d
+min
+(x,v): xivi∈[x0,x1], |xj|∈[x0,x1] for all j̸=i λi(x, v) > 0.
+(b) For |x| ≥ R for some R > 0
+sup
+t∈(0,1),y1,y2∈B(x,t
+√
+d),v,w∈{−1,1}d
+e(t2(|(v+w)T ∇2ψ(y1))i|+2t|(w∇2ψ(y2))i|)γi(x + vt)etvi∂iψ(x) ≤ γ0 < 1.
+(23)
+(c) Denote as B(x, δ
+√
+d) the ball with centre at x and radius δ
+√
+d. Then
+lim
+∥x∥→∞
+sup
+y1,y2∈B(x,δ
+√
+d)
+max{1, ∥∇2ψ(y1)∥}
+|∂iψ(y2)|
+= 0
+for all 0 ≤ δ ≤ δ0, i = 1, . . . , d,
+where δ0 = 2(1 + γ0)−1, for γ0 as in part (b).
+Part (a) in Assumption 3.5 is inspired by [7, Assumption 3] and is used to show that a minorisation
+condition holds. This condition is either a consequence of properties of the target, or else can be
+enforced by taking a non-negative excess switching rate, in which case γi(x) can be chosen to be a
+continuous function γi : Rd → (0, ∞). In principle one could prove a minorisation condition using the
+techniques of [8], but this is beyond the scope of this paper. Part (b) is a condition on the decay of
+the refreshment rate, while Part (c) is similar to Growth Condition 3 in [8] and is satisfied for instance
+if ψ is strongly convex with globally Lipschitz gradient. These two conditions are used to show that
+a drift condition holds.
+Theorem 3.6. Consider the splitting scheme DBD for ZZS. Suppose Assumption 3.5 holds. Then
+there exist C′′, δ0 > 0, ˜κ ∈ (0, 1) and V : Rd × {−1, 1}d → [1, ∞) satisfying
+for all (x, v) ∈ Rd × {−1, 1}d ,
+d
+�
+i=1
+(1 + 2|∂iψ(x)|)− 1
+2 ≤
+V (x, v)
+exp(βψ(x)) ≤
+d
+�
+i=1
+(1 + 2|∂iψ(x)|)
+1
+2
+for all β ∈ (0, 1/2) such that, for all δ ∈ (0, δ0], the following holds:
+(1) Fix (x, v) ∈ Rd × {−1, 1}d. Theorem 3.1 is applicable to P 2
+δ = PδPδ seen as a transition kernel
+on D(x, v), and the inequality (20) holds (with Pδ replaced by P 2
+δ ) with these C′′, ˜κ, V for any
+µ, µ′ having support on D(x, v).
+(2) For any probability measure µ on Rd × {−1, 1}d with µ(V ) < ∞, we have that µP 2n
+δ
+converges
+as n → ∞ to the measure πµ
+δ given by (22) where πx,v
+δ
+is the unique invariant measure of P 2
+δ
+on D(x, v) and we have
+∥µP 2n
+δ
+− µπx,v
+δ ∥V ≤ C′′˜κnδ
+�
+∥δ(x,v) − πx,v
+δ ∥V µ(dx, dv) .
+(24)
+Proof. The proof can be found in Appendix C.1.
+□
+Under similar assumptions we establish geometric ergodicity of schemes DRBRD, RDBDR of
+ZZS, where the switching rates in the B part are λi(x, v) = (vi∂iψ(x))+, i.e. the canonical rates, while
+refreshments in the R part are independent draws from Unif({±1}d) with rate γ(x) : Rd → [0, ∞).
+The rigorous statement of this result, Theorem C.6, and its proof can be found in Appendix C.2.
+
+18
+SPLITTING SCHEMES FOR PDMPS
+Figure 1. Empirical error for the radius statistic t(x) = x2 with a one-dimensional
+standard Gaussian target. The step size is set to δ = 1.0, the number of iterations
+is N = 105, and the experiment is repeated 250 times. The schemes BDB (left) and
+DBD (right) correspond to including the refreshment part in B. In schemes B DR B
+(left) and DR B DR (right) we denote by B the standard bounce part, by DR the
+transition kernel which corresponds to having refreshments and deterministic motion
+together, and we use underscores to divide these two kernels. Here ν is the uniform
+distribution on {±1}.
+4. Expansion of the invariant measure of splitting schemes for BPS
+In this section we investigate the bias in the invariant measure of different splittings of BPS and
+draw conclusions on which schemes perform best. Motivated by Theorems 2.6 and 3.4, we assume
+that the processes corresponding to our splitting schemes have an invariant distribution with density
+µδ(x, v) = µ(x, v)(1 − δ2f2(x, v) + O(δ4)),
+(25)
+where µ(x, v) = ν(v)π(x), π is the target and ν is a distribution satisfying Assumption 4.2 below, e.g.
+the uniform distribution on the unit sphere or the standard Gaussian. It is then our goal to compute
+and compare f2 for different schemes.
+There are several splitting schemes that could be compared, and thus we make a selection of the
+ones it is worth focusing on. The numerical simulations shown in Figure 1 give an idea of the relative
+performance of the schemes. The plots show that the schemes that have DBD as their limit as the
+refreshment rate goes to zero have a smaller bias in the x component compared to those that converge
+to BDB. Naturally the difference between the two schemes is expected to vanish as δ → 0 and also
+appears to be diminishing as the dimension increases (see Figure 4). Based on this result we decide
+to concentrate on schemes RDBDR, DBRBD, DRBRD, as well as BDRDB. Note that all these
+schemes have the same cost of one gradient computation per iteration (in BDRDB it is sufficient to
+keep track of the gradient at the previous iteration).
+Following the approach of [26] , we will show in Section 4.1 (more precisely in Proposition 4.5) that
+the second order of the bias f2 can be computed analytically for one-dimensional targets. We then
+focus on the dependence of f2 on the refreshment rate, which is the only parameter of the algorithm
+(outside of δ).
+As we will see, and as already hinted by Figure 1, some splittings like RDBDR
+and BDRDB are robust to poor choices of the refreshment rate, while others like DBRBD and
+DRBRD have linear or quadratic dependence on λr.
+The numerical experiments of Section 4.2
+confirm the theoretical results of Proposition 4.5 and suggest that the bias behaves similarly in higher
+dimensions, where obtaining f2 analytically is very challenging. In particular, in Figure 3 we show
+that, in the cases we consider, splitting RDBDR is the scheme that shows the best overall behaviour.
+
+SPLITTING SCHEMES FOR PDMPS
+19
+This scheme was shown to be unbiased for standard Gaussian targets in Section 1.2, and is confirmed
+to have f2 = 0 in such cases in Section 4.1. Moreover, we fully characterise the invariant distribution
+of RDBDR in one dimension in Proposition 4.7.
+Note 4.1. In Section 3 we will see cases where a splitting scheme may admit more than one invariant
+measure. In such cases it is not immediately clear what the expansion (25) means. In order to make
+(25) consistent as δ → 0, in those cases we consider µδ as the limit of the law of the splitting scheme
+as the number of steps tends to infinity when the process is started according to µ.
+4.1. Computing f2. Let us discuss briefly how to find f2 with the approach of [26].
+Using the
+Baker-Campbell-Hausdorff (BCH) formula (see e.g. [9]) we can find L2 such that
+Ex,v[f(Xδ, V δ)] = f(x, v) + δLf(x, v) + δ3L2f(x, v) + O(δ4).
+Here L is the infinitesimal generator of the continuous time process.
+Integrating both sides with
+respect to µδ and using that µδ is an invariant measure for the splitting scheme we obtain
+�
+f(x, v)µδ(x, v)dxdv =
+�
+f(x, v)µδ(x, v)dxdv + δ
+�
+Lf(x, v)µδ(x, v)dxdv
++ δ3
+�
+L2f(x, v)µδ(x, v)dxdv + O(δ4).
+Substituting for µδ with the expansion (25) we have
+0 = δ
+�
+Lf(x, v)µ(x, v)dxdv − δ3
+�
+Lf(x, v)µ(x, v)f2(x, v)dxdv + δ3
+�
+L2f(x, v)µ(x, v)dxdv + O(δ4).
+Since µ is an invariant measure for BPS we have
+�
+Lfp dxdv = 0 which gives the equation
+L∗(µf2) = L∗
+2µ.
+(26)
+Here L∗ and L∗
+2 are the adjoints on L and L2 in L2 with respect to Lebesgue measure. Since there is
+not a unique solution to the equation (26) we need to impose a compatibility condition. Since both ˆµ
+and µ are probability densities, integrating (25) gives the requirement
+�
+f2(x, v)µ(x, v)dxdv = 0.
+(27)
+It is then the goal of this section to solve (26) and compare the solutions corresponding to the
+different splitting schemes. We start by computing the term L∗
+2 using the BCH formula. Recall that
+the adjoint of the generator of BPS is given by
+L∗g(x, v) = −⟨v, ∇xg(x, v)⟩ + ((gλ1)(x, R(x)v) − (gλ1)(x, v)) + λr
+�
+ν(v)
+�
+g(x, y)dy − g(x, v)
+�
+.
+We now compute L∗
+2 for the splitting schemes DBRBD, RDBDR, DRBRD, BDRDB. Let us start
+with an assumption on the invariant distribution of the velocity vector.
+Assumption 4.2. The invariant measure for the velocity component ν satisfies the following condi-
+tions:
+(1) Invariance under rotations: ν(w) = ν(v) for any v, w such that |v| = |w|;
+(2) Mean zero: Eν[V ] = 0;
+(3) Isotropic: for some b > 0 it holds that Covν(V ) = bI.
+These properties hold for instance if ν is the standard Gaussian distribution, as well as if ν is the
+uniform on the unit sphere (in that case b = 1/d).
+
+20
+SPLITTING SCHEMES FOR PDMPS
+Proposition 4.3. Let Assumption 4.2 hold and define
+A(x, v) = 3
+2λr
+�
+b tr
+�
+∇ψ(x)∇ψ(x)T − ∇2ψ(x)
+�
++ 2⟨v, ∇ψ(x)⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩
+�
+,
+B(x, v) = 3
+2λ1(x, R(x)v)
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩
+�
++ 1
+2⟨v, ∇x(⟨v, ∇2ψ(x)v⟩)⟩,
+C(x, v) = 3λ1(x, R(x)v)
+�
+− 2⟨v, ∇2ψ(x)v⟩ + ⟨v, ∇ψ(x)⟩2�
+− ⟨v, ∇(⟨v, ∇2ψ(x)v⟩)⟩,
+D(x, v) = 3
+2λr
+�
+b tr
+�
+∇ψ(x)∇ψ(x)T −∇2ψ(x)
+�
++⟨v, ∇2ψ(x)v⟩ + ⟨v, ∇ψ(x)⟩
+�
+3λ1(x, R(x)v) + λ1(x, v)
+��
+.
+The splitting scheme DBRBD satisfies
+L∗
+2µ(x, v) = µ(x, v)
+12
+�
+A(x, v) + B(x, v)
+�
+.
+The splitting scheme RDBDR satisfies
+L∗
+2µ(x, v) = µ(x, v)
+12
+B(x, v).
+The splitting scheme DRBRD satisfies
+L∗
+2µ(x, v) = µ(x, v)
+12
+�
+D(x, v) + B(x, v) + 3
+2λ2
+r⟨v, ∇ψ(x)⟩
+�
+.
+The splitting scheme BDRDB satisfies
+L∗
+2µ(x, v) = µ(x, v)
+12
+�
+− A(x, v) + C(x, v)
+�
+.
+Proof. The proof can be found in Appendix E.
+□
+Note 4.4. Clearly, if L∗
+2µ = 0 then f2 must be a constant that satisfies (27) and hence it must be that
+f2 = 0, i.e. the second order term in µδ is zero. This is the case for instance for scheme RDBDR when
+the target is a multidimensional standard Gaussian. Indeed in Section 1.2 we proved that RDBDR
+is unbiased for standard Gaussian targets, thus this is a consistent result. In the same setting, we
+observe that f2 = 0 for schemes DBRBD and DRBRD when the refreshment rate is λr = 0. This
+is an expected result, as when λr = 0 these schemes coincide with RDBDR. In Figure 2 we confirm
+that, for a one-dimensional standard Gaussian and when λr = 0, the scheme DBD is unbiased, while
+the scheme BDB is of second order.
+Equation (26) is in general hard to solve, as the adjoint of BPS contains both derivatives and
+integrals. Nonetheless, we are able to solve (26) and find f2 in the one-dimensional case, as stated in
+the next Proposition.
+Proposition 4.5. Consider the one-dimensional setting with state space R × {±1} and target distri-
+bution µ(x, v) = π(x)ν(v) with π ∝ exp(−ψ) and ν = Unif({±1}). Let λr ≥ 0 be the refreshment rate.
+Then the function f2 that solves (26) is
+f2(x, +1) = f+
+2 (0) +
+� x
+0
+��λr
+2 + (−∂ψ(y))+
+�
+g(y) − L∗
+2µ(y, +1)
+µ(y, +1)
+�
+dy,
+f2(x, −1) = f2(x, +1) + g(x),
+where
+g(x) = exp (ψ(x))
+� x
+−∞
+�L∗
+2µ(y, +1)
+µ(y, +1)
++ L∗
+2µ(y, −1)
+µ(y, −1)
+�
+exp(−ψ(y))dy,
+f+
+2 (0) = −
+� ∞
+−∞
+�g(x)
+2
++
+� x
+0
+��λr
+2 + (−∂ψ(y))+
+�
+g(y) − L∗
+2µ(y, +1)
+µ(y, +1)
+�
+dy
+�
+π(x)dx.
+
+SPLITTING SCHEMES FOR PDMPS
+21
+Figure 2. Error for the radius statistic for a one-dimensional standard Gaussian tar-
+get. Here λr = 0 for both schemes DBD and BDB. The dashed, blue line corresponds
+to second order convergence. The time horizon is fixed to T = 105 and the number of
+iterations is N = T/δ. The experiment is repeated 250 times.
+Proof. The proof can be found in Appendix D.1.
+□
+Note 4.6. An immediate consequence of Propositions 4.3 and 4.5 is that in the one dimensional case
+the second order term of the bias of scheme RDBDR is always independent of the refreshment rate
+and of v. Indeed applying the propositions we find
+f2(x, v) = f+
+2 (0) − 1
+24
+� x
+0
+ψ(3)(y)dy
+(28)
+with f+
+2 (0) = 1
+24
+� ∞
+−∞
+� x
+0 ψ(3)(y)dyπ(dx).
+In fact, in 1D, for the scheme RDBDR, we can get an explicit expression for the invariant measure.
+Proposition 4.7. Consider the scheme RDBDR for BPS in one dimension, where the velocity is
+refreshed from ν = Unif({±1}). Then for a fixed initial condition x ∈ R and step size δ the distribution
+with support on {y ∈ R : y = x + nδ, n ∈ Z} × {±1} given by
+µδ(y, v) ∝ e−ψδ(y)
+where ψδ(x) = ψ(x) and for y = x + nvδ, n ∈ N
+ψδ(y) = ψ(x) + δ
+n
+�
+ℓ=1
+ψ′(x + (ℓ − 1/2)vδ)
+is stationary for the process. Moreover, under the conditions of Theorem 3.6 we obtain that µδ is
+ergodic, in the sense that for all bounded functions
+lim
+N→∞
+1
+N
+N
+�
+n=1
+f(Xtn, V tn) = µδ(f)
+Px,v − a.s.
+Proof. The proof can be found in Appendix D.2.
+□
+Once again it is clear that the scheme is unbiased in the Gaussian case ψ(x) = x2/(2σ2) (in the
+sense that ψδ(y) = ψ(y) for all y = x + nδv, i.e. the BPS is ergodic with respect to the restriction of
+the true Gaussian target to the grid, and moreover the target measure is invariant for the scheme).
+
+22
+SPLITTING SCHEMES FOR PDMPS
+More generally, for y = x + vnδ with n ∈ N we get
+ψ(y)
+=
+ψ(x) +
+n
+�
+ℓ=1
+� δ/2
+−δ/2
+ψ′(x + v(ℓ − 1/2)δ + u)du
+=
+ψδ(y) + 1
+2
+n
+�
+ℓ=1
+� δ/2
+−δ/2
+u2ψ(3)(x + v(ℓ − 1/2)δ)du + O(nδ5)
+=
+ψδ(y) + δ2
+24
+� y
+x
+ψ(3)(u)du + O(δ4|x − y|) .
+Setting x = 0 this gives ψδ = ψ + δ2f2 + O(δ4) with f2(y) =
+� y
+0 ψ(3)(u)du, which is the same of
+Equation (28). Indeed the term f+
+2 (0) in (28) was introduced to make exp(−ψ)(1+δ2f2) a probability
+distribution and would appear also in the present context. Hence Propositions 4.5 and 4.7 agree.
+4.2. Application to three one-dimensional target distributions. In this section we compare
+the splitting schemes by applying Proposition 4.5 to three one-dimensional target distributions: a
+centred Gaussian distribution, a distribution with non-Lipschitz potential ψ(x) = x4, and a Cauchy
+distribution. The formal statements can be found in the Appendix D.3, correspondingly in Propositions
+D.1, D.2, D.3. Here instead of giving the complicated analytic expressions for f2 in all cases, we give
+plots of the TV distance between µ and µδ as a function of λr as given by Propositions D.1, D.2, D.3.
+The results, both according to the theory and numerical simulations, are shown in Figure 3.
+Let us briefly explain how the TV distance is derived from the analytic expression of f2. We shall
+focus on the position part of µδ, which we denote as πδ. By marginalising and recalling in this context
+ν = Unif({±1}) we obtain
+πδ(x) = π(x)
+�
+1 − δ2
+2 (f2(x, +1) + f2(x, −1))
+�
++ O(δ4),
+(29)
+Using (29) we can express the TV distance between π and πδ as
+∥π − πδ∥TV = δ2
+2 sup
+A
+����
+�
+A
+(f2(x, +1) + f2(x, −1))π(x)dx
+���� + O(δ4).
+(30)
+The δ2 contribution of the rhs can be computed by plugging in the expressions for f2 found in Propo-
+sitions D.1, D.2, and D.3. We neglect higher order terms.
+Let us comment on these results. First of all, the theoretical results are consistent with the numerical
+experiments of Figure 3. Indeed, it is clear that the schemes RDBDR and BDRDB have a bias that
+is independent of the refreshment rate, while DBRBD and DRBRD have respectively linear and
+quadratic dependence. In the one-dimensional case, the plots show that it is best to choose λr = 0,
+which is possible as in this case BPS is irreducible. However, in higher dimensional settings it is
+necessary to take λr > 0 as shown in Figure 4. Since choosing a good value of λr is difficult and
+depends on the target distribution, it is desirable to use schemes that have good performance for most
+values of λr. Moreover, it is clear from Figure 3 that RDBDR is indeed unbiased in the Gaussian
+case, and also has the smallest bias out of all the considered splittings with the exception of the Cauchy
+target, where the difference in performance between RDBDR and BDRDB is almost negligible and
+seems to slightly favour the latter in experiments. In this case, we also see a small dependence on λr
+for RDBDR and BDRDB, which could be due to higher order terms.
+The experiments in Figure 4 suggest that the findings of the one-dimensional case extend to multi-
+dimensional targets. In particular, RDBDR has either a better performance than other splittings
+or behaves very similarly to BDRDB both on an independent as well as a correlated Gaussian.
+Moreover, the independence on λr of the bias of schemes DBRBD and DRBRD is confirmed also
+when d > 1.
+
+SPLITTING SCHEMES FOR PDMPS
+23
+(a) TV distance according to Proposition D.1
+(b) Absolute value of the error for the radius statistic.
+(c) TV distance according to Proposition D.2.
+(d) Absolute value of the error for the radius statistic.
+(e) TV distance according to Proposition D.3.
+(f) Absolute value of the error for min{4, x2}.
+Figure 3. Total variation distance to the true target as given by the second order
+term in (30) (left) and numerical simulations (right) for the various splittings. The top
+row is obtained with standard Gaussian target, the middle row with ψ(x) = x4, and
+the bottom row with a one dimensional Cauchy target with γ = 1. In all plots δ = 0.5
+and the number of iterations is N = 2 · 105. In the Gaussian and Cauchy cases we
+initialise the processes at µ.
+
+24
+SPLITTING SCHEMES FOR PDMPS
+(a) d = 2, ρ = 0.
+(b) d = 10, ρ = 0.
+(c) d = 2, ρ = 0.7.
+(d) d = 10, ρ = 0.7.
+Figure 4. Error of estimators of the radius with splitting schemes for BPS with a
+Gaussian target with covariance Σii = 1, Σij = ρ for i ̸= j. The step size is δ = 0.5
+and the number of iterations is N = 2 · 105. The processes are initialised with a draw
+from µ.
+In conclusion, we have conducted a detailed analysis of the bias in the invariant measure, both
+theoretical in Propositions D.1, D.2, D.3, and empirical in Figures 1, 2, 3, 4, and the evidence suggests
+that RDBDR is the best candidate out of the pool of splitting schemes that are available. The closest
+competitor BDRDB shows similar performance in some settings, but a larger bias in others in which
+RDBDR enjoys desirable properties.
+5. Numerical experiments
+In this section we discuss some numerical simulations for the proposed samplers. The codes for all
+these experiments can be found at https://github.com/andreabertazzi/splittingschemes_PDMP.
+5.1. Gaussian target. Here we study the behaviour of the proposed algorithms on two types of
+Gaussian targets. The first type is a correlated Gaussian, for which the covariance matrix has unitary
+variances and correlation ρ between all components. The second type is an independent Gaussian,
+where components i ≥ 2 have unitary variance, while the first component has (small) variance σ2.
+We study the performance of our algorithms as a function of ρ and σ2, as well as of the step size
+δ and the dimension of the target. In particular we first focus on the number of rejections in the
+Metropolised algorithms, that is Algorithms 3 and 4, and then we focus on the error in the estimation
+of the expected radius for all our algorithms.
+
+SPLITTING SCHEMES FOR PDMPS
+25
+(a) Here δ = 0.3 and d = 20.
+(b) Here δ = 0.3 and d = 20.
+(c) Here ρ = 0.5 and d = 20.
+(d) Here σ2 = 0.1 and d = 20.
+(e) Here δ = 0.3 and ρ = 0.5.
+(f) Here δ = 0.3 and σ2 = 0.1.
+Figure 5. Fraction of rejected proposals in the Metropolis step for Algorithms 3 and
+4. The plots on the left are obtained running the algorithms with a Gaussian target
+with covariance Σii = 1, Σij = ρ for j ̸= i, while the plots on the right with diagonal
+covariance Σ11 = σ2, Σii = 1 for i ̸= 1. In all experiments the refreshment rate for BPS
+is λr = 0.5.
+Figure 5 shows the number of rejections in adjusted algorithms, with the left part of the plot showing
+the first type of Gaussian target and the right part showing the second. This experiment allows us
+to understand the efficiency of the Metropolis adjusted algorithms, as a larger fraction of rejections
+corresponds to more computations required to obtain an accepted state. What we observe is that the
+adjusted ZZS defined in Algorithm 3 is exact for targets with diagonal covariance as expected, but the
+number of rejections increases with the correlation between components of the target. It is well known
+
+26
+SPLITTING SCHEMES FOR PDMPS
+(a) Here δ = 0.3 and d = 20.
+(b) Here δ = 0.3 and d = 20.
+(c) Here ρ = 0.5 and d = 20.
+(d) Here σ2 = 0.1 and d = 20.
+(e) Here δ = 0.3 and ρ = 0.5.
+(f) Here δ = 0.3 and σ2 = 0.1.
+Figure 6. Error for the radius statistic for Algorithms 1, 2, 3, 4, as well as the
+continuous ZZS and BPS. The plots on the left are obtained running the algorithms
+with a Gaussian target with covariance Σii = 1, Σij = ρ for j ̸= i, while the plots on
+the right with diagonal covariance Σ11 = σ2, Σii = 1 for i ̸= 1. In all experiments the
+refreshment rate for BPS is λr = 0.5. The processes are started from a draw of the
+target. The time horizon is T = 103 and the number of iterations is N = ⌈T/δ⌉. The
+radius is estimated with the usual Monte Carlo averages.
+
+SPLITTING SCHEMES FOR PDMPS
+27
+that the continuous time ZZS has lower efficiency for correlated targets (see [2]), and in the case of
+Algorithm 3 this is seen as a large number of reflections. On the other hand, the adjusted BPS given
+by Algorithm 4 appears to suffer when σ2 is small, while the number of rejections remains controlled
+for large correlation ρ.
+Figure 6 shows the error in the estimation of the expected radius for the adjusted and unadjusted
+algorithms, as well as for the continuous BPS and ZZS. As expected, ZZS is sensitive to high correlation
+between components and its error increases with ρ. It is possible to improve in these cases by applying
+the adaptive schemes proposed in [2], which learn the covariance structure of the target and use this
+information to tune the set of velocities of the ZZS suitably. It seems also clear that the schemes based
+on ZZS are more robust when the target is very narrow in some components. This is a reasonable
+behaviour, as DBD schemes for ZZS essentially decompose the target in one dimensional problems,
+hence the chain can explore efficiently some components while being stuck in others. As a consequence,
+in the second type of Gaussian target the chain will rarely move in the component with small variance,
+but it can freely move in the other components. On the other hand, in BPS the switching rate and
+reflection operator are dominated by the component with small variance, thus the whole chain is
+affected by settings with e.g. small variances of some components. We observe that the adjusted BPS
+given by Algorithm 4 is more robust than its unadjusted counterpart. For these reasons, Algorithms
+1, 3, and 4 are to be preferred in case of stiff targets.
+5.2. Image deconvolution using a total variation prior. In this section we test Algorithm 3 on
+an imaging inverse problem, which we solve with a Bayesian approach. In the following we shall refer
+to an image either as a N × N matrix or as a vector of length N2, which is obtained by placing each
+column of the matrix below the previous one. We set d = N2. In both cases each entry corresponds to
+a pixel. Now denote as x ∈ Rd the image we are interested in estimating and y ∈ Rd the observation.
+The observation is related to x via the statistical model
+y = Ax + ξ,
+where A is a d × d-dimensional matrix which may be degenerate and ill-conditioned and ξ a d-
+dimensional Gaussian random variable with mean zero and variance σ2Id. The forward problem we
+consider is given by a blurring operator, i.e. A acts by a discrete convolution with a kernel h. In our
+examples h will be a uniform blur operator with blur length either 9 or 25. The likelihood of y given
+x is given by
+p(y|x) ∝ e−fy(x),
+fy(x) =
+1
+2σ2 ∥Ax − y∥2.
+In the Bayesian approach one then has to place a prior distribution on x. Here we choose the total
+variation prior:
+p(x) ∝ e−g(x),
+where θ > 0 and g(x) = θ∥x∥TV is the total variation of the image x (see [41]) and is given by
+∥x∥TV :=
+N
+�
+i,j=1
+(|xi+1,j − xi,j| + |xi,j+1 − xi,j|).
+The total variation prior corresponds to the ℓ1-norm of the discrete gradient of the image and therefore
+promotes piecewise constant reconstructions. Note that this prior is not smooth and hence we cannot
+directly apply the gradient based algorithms such as the unadjusted Langevin algorithm (ULA), BPS
+or ZZS. Therefore we approximate g with a Moreau-Yosida envelope
+gλ(x) = min
+z∈Rd
+�
+g(z) + 1
+2λ∥x − z∥2
+�
+.
+
+28
+SPLITTING SCHEMES FOR PDMPS
+By [40, Proposition 12.19] we have that gλ is Lipschitz differentiable with Lipschitz constant λ−1 and
+∇gλ(x) = 1
+λ(x − proxλ
+g(x)),
+proxλ
+g(x) = arg min
+z∈Rd
+�
+g(z) + 1
+2λ∥x − z∥2
+�
+.
+Using Bayes theorem, we have the posterior distribution
+π(x) := p(x|y) ∝ e−fy(x)−θgλ(x).
+(31)
+We select the optimal θ by using the SAPG algorithm [44, 17] and we choose λ based on the guidelines
+given in [22], which set λ = 1/Lf where Lf is the Lipschitz constant of fy. Sampling from this model
+using MCMC schemes is difficult because x is usually very high dimensional and the problem is ill-
+conditioned. In this case the unadjusted Langevin algorithm can be very expensive to run since the
+step size is limited by 2/L, where L = Lf + λ−1 is the Lipschitz constant of ∇ log π. Note that we do
+not consider an unadjusted underdamped Langevin algorithm since this algorithm scales poorly (see
+[11, 21]) with the conditioning number which is very large in these examples.
+We are now interested in drawing samples from the posterior (31), and in particular we compare the
+unadjusted ZZS (Algorithm 1, abbreviated as UZZS in the plots)), the unadjusted Langevin algorithm
+(ULA), as well as the continuous ZZS. Indeed, we can compute the Lipschitz constant of the gradient
+of the negative log-posterior, L, and thus we can implement the exact ZZS using the Poisson thinning
+technique based on the simple bound
+λi(x + vt, v) ≤ tL
+√
+d + λi(x, v).
+(32)
+In order to compare the computational cost of the continuous ZZS to the unadjusted ZZS and to ULA
+we count each proposal for an event time obtained by Poisson thinning as a gradient evaluation and
+thus as an iteration. Indeed, an update of the computational bounds requires the evaluation of λi(x, v)
+for all i = 1, . . . , d and thus the full gradient has to be computed. To estimate the posterior mean for
+the continuous ZZS we compute the time average T −1 � T
+0 Xtdt.
+In Figures 7 and 9 we show the original images, the observed images after blurring and adding
+noise, and the estimated posterior mean using the different samplers. Figure 8 shows the mean square
+error (MSE) between the true image and the estimated posterior mean as a function of the number of
+iterations. The MSE is computed for two images x, y ∈ [0, 1]N×N as
+MSE(x, y) =
+1
+N2
+N
+�
+i=1
+N
+�
+j=1
+(xij − yij)2.
+(33)
+It is clear in both Figures 7 and 9 that the unadjusted ZZS shows the fastest convergence to the
+posterior mean. This is clear from visual inspection of the reconstructed images, as the reconstruction
+of ZZS are less noisy after just a few thousand iterations, and even more evident from the MSE shown
+in Figure 8. In the case of Figure 7 it appears ZZS has essentially converged after 105 iterations, while
+it takes around 4 × 105 iterations for ULA to obtain a comparable approximation of the posterior
+mean. This is likely due to the fact that the step size of ULA must be very small or otherwise the
+process tends to infinity, while for ZZS larger step sizes can be selected. For instance in the context of
+Figure 7 the step size for ULA is approximately 1.9×10−6, which is the largest available value without
+going over the stability barrier, while for ZZS the step size is 2.9 × 10−3. This constitutes a major
+difference because every iteration is very computationally intensive, as the target distribution e.g. for
+the context of Figure 9 is of dimension 65536. Notably each iteration involves solving an optimisation
+problem, which is solved by the SAPG algorithm. A similar behaviour is observed for the cameraman
+image shown in Figure 9. In this case the unadjusted ZZS needs around 6×104 gradient evaluations to
+converge, while ULA still has not achieved the same accuracy after 2 × 105 iterations. This difference
+
+SPLITTING SCHEMES FOR PDMPS
+29
+Figure 7. Results for the reconstruction of one of the MNIST handwritten digits using
+a TV prior. The observed image is obtained with a blur length of 9 pixels and then
+adding Gaussian noise with standard deviation σ = 0.0014. The step size for ULA is
+1.98/L, for ZZS is 3000/L, where L ≈ 1042855.81. Mean after 2×103 iterations (second
+column) and after 106 iterations (third column) of the samplers based on the states at
+iterations n × 103 for n = 0, . . . , 100.
+Figure 8. Mean square errors as defined in (33) for the setting of Figure 7 (left) and
+of Figure 9 (right).
+can also be seen from the reconstructions after 2 × 103 iterations shown in Figure 9, as indeed the
+unadjusted ZZS gives a clearly better estimate for the posterior mean. Finally, let us compare the
+unadjusted ZZS with the continuous time ZZS. It is clear from our experiments that ZZS performs
+poorly compared to its discretisation. The reason is twofold. First, the major drawback of Poisson
+thinning using the bounds (32) is that a considerable proportion of the proposed event times are
+rejected (in our examples the rejection rate is around 70 − 80%). Moreover, the rates λi are very
+large in the current framework and the process can have even 109 switches per continuous time unit.
+This means that many gradient computations are required to travel a decent distance and thus the
+
+30
+SPLITTING SCHEMES FOR PDMPS
+Figure 9. Results for the reconstruction of full 256 × 256 pixels cameraman image
+using a TV prior. The observed image is obtained with a blur length of 25 pixels and
+then adding Gaussian noise with standard deviation σ = 0.0021. The reconstructions
+show the estimated mean after 2 × 103 iterations (top row) and after 2 × 105 iterations
+(bottom row) of the samplers based on the states at iterations n×103 for n = 0, . . . , 200.
+The step size for ULA is 1.98/L, for ZZS is 1000/L, where L ≈ 518349.52.
+process itself is expensive to run. The combination of these two phenomena implies an important loss
+of efficiency, which explains the results of our simulations.
+5.3. Chain of interacting particles. Finally, let us consider a problem which will serve as an
+illustration of a typical context where ZZS is favored with respect to other samplers. This is a toy
+model that presents in a simpler form features which are similar to the molecular system considered in
+[37], where splitting schemes involving velocity bounces have proven efficient. We consider a chain of N
+particles in 1D, labeled from 1 to N. The particles interact through two potentials: a chain interaction,
+where the particle i interacts with the particles i − 1 and i + 1; and a mean-field interaction, where
+each particle interacts with all the others. For x ∈ RN, the potential is thus of the form
+ψ(x) =
+N−1
+�
+i=1
+V (xi − xi+1) + a
+N
+N
+�
+i,j=1
+W(xi − xj) ,
+where a > 0 measures the strength of the mean-field interaction, V is the chain potential and W is
+the mean-field potential. In the following we take
+V (s) = s4,
+W(s) = −
+�
+1 + s2 ,
+for s ∈ R, i.e. the chain interaction is an anharmonic quartic potential which constrains two consecutive
+particles in the chain to stay close, while the mean-field interaction induces a repulsion from the rest
+
+SPLITTING SCHEMES FOR PDMPS
+31
+of the system. Although this specific ψ is an academic example meant for illustration purpose, its
+general form is classical in statistical physics.
+Notice that ψ is invariant by translation of the whole system, so that e−ψ is not integrable on RN.
+However we are not interested in the behavior of the barycentre ¯x =
+1
+N
+�N
+i=1 xi, so we consider e−ψ
+as a probability density on the subspace {x ∈ RN, ¯x = 0}, which amounts to looking at the system of
+particles from its center of mass. Anyway, in practice, we run particles in RN without constraining
+their barycentre to zero, which does not change the output as long as we estimate the expectations of
+translation-invariant functions. Specifically, here, we consider the empirical variance of the system
+v(x) =
+1
+N2
+N
+�
+i,j=1
+(xi − xj)2 .
+The important points concerning this model (which are typically met in real molecular dynamics
+as in [37]) are the following.
+The forces ∇ψ can be decomposed in two parts, one of which (the
+chain interaction) is unbounded and not globally Lipschitz but is relatively cheap to compute (with
+a complexity O(N)), while the second part (the mean-field interaction) is bounded but numerically
+expensive (with a complexity O(N2)). If this decomposition is not taken into account, so that we
+simply run a classical MCMC sampler based on the computation of ∇ψ, then the step size has to be
+very small because of the non-Lipschitz part of the forces, and then each step is very costly because of
+the mean-field force. Besides, due to the non-Lipschitz part, sampling a continuous-time PDMP via
+thinning would not be very efficient (in fact in this specific simple case it could be possible to design
+a suitable thinning procedure with some effort, but this would be more difficult with 3D particles and
+singular potentials such as the Lennard-Jones one [37]).
+Now, as was already discussed in Section 1.3 for subsampling, PDMPs and their splitting schemes
+can be used with a splitting of the forces. In the present case, we consider a ZZS where the switching
+rate of the i-th velocity is given by
+λi(x, v) =
+�
+vi(V ′(xi − xi+1) − V ′(xi−1 − xi))
+�
++ + a
+N
+�
+j̸=i
+�
+viW ′(xi − xj)
+�
++ ,
+where, for the particles 1 and N, we set x0 = x1 and xN+1 = xN to cancel out the corresponding
+terms. The corresponding continuous-time ZZS has the correct invariant measure (once centered). We
+consider the DBD splitting to approximate this ZZS (although several other choices are possible, e.g.
+including the Poisson thinning part in the D step and having only jumps according to the potential
+V in the B part). To sample the jump times of the i-th velocity, using that |W ′(s)| ⩽ 1 for all s ∈ R,
+we sample two jump times with rates respectively (vi(V ′(xi − xi+1) − V ′(xi−1 − xi)))+ and a. If both
+times are larger than the step size δ, then the velocity is not flipped. Else, if the time corresponding to
+the first rate is smaller than δ and than that corresponding to the second, then we flip the i-th velocity.
+Alternatively, if the second time is smaller than δ and than the first, we draw J ∼ Unif({1, . . . , N})
+and we flip the sign of the i-th velocity with probability
+(viW ′(xi−xJ))+
+a
+(note that if J = i then this
+probability is indeed 0). Since in this case the rates are not canonical due to the splitting of forces,
+this procedure is repeated until there are no events before the end of the time step. This results in
+O(1) computations per particle on average, hence O(N) for the whole system.
+We compare this scheme with an HMC sampler implemented in the Julia package [46]. This contains
+state of the art techniques and implementation of HMC.
+The results of our simulations are shown in Figure 10. Initially, particles are i.i.d. standard Gaussian
+variables. This gives a configuration which is far from the modes of the target distribution, where
+particles are organized so that i �→ xi is close to the monotonous profile which minimizes ψ (by the
+symmetry i ←→ N + 1 − i there are two such modes, but the empirical variance is unchanged by
+this symmetry so that it is sufficient to see one of them). For both ZZS and HMC, the convergence
+
+32
+SPLITTING SCHEMES FOR PDMPS
+(a) N = 25, a = 0.01.
+(b) N = 25, a = 1.
+(c) N = 25, a = 10.
+(d) N = 50, a = 1.
+(e) N = 100, a = 1.
+Figure 10. Empirical variance (on the y-axis) and runtime in seconds (on the x-
+axis) for various values of N and a in the setting of Section 5.3.
+The green line
+corresponds to the unadjusted ZZS, the red line corresponds to HMC, and the dashed
+black line corresponds to the estimated empirical variance with a long run of HMC
+(when computationally feasible).
+of the estimator is thus essentially driven by a deterministic motion from the biased initial condition
+towards a mode. It is clear that our algorithm gives considerably cheaper yet accurate estimates of the
+empirical variance for all values of a and N considered. This is the result of the subsampling procedure,
+which reduces the cost per iteration from O(N2) to O(N), whereas in each iteration of HMC the full
+mean-field interaction needs to be computed. Notably, as expected the gain in performance increases
+with the number of particles N, which makes the required runtime of HMC prohibitive for large values
+of N.
+Funding. AB acknowledges funding from the Dutch Research Council (NWO) as part of the research
+programme ‘Zigzagging through computational barriers’ with project number 016.Vidi.189.043. PD
+acknowledges funding from the Engineering and Physical Sciences Research Council (EPSRC) grant
+EP/V006177/1. PM acknowledges funding from the French ANR grant SWIDIMS (ANR-20-CE40-
+0022) and from the European Research Council (ERC) under the European Union’s Horizon 2020
+research and innovation program (grant agreement No 810367), project EMC2.
+References
+[1] Christophe Andrieu and Samuel Livingstone. Peskun–tierney ordering for markovian monte carlo: beyond the
+reversible scenario. Annals of Statistics, 49(4):1958–1981, 2021.
+[2] Andrea Bertazzi and Joris Bierkens. Adaptive schemes for piecewise deterministic Monte Carlo algorithms. Bernoulli,
+28(4):2404 – 2430, 2022.
+[3] Andrea Bertazzi, Joris Bierkens, and Paul Dobson. Approximations of piecewise deterministic markov processes and
+their convergence properties. Stochastic Processes and their Applications, 154:91–153, 2022.
+[4] Joris Bierkens, Alexandre Bouchard-Cˆot´e, Arnaud Doucet, Andrew B. Duncan, Paul Fearnhead, Thibaut Lienart,
+Gareth Roberts, and Sebastian J. Vollmer. Piecewise deterministic markov processes for scalable monte carlo on
+restricted domains. Statistics & Probability Letters, 136:148–154, 2018. The role of Statistics in the era of big data.
+
+SPLITTING SCHEMES FOR PDMPS
+33
+[5] Joris Bierkens, Paul Fearnhead, and Gareth Roberts. The zig-zag process and super-efficient sampling for bayesian
+analysis of big data. Annals of Statistics, 47, 2019.
+[6] Joris Bierkens, Sebastiano Grazzi, Frank van der Meulen, and Moritz Schauer. Sticky pdmp samplers for sparse and
+local inference problems. Statistics and Computing, 33(1):8, 2022.
+[7] Joris Bierkens and Gareth Roberts. A piecewise deterministic scaling limit of lifted metropolis–hastings in the
+curie–weiss model. The Annals of Applied Probability, 27(2):846–882, 2017.
+[8] Joris Bierkens, Gareth O Roberts, and Pierre-Andr´e Zitt. Ergodicity of the zigzag process. The Annals of Applied
+Probability, 29(4):2266–2301, 2019.
+[9] Andrea Bonfiglioli and Roberta Fulci. Topics in noncommutative Algebra: The Theorem of Campbell, Baker, Haus-
+dorff and Dynkin, volume 2034. Springer, 01 2012.
+[10] Alexandre Bouchard-Cˆot´e, Sebastian J. Vollmer, and Arnaud Doucet. The bouncy particle sampler: A nonreversible
+rejection-free markov chain monte carlo method. Journal of the American Statistical Association, 113(522):855–867,
+2018.
+[11] Xiang Cheng, Niladri S Chatterji, Peter L Bartlett, and Michael I Jordan. Underdamped langevin mcmc: A non-
+asymptotic analysis. In Conference on learning theory, pages 300–323. PMLR, 2018.
+[12] Augustin Chevallier, Sam Power, Andi Q. Wang, and Paul Fearnhead. Pdmp monte carlo methods for piecewise-
+smooth densities. arXiv.2111.05859, 2021.
+[13] Cloez, Bertrand, Dessalles, Renaud, Genadot, Alexandre, Malrieu, Florent, Marguet, Aline, and Yvinec, Romain.
+Probabilistic and piecewise deterministic models in biology. ESAIM: Procs, 60:225–245, 2017.
+[14] Alice Corbella, Simon E F Spencer, and Gareth O Roberts. Automatic zig-zag sampling in practice. arxiv:2206.11410,
+2022.
+[15] M. H. A. Davis. Piecewise-Deterministic Markov Processes: A General Class of Non-Diffusion Stochastic Models.
+Journal of the Royal Statistical Society. Series B (Methodological), 46(3):353–388, 1984.
+[16] M.H.A. Davis. Markov Models & Optimization. Chapman & Hall/CRC Monographs on Statistics & Applied Prob-
+ability. Taylor & Francis, 1993.
+[17] Valentin De Bortoli, Alain Durmus, Marcelo Pereyra, and Ana Fernandez Vidal. Maximum likelihood estimation of
+regularization parameters in high-dimensional inverse problems: an empirical bayesian approach. part ii: Theoretical
+analysis. SIAM Journal on Imaging Sciences, 13(4):1990–2028, 2020.
+[18] Persi Diaconis, Susan Holmes, and Radford M. Neal. Analysis of a nonreversible Markov chain sampler. The Annals
+of Applied Probability, 10(3):726 – 752, 2000.
+[19] Alain Durmus, Arnaud Guillin, and Pierre Monmarch´e. Geometric ergodicity of the bouncy particle sampler. The
+Annals of Applied Probability, 30(5):2069–2098, 10 2020.
+[20] Alain Durmus, Arnaud Guillin, and Pierre Monmarch´e. Piecewise deterministic Markov processes and their invariant
+measures. Annales de l’Institut Henri Poincar´e, Probabilit´es et Statistiques, 57(3):1442 – 1475, 2021.
+[21] Alain Durmus and Eric Moulines. Sampling from strongly log-concave distributions with the unadjusted langevin
+algorithm. arXiv preprint arXiv:1605.01559, 2016.
+[22] Alain Durmus, ´Eric Moulines, and Marcelo Pereyra. Efficient bayesian computation by proximal markov chain monte
+carlo: When langevin meets moreau. SIAM Journal on Imaging Sciences, 11(1):473–506, 2018.
+[23] Martin Hairer and Jonathan C. Mattingly. Yet another look at Harris’ ergodic theorem for Markov chains. In
+Seminar on Stochastic Analysis, Random Fields and Applications VI, volume 63 of Progr. Probab., pages 109–117.
+Birkh¨auser/Springer Basel AG, Basel, 2011.
+[24] K Hukushima and Y Sakai. An irreversible markov-chain monte carlo method with skew detailed balance conditions.
+Journal of Physics: Conference Series, 473:012012, dec 2013.
+[25] Ben Leimkuhler, Daniel T. Margul, and Mark E. Tuckerman. Stochastic, resonance-free multiple time-step algorithm
+for molecular dynamics with very large time steps. Molecular Physics, 111(22-23):3579–3594, 2013.
+[26] Benedict Leimkuhler and Charles Matthews. Rational Construction of Stochastic Numerical Methods for Molecular
+Sampling. Applied Mathematics Research eXpress, 2013(1):34–56, 06 2012.
+[27] Benedict Leimkuhler, Charles Matthews, and Gabriel Stoltz. The computation of averages from equilibrium and
+nonequilibrium Langevin molecular dynamics. IMA J. Numer. Anal., 36(1):13–79, 2016.
+[28] Vincent Lemaire, Mich`ele Thieullen, and Nicolas Thomas. Exact simulation of the jump times of a class of piecewise
+deterministic markov processes. Journal of Scientific Computing, 2017.
+[29] Vincent Lemaire, Mich´ele Thieullen, and Nicolas Thomas. Thinning and multilevel monte carlo methods for piece-
+wise deterministic (markov) processes with an application to a stochastic morris–lecar model. Advances in Applied
+Probability, 52(1):138–172, 2020.
+[30] P.A.W. Lewis and G.S. Shedler. Simulation of nonhomogeneous poisson processes by thinning, 1978.
+[31] Eva L¨ocherbach and Pierre Monmarch´e. Metastability for systems of interacting neurons. Annales de l’Institut Henri
+Poincar´e, Probabilit´es et Statistiques, 58(1):343 – 378, 2022.
+
+34
+SPLITTING SCHEMES FOR PDMPS
+[32] Sean P. Meyn and R. L. Tweedie. Stability of markovian processes iii: Foster–lyapunov criteria for continuous-time
+processes. Advances in Applied Probability, 25(3):518–548, 1993.
+[33] M. Michel, S. C. Kapfer, and W. Krauth. Generalized event-chain Monte Carlo: Constructing rejection-free global-
+balance algorithms from infinitesimal steps. The Journal of Chemical Physics, 140(5), 2014.
+[34] Pierre Monmarch´e. Piecewise deterministic simulated annealing. ALEA, 13(1):357–398, 2016.
+[35] Pierre Monmarch´e. Kinetic walks for sampling. ALEA Lat. Am. J. Probab. Math. Stat., 17:491–530, 2020.
+[36] Pierre Monmarch´e. High-dimensional MCMC with a standard splitting scheme for the underdamped Langevin
+diffusion. Electronic Journal of Statistics, 15(2):4117 – 4166, 2021.
+[37] Pierre Monmarch´e, Jeremy Weisman, Louis Lagard`ere, and Jean-Philip Piquemal. Velocity jump processes: An
+alternative to multi-timestep methods for faster and accurate molecular dynamics simulations. The Journal of
+Chemical Physics, 153(2):024101, 2020.
+[38] Filippo Pagani, Augustin Chevallier, Sam Power, Thomas House, and Simon Cotter. Nuzz: numerical zig-zag
+sampling for general models. arXiv:2003.03636, 2020.
+[39] E. A. J. F. Peters and G. De With. Rejection-free Monte Carlo sampling for general potentials. Physical Review E
+- Statistical, Nonlinear, and Soft Matter Physics, 85(2):1–5, 2012.
+[40] R Tyrrell Rockafellar and Roger J-B Wets. Variational analysis, volume 317. Springer Science & Business Media,
+2009.
+[41] Leonid I Rudin, Stanley Osher, and Emad Fatemi. Nonlinear total variation based noise removal algorithms. Physica
+D: nonlinear phenomena, 60(1-4):259–268, 1992.
+[42] Konstantin S. Turitsyn, Michael Chertkov, and Marija Vucelja. Irreversible monte carlo algorithms for efficient
+sampling. Physica D: Nonlinear Phenomena, 240(4):410–414, 2011.
+[43] Paul Vanetti, Alexandre Bouchard-Cˆot´e, George Deligiannidis, and Arnaud Doucet. Piecewise-deterministic markov
+chain monte carlo. arXiv:1707.05296, 2017.
+[44] Ana Fernandez Vidal, Valentin De Bortoli, Marcelo Pereyra, and Alain Durmus. Maximum likelihood estimation of
+regularization parameters in high-dimensional inverse problems: An empirical bayesian approach part i: Methodol-
+ogy and experiments. SIAM Journal on Imaging Sciences, 13(4):1945–1989, 2020.
+[45] Marija Vucelja. Lifting – a nonreversible markov chain monte carlo algorithm. American Journal of Physics, 84, 12
+2014.
+[46] Kai Xu, Hong Ge, Will Tebbutt, Mohamed Tarek, Martin Trapp, and Zoubin Ghahramani. Advancedhmc. jl: A ro-
+bust, modular and efficient implementation of advanced hmc algorithms. In Symposium on Advances in Approximate
+Bayesian Inference, pages 1–10. PMLR, 2020.
+Appendix A. Proofs of Section 2
+A.1. Proof of Theorem 2.6.
+Proof. Fix g ∈ C2,0
+Φ ∩ D(L). By a telescoping sum we have
+Ez[g(Ztn)] − Ez[g(Ztn)] =
+n−1
+�
+k=0
+(Ez[Ptn−tk+1g(Ztk+1)] − Ez[Ptn−tkg(Ztk)]).
+For each k ∈ {0, . . . , n − 1}, set fk(y, s) = Ptn−tk−sg(y) then we have
+Ez[g(Ztn)] − Ez[g(Ztn)] =
+n−1
+�
+k=0
+Ez[fk(Ztk+1, δ) − fk(Ztk, 0)].
+By conditioning on Ztk it is sufficient to prove that
+|Ez[fk(Zδ, δ)] − fk(z, 0)| ≤ R(1 + |z|M)∥g∥C2,0
+Φ δ3.
+(34)
+Indeed if we have that (34) holds then by Assumption 2.5 we have
+|Ez[g(Ztn)] − Ez[g(Ztn)]| ≤ Cδ3
+n−1
+�
+k=0
+eR(tn−tk)Ez[G(Ztk)]
+≤ C∥g∥C2,0
+Φ eRtnδ3nGM(z),
+which gives the desired result. It remains to show that (34) holds.
+
+SPLITTING SCHEMES FOR PDMPS
+35
+As done in [3] we rewrite the lhs as
+Ez[fk(Zδ, δ)] − fk(z, 0) = Ez[fk(Zδ, δ)] − fk(ϕδ(z), δ) + fk(ϕδ(z), δ) − fk(z, 0).
+(35)
+In particular, with identical steps to [3] we can rewrite the last two terms on the left hand side of (35)
+using the fundamental theorem of calculus and that ∂sfk(z, s) = −Lfk(z, s):
+fk(ϕδ(z), δ) − fk(z, 0) = −
+� δ
+0
+λ(ϕr(z))[Q(fk(·, r))(ϕr(z)) − fk(ϕr(z), r)]dr.
+Then we compute the expectation in the right hand side of (35), collecting a term for the case of no
+jumps, a single jump and the case of multiple jumps
+Ez[fk(Zδ, δ)] − fk(z, δ) =
+=
+� � δ
+0
+Q(ϕδ/2(z), d˜z)
+�
+fk(ϕδ/2(˜z), δ) − fk(ϕδ(z), δ)
+�
+λ(ϕδ/2(z))e−sλ(ϕδ/2(z))e−(δ−s)λ(˜z)ds
+(†)
++
+∞
+�
+ℓ=2
+Ez[(fk(Zδ, δ) − fk(ϕδ(z), δ))1{ℓ events}]
+(‡)
+−
+� δ
+0
+λ(ϕr(z))[Q(fk(·, r))(ϕr(z)) − fk(ϕr(z), r)]dr.
+(‡†)
+Observe that the sum in the second term (‡) can be truncated from ℓ = 3 onward as we only wish to
+get an order δ3 local error. Indeed, we have |fk| ≤ ∥g∥∞ and hence
+�����
+∞
+�
+ℓ=3
+Ez[(fk(Zδ, δ) − fk(ϕδ(z), δ))1{ℓ events}]
+����� ≤ 2∥g∥∞Pz(ℓ ≥ 3 events)
+≤ 2∥g∥∞
+� � δ
+0
+� δ−s1
+0
+� δ−s1−s2
+0
+λ(ϕδ/2(z))e−s1λ(ϕδ/2(z))λ(z1)e−s2λ(z1)λ(z2)e−s3λ(z2)
+Q(ϕδ/2(z), dz1)Q(z1, dz2)Q(z2, dz3) ds1ds2ds3
+≤ 2∥g∥∞
+�
+(1 − e−δλ(ϕδ/2(z)))(1 − e−δλ(z1))(1 − e−δλ(z2))Q(ϕδ/2(z), dz1)Q(z1, dz2)
+≤ 2δ3∥g∥∞
+�
+λ(ϕδ/2(z))λ(z1)λ(z2)Q(ϕδ/2(z), dz1)Q(z1, dz2)
+≤ 2δ3∥g∥∞
+�
+λ(ϕδ/2(z))λ(z1)Q(ϕδ/2(z), dz1)Qλ(z1)
+≤ 2δ3∥g∥∞ λ(ϕδ/2(z))Q(λ(·)Qλ(·))(ϕδ/2(z))
+where we used that 1 − exp(−z) ≤ z for z ≥ 0.
+By Assumption 2.3 we have that λQ(λQλ) is
+polynomially bounded and therefore we can bound (‡) by
+(‡) =
+�
+Q(ϕδ/2(z), dz1)Q(z1, dz2)
+�
+fk(ϕδ/2(z2), δ) − fk(ϕδ(z), δ)
+�
+·
+·
+� δ
+0
+� δ−s1
+0
+λ(ϕδ/2(z))e−s1λ(ϕδ/2(z))λ(z1)e−s2λ(z1)e−(δ−s1−s2)λ(z2)ds2 ds1
++ O(∥g∥∞(1 + |z|3mλ)δ3).
+Here and throughout we understand F(z, δ, g) = O(∥g∥C2
+Φ(1 + |z|m)δn) to mean that
+lim sup
+δ→0
+sup
+z∈E
+sup
+g
+|F(z, δ, g)|
+∥g∥C2
+Φδn(1 + |z|m) ≤ C.
+
+36
+SPLITTING SCHEMES FOR PDMPS
+We Taylor expand several terms in order to verify that the local error is of order δ3.
+We use
+repeatedly the following expansions:
+λ(ϕs(z)) = λ(z) + sDΦλ(z) + s2R(z, ˜s; λ),
+fk(ϕs(z), δ) = fk(z, δ) + sDΦfk(z, δ) + s2R(z, ˜s; fk),
+R(z, ˜s; g) = D2
+Φg(ϕ˜s(z))/2,
+fk(z, s) = fk(z, 0) − sLfk(z, 0) + s2L2fk(z, ˜s)/2,
+for some ˜s ∈ [0, s] (note that ˜s may vary with each term so when we use these expansions we include
+an index to distinguish different incidents of ˜s). Note that by Assumption 2.4 we have ∥fk∥C2
+Φ ≤
+CeR(tn−tk)(1+|z|mP)∥g∥C2,0
+Φ
+which gives us a bound on the remainder terms. Applying the expansions
+above to (†) we obtain
+(†) =
+�
+Q(ϕδ/2(z), d˜z)
+�
+fk(˜z, 0) − δLfk(˜z, 0) + δ2
+2 L2fk(˜z, ˜s2) + δ
+2DΦfk(˜z, δ) + 1
+8δ2R(˜z, ˜s2; fk)
+− fk(z, 0) + δLfk(z, 0) − 1
+2δ2L2fk(z, ˜s3) − δDΦfk(z, δ) − 1
+2δ2R(z, ˜s3; fk)
+�
+� δ
+0
+�
+λ(z) + δ
+2DΦλ(z) + 1
+8δ2R(z, ˜s4; λ)
+�
+�
+1 − sλ(ϕδ/2(z)) − (δ − s)λ(˜z) + 1
+2(sλ(ϕδ/2(z)) + (δ − s)λ(˜z))2e−η�
+ds
+=
+�
+Q(ϕδ/2(z), d˜z)
+�
+fk(˜z, 0) − fk(z, 0)
+� � δ
+0
+�
+λ(z) + δ
+2DΦλ(z)
+��
+1 − sλ(z) − (δ − s)λ(˜z)
+�
+ds
++ δ
+�
+Q(ϕδ/2(z), d˜z)
+�
+− Lfk(˜z, 0) + 1
+2DΦfk(˜z, 0) + Lfk(z, 0) − DΦfk(z, 0)
+�
+� δ
+0
+�
+λ(z) + δ
+2DΦλ(z)
+��
+1 − sλ(z) − (δ − s)λ(˜z)
+�
+ds + eR(tn−tk)O(∥g∥C2,0
+Φ (1 + |z|M)δ3)
+where we used
+η ∈ [0, sλ(ϕδ/2(z)) + (δ − s)λ(˜z)]
+in the first equality and further Taylor expansions to obtain the second equality. Here M = 3mλ+mP.
+Now using Assumption 2.3 we can expand the Q term
+(†) =
+� �
+Q(z, d˜z) + δ
+2DΦQ(z, d˜z)
+��
+fk(˜z, 0) − fk(z, 0)
+� � δ
+0
+�
+λ(z) + δ
+2DΦλ(z)
+��
+1 − sλ(z) − (δ − s)λ(˜z)
+�
+ds
++ δ
+�
+Q(z, d˜z)
+�
+− Lfk(˜z, 0) + 1
+2DΦfk(˜z, 0) + Lfk(z, 0) − DΦfk(z, 0)
+�
+� δ
+0
+�
+λ(z) + δ
+2DΦλ(z)
+��
+1 − sλ(z) − (δ − s)λ(˜z)
+�
+ds + eR(tn−tk)O(∥g∥C2,0
+Φ (1 + |z|M)δ3)
+Term (‡) can be expanded as
+(‡) =
+�
+Q(ϕδ/2(z), d˜z)Q(z1, dz2)
+�
+fk(z2, 0) − fk(z, 0)
+� � δ
+0
+� δ−s1
+0
+�
+λ(z) + δDΦ(λ)(ϕ˜s4(z))
+�
+λ(z1)
+�
+1 + (−s1λ(ϕδ/2(z)) − s2λ(z1) − (δ − s1 − s2)λ(z2))e−ξ�
+ds2 ds1 + eR(tn−tk)O(∥g∥C1,0
+Φ (1 + |z|mP+3mλ)δ3)
+=δ2
+2
+�
+Q(ϕδ/2(z), d˜z)Q(z1, dz2)
+�
+fk(z2, 0) − fk(z, 0)
+�
+λ(z)λ(z1) + eR(tn−tk)O(∥g∥C1,0
+Φ (1 + |z|mP+3mλ)δ3)
+
+SPLITTING SCHEMES FOR PDMPS
+37
+=δ2
+2
+�
+Q(z, dz1)Q(z1, dz2)
+�
+fk(z2, 0) − fk(z, 0)
+�
+λ(z)λ(z1) + eR(tn−tk)O(∥g∥C1,0
+Φ (1 + |z|mP+3mλ)δ3),
+where
+ξ ∈ [0, s1λ(ϕδ/2(z)) + s2λ(z1) + (δ − s1 − s2)λ(z2)].
+By Assumption 2.3 we can expand the term (‡†) as follows:
+(‡†) = −
+� δ
+0
+λ(ϕr(z))
+�
+Qfk(·, r)(z) + r
+�
+DΦQ(z, d˜z)fk(˜z, r) − fk(ϕr(z), r)
+�
+dr
++ eR(tn−tk)O(δ3∥g∥C2,0
+Φ (1 + |z|mλ)).
+By Taylor’s theorem
+(‡†) = −
+� δ
+0
+λ(ϕr(z))
+�
+Qfk(·, 0)(z) − r
+�
+Q(z, d˜z)Lfk(˜z, 0) + r
+�
+DΦQ(z, d˜z)(fk(˜z, 0) − rLfk(˜z, ˜r))
+− (fk(z, 0) + rDΦfk(z, 0) − rLfk(z, 0))) dr
++ eR(tn−tk)O(δ3∥g∥C2,0
+Φ (1 + |z|mλ+mP).
+Note that
+�
+DΦQ(z, d˜z)Lfk(˜z, ˜r) = Q(DΦLfk(·, ˜r))(z) = eR(tn−tk)O((1 + |z|mP)∥g∥C2,0
+Φ ). Using this
+and Taylor expanding λ(ϕr(z)) we have
+(‡†) = −
+� δ
+0
+λ(z)
+�
+Qfk(·, 0)(z) − r
+�
+Q(z, d˜z)Lfk(˜z, 0) + r
+�
+DΦQ(z, d˜z)fk(˜z, 0)
+− (fk(z, 0) + rDΦfk(z, 0) − rLfk(z, 0))
+�
+dr −
+� δ
+0
+rDΦλ(z) (Qfk(·, 0)(z) − fk(z, 0)) dr
++ eR(tn−tk)O(δ3∥g∥C2,0
+Φ (1 + |z|mλ+mP).
+Evaluating the integral over r
+(‡†) = − λ(z)
+�
+Qfk(·, 0)(z)δ − 1
+2δ2
+�
+Q(z, d˜z)Lfk(˜z, 0) + 1
+2δ2
+�
+DΦQ(z, d˜z)fk(˜z, 0)
+−
+�
+fk(z, 0)δ + 1
+2δ2DΦfk(z, 0) − 1
+2δ2Lfk(z, 0)
+��
+− 1
+2δ2DΦλ(z) (Qfk(·, 0)(z) − fk(z, 0))
++ eR(tn−tk)O(δ3∥g∥C2,0
+Φ (1 + |z|mλ+mP).
+First order terms. Terms of order δ appear only in (†) and (‡†) and clearly they cancel out.
+Second order terms. In (†) we can further expand terms of the form DΦ(f)(˜z, δ) and rearrange
+as
+Order δ2 of (†) =δ2
+� �
+λ(z)
+�
+− Q(z, d˜z)Lfk(˜z, 0) + Q(z, d˜z)1
+2DΦfk(˜z, 0)
++ 1
+2DΦQ(z, d˜z)(fk(˜z, 0) − fk(z, 0))
+�
++ λ(z)(Lfk(z, 0) − DΦfk(z, 0)⟩)
++
+�
+Q(z, d˜z)(fk(˜z, 0) − fk(z, 0))
+�
+− 1
+2λ(z)(λ(˜z) + λ(z)) + 1
+2DΦλ(z)
+��
+=δ2
+� �
+λ(z)
+�
+− Q(z, d˜z)Lfk(˜z, 0) + Q(z, d˜z)1
+2DΦfk(˜z, 0)⟩
+
+38
+SPLITTING SCHEMES FOR PDMPS
++ 1
+2DΦQ(z, d˜z)(fk(˜z, 0) − fk(z, 0))
+�
+��
+�
+Term A
+�
++ λ(z)
+�
+Lfk(z, 0) − DΦfk(z, 0)) − 1
+2λ(z)
+�
+Q(z, d˜z)(fk(˜z, 0) − fk(z, 0))
+�
+�
+��
+�
+Term B
++ 1
+2DΦλ(z)
+�
+Q(z, d˜z)(fk(˜z, 0) − fk(z, 0))
+�
+��
+�
+Term C
+− 1
+2
+�
+Q(z, d˜z)(fk(˜z, 0) − fk(z, 0)
+� �� �
+Term D
+)λ(z)λ(˜z)
+�
+.
+For term (‡) we have
+Order δ2 of (‡) = 1
+2δ2
+�
+Q(z, dz1)Q(z1, dz2)(fk(z2, 0) − fk(z, 0)
+� �� �
+Term D
+)λ(z)λ(z1).
+Similarly for (‡†) we have
+Order δ2 of (‡†) = − δ2
+2
+� �
+DΦ(Q)(z)λ(z)(fk(˜z, 0) − fk(z, 0))
+�
+��
+�
+Term A
++
+�
+Q(z, d˜z)DΦ(λ)(z)(fk(˜z, 0) − fk(z, 0))
+�
+��
+�
+Term C
++
+�
+Q(z, d˜z)λ(z)(−Lfk(˜z, 0) + Lfk(z, 0) − DΦ(fk)(z, 0)
+�
+��
+�
+Term B
+�
+.
+After cancellations we obtain
+Order δ2 of (†) + (‡) + (‡†) = δ2λ(z)
+� � �
+− Q(z, d˜z)Lfk(˜z, 0) + Q(z, d˜z)1
+2DΦ(fk)(˜z, 0)
+�
+− 1
+2
+�
+Q(z, d˜z)fk(˜z, 0)λ(˜z) + 1
+2
+�
+Q(z, d˜z)Q(˜z, dz2)fk(z2, 0)λ(˜z) + 1
+2
+�
+Q(z, d˜z)Lfk(˜z, 0)
+�
+= 1
+2δ2λ(z)
+� �
+Q(z, d˜z)
+�
+− Lfk(˜z, 0) + DΦ(fk)(˜z, 0) + λ(˜z)
+�
+Q(˜z, dz2)[fk(z2, 0) − fk(˜z, 0)]
+��
+= 0,
+where the last equality follows by the definition of the generator of the PDMP. Therefore we have
+shown that second order terms cancel out.
+□
+A.2. Proofs for Example 2.7. Let us verify that Assumption 2.3 holds. Note that
+(DΦQ)(g)(x, v) = vT
+d
+�
+i=1
+∇x
+�λi(x, v)
+λ(x, v) g(x, Fiv)
+�
+.
+
+SPLITTING SCHEMES FOR PDMPS
+39
+Therefore by Taylor’s theorem we have for some η between x and x + sv
+Qg(x + sv, v) − Qg(x, v) − (DΦQ)(g)(x, v) = 1
+2s2
+d
+�
+i=1
+vT ∇2
+x
+�λi(x, v)
+λ(x, v) g(x, Fiv)
+� ����
+x=η
+v.
+It remains to show that λi
+λ , ∇x
+�
+λi(x,v)
+λ(x,v)
+�
+, ∇2
+x
+�
+λi(x,v)
+λ(x,v)
+�
+are bounded. It is clear that 0 < λi/λ ≤ 1. Let
+us consider the first derivative. Set Ξ(s) = log(1 + es) so that λi(x, v) = Ξ(vi∂iψ(x)) and note that
+0 ≤ Ξ′(s)
+Ξ(s) ≤ 1,
+0 ≤ Ξ′′(s)
+Ξ(s) ≤ 1.
+Now
+����∇x
+�λi(x, v)
+λ(x, v)
+����� =
+����
+∇xλi(x, v)
+λ(x, v)
+− λi(x, v)∇xλ(x, v)
+λ(x, v)2
+���� ≤
+����
+∇xλi(x, v)
+λ(x, v)
+���� +
+d
+�
+j=1
+����
+∇xλj(x, v)
+λ(x, v)
+���� .
+So it remains to show that ∇xλi/λ is bounded. Using the bounds on Ξ we have
+����
+∇xλi(x, v)
+λ(x, v)
+���� ≤
+����
+∇xλi(x, v)
+λi(x, v)
+���� =
+����
+Ξ′(vi∂iψ(x))
+Ξ(vi∂iψ(x)) vi∇x∂iψ(x)
+���� ≤ |∇x∂iψ(x)| .
+This is bounded by our assumptions on ψ. Let us now consider ∇2
+x
+�
+λi(x,v)
+λ(x,v)
+�
+:
+����∇2
+x
+�λi(x, v)
+λ(x, v)
+����� =
+�����
+∇2
+xλi(x, v)
+λ(x, v)
+− 2∇xλi(x, v)∇xλ(x, v)T
+λ(x, v)2
++ 2λi(x, v)∇xλ(x, v)∇xλ(x, v)T
+λ(x, v)3
+− λi(x, v)∇2
+xλ(x, v)
+λ(x, v)2
+�����.
+Using the bound on ∇xλi/λ we can bound all the terms aside from the one involving the second
+derivative of λ so it suffices to consider
+����
+∇2
+xλi(x, v)
+λ(x, v)
+���� =
+�����
+Ξ′′(vi∂iψ(x))
+�
+j Ξ(vj∂jψ(x))∇x∂iψ(x)∇x∂iψ(x)T +
+Ξ′(vi∂iψ(x))
+�
+j Ξ(vj∂jψ(x))vi∇2
+x∂iψ(x)
+�����
+≤
+��∇x∂iψ(x)∇x∂iψ(x)T �� +
+��∇2
+x∂iψ(x)
+��
+This is bounded since ψ has bound second and third order derivatives.
+Appendix B. Proofs of ergodicity for splitting schemes of the BPS
+To fix ideas, we consider a splitting scheme for a BPS with unitary velocity with generator L
+decomposed as L = L1 + L2 + L3 with
+L1f(x, v) = vT ∇xf(x, v),
+L2f(x, v) =
+�
+vT ∇ψ(x)
+�
++ (f(x, R(x)v) − f(x, v)) ,
+L3f(x, v) = λr
+�
+Sd−1 (f(x, w) − f(x, v)) dw .
+Write P j
+t = etLj the associated semi-groups for j ∈ �1, 3�. We shall show the following statement,
+which implies Theorem 3.4.
+Theorem B.1. Consider any scheme of the BPS based on the decomposition D,R,B. Under Assump-
+tion 3.3, the following hold:
+
+40
+SPLITTING SCHEMES FOR PDMPS
+(1) There exist a, b, C, δ0 > 0 and a function V (with a, b, C, δ0, V depending only on ψ and λr,
+but not on δ) such that, for all x, v,
+e|x|/a/a ⩽ V (x, v) ⩽ aea|x|
+and for all δ ∈ (0, δ0] and all x, v,
+PδV (x, v) ⩽ (1 − bδ) V (x, v) + Cδ .
+(2) For all R > 0, there exist c, δ0 > 0 and a probability measure ν on E such that for all x, v with
+|x| ⩽ R and all δ ∈ (0, δ0], setting n∗ = ⌈4R/δ⌉,
+δx,vP n∗
+δ
+⩾ cν .
+B.1. Lyapunov function. Under Assumption 3.3, let W(x) =
+�
+ψ(x) so that ∇W is bounded and,
+as |x| → ∞, lim inf |∇W(x)| > 0 and ∇2W(x) → 0.
+Parameters. Let cW > 0 be such that |∇W(x)| ⩾ cW for |x| large enough. Let
+q =
+�
+Sd−1 1w1⩽−1/2dw > 0 .
+Let ϕ ∈ C2(R) be a non-decreasing function such that ϕ′(0) > 0 and
+ϕ(θ)
+�
+�
+�
+= 1
+for θ ⩽ − 1
+2cV
+⩾ 2 − q
+2
+for θ ⩾ − 1
+4cV
+= 2
+for θ ⩾ 1.
+Let
+κ = 4λr
+cW
+,
+M =
+κ
+ϕ
+�
+qλr
+4κ
+�
+− ϕ
+�
+− qλr
+4κ
+�
+(where we used that ϕ′(0) > 0 so that the denominator is positive). Finally, let
+ε =
+λrq
+16∥ϕ′∥∞
+and let R > 0 be such that for all x ∈ Rd with |x| ⩾ R,
+|∇W(x)| ⩾ cW ,
+2W(x) ⩾ M,
+and
+∥∇2W(x)∥ ⩽ ε .
+Preliminary computation. In the following we denote by the same letter C various constants (inde-
+pendent from t) and we assume that t ∈ (0, δ0] with δ0 = 1/(∥∇W∥∞κ). We consider a Lyapunov
+function
+V (x, v) = eκW(x)ϕ(θ(x, v)) ,
+θ(x, v) = vT ∇W(x) .
+Notice that |θ| is bounded by ∥∇W∥∞. More generally, for a non-negative C2 function g, we bound
+P 1
+t
+�
+eκW g ◦ θ
+�
+(x, v)
+=
+eκW(x+tv)g(vT ∇W(x + tv)),
+using that etr ⩽ 1 + tr + r2t2erδ0/2 for t ⩽ δ0 as
+eκW(x+tv)
+⩽
+eκ(W(x)+tvT ∇W(x)+ 1
+2 t2∥∇2W∥∞)
+⩽
+eκW(x) �
+1 + tκvT ∇W(x) + Ct2�
+and, similarly,
+g(v · ∇W(x + tv))
+⩽
+g(θ(x, v)) + tvT ∇2W(x)vg′(θ(x, v)) + Ct2
+⩽
+g(θ(x, v)) +
+�
+ε∥g′∥∞ + C1|x|⩽R
+�
+t + Ct2
+
+SPLITTING SCHEMES FOR PDMPS
+41
+(here and below the constants C implicitly involve the suprema of g, |g′| and |g′′| over [−∥∇W∥∞, ∥∇W∥∞]).
+Combining these two bounds and using that |tκθ| ⩽ 1, we get
+P 1
+t
+�
+eκW g ◦ θ
+�
+⩽ eκW �
+(1 + tκθ)g ◦ θ + ε∥g′∥∞t + Ct2�
++ Ct .
+(36)
+Second, for any function g,
+P 2
+t
+�
+eκW g ◦ θ
+�
+(x, v)=eκW(x) �
+e−t(vT ∇ψ(x))+g
+�
+vT ∇W(x)
+�
++
+�
+1 − e−t(vT ∇ψ(x))+�
+g
+�
+(R(x)v)T ∇W(x)
+��
+=eκW(x) �
+g(θ(x, v)) +
+�
+1 − e−2tW(x)θ+(x,v)�
+(g (−θ(x, v)) − g (θ(x, v)))
+�
+.
+In the second equality we used (vT ∇ψ(x))+ = 2W(x)(vT ∇W(x))+. In particular, if g is non-decreasing,
+P 2
+t
+�
+eκW g ◦ θ
+�
+(x, v) ⩽ eκW(x) �
+g(θ(x, v)) +
+�
+1 − e−tMθ+(x,v)�
+(g (−θ(x, v)) − g (θ(x, v))) + tC1|x|⩽R
+�
+⩽ eκV (x) �
+g(θ(x, v)) + tMθ+(x, v) (g (−θ(x, v)) − g (θ(x, v))) + Ct2�
++ tC1|x|⩽R .
+(37)
+Third, for any function g, similarly,
+P 3
+t
+�
+eκW g ◦ θ
+�
+(x, v) = eκW(x)
+�
+g (θ(x, v)) +
+�
+1 − e−λrt� ��
+Sd−1 g(wT ∇W(x))dw − g (θ(x, v))
+��
+⩽ eκW(x)
+�
+g (θ(x, v)) + λrt
+��
+Sd−1 g(w · ∇W(x))dw − g (θ(x, v))
+�
++ Ct2
+�
+(38)
+In particular, for g = ϕ, using the rotation-invariance of the uniform measure on the sphere and the
+fact that ϕ is increasing and with values in [1, 2], we bound
+�
+Sd−1 ϕ(w · ∇W(x))dw =
+�
+Sd−1 ϕ(w1|∇W(x)|)dw ⩽ 1|x|⩽R + A
+with
+A = sup
+|x|⩾R
+�
+Sd−1 ϕ(w1|∇W(x)|)dw .
+Combining the computations. Now, to fix ideas, let us start with the DRBRD case. From our previous
+computations, using first (36) and then (38) (keeping track only of the terms of order 0 or 1 in t, the
+higher order ones being bounded and gathered in the term t2C),
+P 1
+t P 3
+t P 2
+2tP 3
+t P 1
+t V
+⩽
+P 1
+t P 3
+t P 2
+2tP 3
+t
+�
+eκW �
+(1 + tκθ)ϕ(θ) + tε∥ϕ′∥∞ + t2C
+��
++ tC
+⩽
+P 1
+t P 3
+t P 2
+2t
+�
+eκW �
+(1 + tκθ)ϕ(θ) + λrt(A − ϕ(θ)) + tε∥ϕ′∥∞ + t2C
+��
++ tC
+:=
+P 1
+t P 3
+t P 2
+2t
+�
+eκW Ψ(θ)
+�
++ tC .
+Assume that t ⩽ 1/(2λr + 2∥W∥∞κ). Under this condition Ψ(θ) is non-decreasing in θ, since ϕ is
+non-decreasing in θ and (1 + t(κθ − λr)) > 0. Thus, applying (37) and then (38) and (36) again,
+P 1
+t P 3
+t P 2
+2tP 3
+t P 1
+t V
+⩽
+P 1
+t P 3
+t
+�
+eκW �
+Ψ(θ) + 2tMθ+ (Ψ(−θ) − Ψ(θ)) + t2C
+��
++ tC
+⩽
+P 1
+t P 3
+t
+�
+eκW �
+Ψ(θ) + 2tMθ+ (ϕ(−θ) − ϕ(θ)) + t2C
+��
++ tC
+⩽
+P 1
+t
+�
+eκW �
+(1 + tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tMθ+ (ϕ(−θ) − ϕ(θ)) + tε∥ϕ′∥∞ + t2C
+��
++ tC
+⩽
+eκW �
+(1 + 2tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tMθ+ (ϕ(−θ) − ϕ(θ)) + 2tε∥ϕ′∥∞ + t2C0
+�
++ tC ,
+
+42
+SPLITTING SCHEMES FOR PDMPS
+for some C0, C > 0 (in the last inequality, we use that t is small enough so that the function which
+multiplies eκW is non-negative in order to apply (36)). The choice of ε ensures that
+4ε∥ϕ′∥∞ ⩽ λrq/4 =: η.
+So, if we have that, for all r ∈ R,
+κr + λ(A − ϕ(r)) + Mr+ (ϕ(−r) − ϕ(r)) ⩽ −η ,
+(39)
+then, using that ϕ(θ) ∈ [1, 2], we will conclude that, for t ⩽ min(η/C0, 1/(λr + ∥W∥∞κ))/2,
+P 1
+t P 3
+t P 2
+2tP 3
+t P 1
+t V ⩽ eκW
+�
+ϕ(θ) − 2ηt + 1
+2ηt + 1
+2ηt
+�
++ tC ⩽
+�
+1 − tη
+2
+�
+V + tC ,
+which will conclude the proof of the first part of Theorem 3.4 for the scheme DRBRD.
+Let us check what happens for different splitting orders. The computations are similar, in particular
+an important point is to check that the bound for P 1
+t (resp. P 2
+t ) is always used for a function g which is
+non-negative (resp. non-decreasing in θ), which is ensured each time by the assumption that t is small
+enough, as in the previous DRBRD case. For instance, for BDRDB, we get, for t small enough,
+P 2
+t P 1
+t P 3
+2tP 1
+t P 2
+t V
+⩽
+P 2
+t P 1
+t P 3
+2tP 1
+t
+�
+eκW �
+ϕ(θ) + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C
+��
++ tC
+⩽
+P 2
+t P 1
+t P 3
+2t
+�
+eκW �
+(1 + tκθ)ϕ(θ) + tε∥ϕ′∥∞ + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C
+��
++ tC
+⩽
+P 2
+t P 1
+t
+�
+eκW �
+(1 + tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + tε∥ϕ′∥∞ + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C
+��
++ tC
+⩽
+P 2
+t
+�
+eκW �
+(1 + 2tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tε∥ϕ′∥∞ + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C
+��
++ tC
+⩽
+eκW �
+(1 + 2tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tε∥ϕ′∥∞ + 2tMθ+(ϕ(−θ) − ϕ(θ)) + t2C
+�
++ tC ,
+which is exactly the same bound as in the DRBRD case, as expected since we only keep track of the
+first order terms in t, which coincide for all splitting orders. The other cases are similar.
+Conclusion. Let us check that (39) holds with our choices of parameters. Using that ϕ(r) ⩽ 2 for
+all r ∈ R and ϕ(r) = 1 for all r ⩽ −cV /2 and that |∇V (x)| ⩾ cV if |x| ⩾ R,
+A ⩽ q × 1 + (1 − q) × 2 = 2 − q .
+Then, for r ⩾ −cV /4, since ϕ(r) ⩾ 2 − q/2,
+A − ϕ(r) ⩽ −q/2 .
+The choice of κ ensures that, for all r ⩽ −cV /4,
+κr + λr(A − ϕ(r)) ⩽ κr + λr(1 − q) ⩽ −λrq .
+For r ∈ [cV /4, qλr/(4κ)],
+κr + λr(A − ϕ(r)) ⩽ κr − λrq/2 ⩽ −λrq/4 .
+Finally, for r ⩾ qλr/(4κ), the choice of M ensures that
+κr + λr(A − ϕ(r)) + Mr (ϕ(−r) − ϕ(r))
+⩽
+�
+κ − M
+�
+ϕ
+�qλr
+4κ
+�
+− ϕ
+�
+−qλr
+4κ
+���
+r − λrq/2
+⩽
+−λrq/2 ,
+which concludes the proof of the first part of Theorem 3.4.
+
+SPLITTING SCHEMES FOR PDMPS
+43
+B.2. Doeblin condition. The proof is an adaptation from [34, Lemma 5.2] (see also [19, Lemma
+11]), with the additional difficulty that time is discrete. Fix R > 0 and consider an initial condition
+(x, v) ∈ Rd × Sd−1 with |x| ⩽ R. We want to prove that the law of the process at time n∗ = ⌈4R/δ⌉ is
+bounded below by a positive constant (independent from δ small enough) times the Lebesgue measure
+on some domain, uniformly in x with |x| ⩽ R. Since the process moves at unitary speed, for all n ∈ N,
+|Xn| ⩽ R + nδ, and thus
+P (no bounce in the n first steps) ⩾ e−h(nδ)
+with h(t) = sup{∥∇ψ(x)|, |x| ⩽ R + t}. In the absence of bounces, depending on the splitting order,
+the chain behaves either as the chain given by the splitting DRD or RDR. To fix ideas, we only
+consider the DRD case in the following (corresponding either to DRBRD, DBRBD or BDRDB),
+the proof being similar in the other case.
+For k ⩾ 1, denote by Tk the number of transitions of the chain between the (k − 1)th refreshment
+and the kth one, so that (Tk)k∈N is an i.i.d. sequence of geometric random variables with parameter
+1−e−δλr, independent from the random variables used to define the bounces. Similarly, let (Vk)k∈N be
+the i.i.d. sequence of random variables uniformly distributed on the sphere used to define the velocities
+at the refreshment times. For a small parameter η > 0 to be fixed later on, consider the events
+A1 = {T1 + T2 + T3 ⩽ n∗, |δTj − 2R| ⩽ ηR for j = 2, 3}
+A2 = {δ(T4 − 1) > 2ηR, no bounce in the n∗ first steps} .
+In particular, under A1 ∩ A2, exactly three refreshments occur during the n∗ first transitions. Then,
+for all Borel set B of Rd × Sd−1,
+P
+�
+Zn∗ ∈ B
+�
+⩾
+P
+�
+Zn∗ ∈ B, A1, A2
+�
+=
+P
+�
+( ˜X, V3) ∈ B, A1, A2
+�
+=
+P(A2)P
+�
+( ˜X, V3) ∈ B, A1
+�
+with, in the DRD case for instance,
+˜X = x + δ
+��
+T1 − 1
+2
+�
+v + T2V1 + T3V2 +
+�
+n∗ − T1 − T2 − T3 + 1
+2
+�
+V3
+�
+.
+(Indeed, the DRD scheme starts and ends with a half step of transport). Notice that P(A2) is lower
+bounded uniformly in δ ∈ (0, 1] as
+P(A2) ⩾ e−h(4R+1)e−λr(2ηR+1) .
+On the other hand, writing
+I = {(t1, t2, t3) ∈ N3, t1 + t2 + t3 ⩽ n∗, |δtj − 2R| ⩽ ηR for j = 2, 3} ,
+we get
+P
+�
+( ˜X, V3) ∈ B, A1
+�
+=
+�
+I
+(1 − e−δλr)3e−δλr(t1+t2+t3−3)
+�
+(Sd−1)3 1B
+�
+x + δ
+��
+t1 − 1
+2
+�
+v + t2v1 + t3v2 +
+�
+n∗ − t1 − t2 − t3 + 1
+2
+�
+v3
+�
+, v3
+�
+dv1dv2dv3
+⩾
+�
+I
+(1 − e−δλr)3e−δλr(t1+t2+t3−3)
+inf
+|x′|⩽R(1+2η)
+�
+(Sd−1)3 1B
+�
+x′ + δ(t2v1 + t3v2), v3
+�
+dv1dv2dv3 .
+
+44
+SPLITTING SCHEMES FOR PDMPS
+By the rotation invariance of the uniform measure on the sphere, for fixed t, s > 0, sV1 + tV2 has the
+same distribution as V1|sw+tV2| where w is a fixed unitary vector. Then, |sw+tV2|2 = s2+t2+2stV T
+2 w
+and V T
+2 w has on [−1, 1] the probability density proportional to u �→ (1 − u2)d/2−1 (this is here we use
+that d ≥ 2), which is lower bounded by a positive constant on [−1 + ε, 1 − ε] for all ε > 0. Assuming
+that η ⩽ 1/4 and considering for y ∈ Rd the ring R(y) = {x ∈ Rd, 4ηR ⩽ |x − y| ⩽ 4R(1 − 2η)},
+we get that sV2 + tV3 has a density which is lower bounded on R(0) by a constant α > 0 which is
+independent from t, s ∈ [R(2 − η), R(2 + η)].
+As a consequence, for (t1, t2, t3) ∈ I and x′ ∈ Rd with |x′| ⩽ R(1 + 2η), the law of (x′ + δ(t2V1 +
+t3V2), V3) has a density lower bounded by α on R(x′) × Sd−1. Assuming that η ⩽ 1/16, let R′ = {y ∈
+Rd, R(1 + 6η) ⩽ |y| ⩽ R(3 − 10η)} (which has a non-zero Lebesgue measure). The triangle inequality
+implies that R′ ⊂ R(x′) if |x′| ⩽ R(1 + 2η). As a consequence,
+P
+�
+( ˜X, V3) ∈ B, A1
+�
+⩾ P (A1) α
+�
+R′×Sd−1 1B(y, v)dydv ,
+which concludes since P(A1) converges to a positive constant as δ → 0.
+Appendix C. Ergodicity for splitting schemes of ZZS
+C.1. Splitting DBD. In this Section we focus on splitting scheme DBD for ZZS. We shall prove the
+following, which implies Theorem 3.6.
+Theorem C.1. Consider the splitting scheme DBD for ZZS. Suppose Assumption 3.5 holds. Then
+the following hold:
+(1) There exists a function V : Rd × {±1}d → R, and constants b ∈ (0, 1), C < ∞ such that for all
+(x, v) and all δ ∈ (0, δ0) with δ0 = 2(1 + γ0)−1, where γ0 is as in Assumption 3.5(b), it holds
+that
+PδV (x, v) ≤ (1 − bδ)V (x, v) + Cδ.
+(40)
+(2) For any R > 0 consider a set C = [−R, R]d. For some L > 0 let
+n∗ = 2 + 4x0 + 2R
+δ
++ 2
+�L
+δ
+�
+∈ 2N.
+For (x, v) ∈ C × {±1}d define the set D(x, v) given by (21). Then for any (y, w) ∈ D(x, v) ∩
+(C × {±1}d) and δ ∈ (0, δ0] for δ0 > 0 it holds that
+δy,wP n∗
+δ
+⩾ cν .
+where c is independent of δ and ν is uniform over D(x, v) ∩ (C × {±1}d).
+In Section C.1.1 we prove the minorisation condition, in Section C.1.2 we prove the drift condition,
+while in Section C.1.3 we prove Equation (24).
+C.1.1. Minorisation condition. We now prove a minorisation condition for splitting scheme DBD of
+ZZS. In the following Lemma we consider the one-dimensional setting, for which the reasoning is
+similar to that of the proof of a minorisation condition for the continuous ZZS done in [2, Lemma
+B.2].
+Lemma C.2. Consider the splitting scheme DBD of ZZS with step size δ ≤ δ0. Suppose Assumption
+3.5(a) holds for some x0 ≥ 0 and consider a set C = [−R, R] for R > 0. For L > 0 let
+N = 2 + 4x0 + 2R
+δ
++ 2
+�L
+δ
+�
+∈ 2N.
+(41)
+
+SPLITTING SCHEMES FOR PDMPS
+45
+For (x, v) ∈ C × {±1} define the set D(x, v) := D+(x, v) ∪ D−(x, v), where
+D+(x, v) := {(y, w) : w = v, y = x + mδ, m ∈ 2Z},
+D−(x, v) := {(y, w) : w = −v, y = x + mδ, m ∈ 2Z + 1}.
+(42)
+Then for any (y, w) ∈ D(x, v) ∩ (C × {±1}) it holds that
+P(y,w)((XN, VN) ∈ ·) ≥ bν(·)
+where b is independent of δ and ν is uniform over D(x, v) ∩ (C × {±1}).
+Proof. Let C = [−R, R] for a fixed R > 0 and let x ∈ C. We shall consider only the case of v = +1, as
+the same arguments extend to the symmetric case v = −1. In particular observe that if the process is
+started in set D(x, +1) (respectively D(x, −1)), then after an even number of iterations it will again be
+in D(x, +1) (D(x, −1)). This means that the process lives on D(x, +1) (respectively on D(x, −1)). To
+shorten the notation we denote by D+, D− the sets D+(x, +1), D−(x, +1) as defined in (42). Below
+we focus on the case of an initial condition in D+, while the case of D− follows with an identical
+reasoning and obvious changes.
+Fix N ∈ 2N and define
+λ := max
+x∈C
+max
+y:|y−x|≤Nδ,v=±1 λ(y, v)
+which is the largest switching rate that can be reached within N iterations starting in C. Note that
+taking N as in (41) implies that Nδ is upper bounded by a constant as δ ≤ δ0 and thus λ can be
+chosen independently of the step size δ. Recall λ > 0.
+From here on we shall denote the initial condition as (y, w) ∈ D+, and without loss of generality we
+shall assume x0 = x+ℓδ for some ℓ ∈ N, where x0 is as in Assumption 3.5(a). We want to lower bound
+the probability that after N iterations the process is in measurable sets B ⊂ D. We consider two cases:
+in the first one the final state of the process is of the form (XN, VN) = (z, −1) ∈ D− ∩ (C × {±1}),
+while in the second case (XN, VN) = (z, +1) ∈ D+ ∩ (C × {±1}).
+First case. Consider the case in which the final state has negative velocity, i.e. VN = −1. To lower
+bound the probability of reaching this state, we consider the case in which only one switching event
+takes place. Let z = y + mδ with m ∈ N odd. Then in order to have (XN, VN) = (z, −1) with exactly
+one event taking place at time N1 it must be that
+y + (N1 − 1)δ − (N − N1)δ = z.
+Thus we find that the event should take place at
+N1 = z − y
+2δ
++ N + 1
+2
+.
+In order to guarantee the switching rate is strictly positive it must also be that XN1 ≥ x0, i.e.
+y + (N1 − 1)δ ≥ x0 and thus N1 ≥ 1 + (x0 − y)/δ. Note N1 < N, where N is as in (41). Denote the
+position at the time of the switching event by ˜x = y + δ(N1 − 1/2). Then probability of exactly one
+event taking place at iteration N1 is given by
+� δ
+0
+λ(˜x, 1) exp(−sλ(˜x, 1)) exp(−(δ − s)λ(˜x, −1))ds ≥ δλ exp(−δλ).
+The probability of this path is simple to lower bound, since upper bounding the switching rates gives
+a smaller probability:
+P(y,+1)((XN, VN) = (z, −1)) ≥
+N1−1
+�
+n=0
+exp(−δλ(y + (n + 1/2)δ))
+�
+��
+�
+no jumps before N1
+× δλ exp(−δλ)
+�
+��
+�
+a jump at N1
+×
+
+46
+SPLITTING SCHEMES FOR PDMPS
+×
+N−N1
+�
+n=0
+exp(−δλ(y + (N1 − 1 − n)δ))
+�
+��
+�
+no jumps after N1
+≥ exp(−(N1 − 1)δλ) δλ exp(−δλ) exp(−(N − N1)δλ)
+≥ 2 exp(−(N − 1)δλ) λ exp(−δ0λ) ×
+�1
+2
+δM
+M
+�
+≥ 2 exp(−(N − 1)δλ) λ exp(−δ0λ)(2R − δ)
+�
+ν(−1) × 1
+M
+�
+,
+where M ∈ N is the number of points in D+ ∩ (C × {±1}). In the last line we used that δM ≥ 2R − δ.
+Recall that δ ≤ δ0 and that N is given by (41). This concludes as (N − 1)δ ≤ 4x0 + R + 2L + 3δ0 and
+2R − δ ≥ 2R − δ0.
+Second case. We now focus on the case in which VN = +1. We shall find an appropriate lower bound
+by restricting to the case in which exactly two switching events take place. Denoting the times of the
+two events as N1, N2, if the final position is z it must be
+y + (N1 − 1)δ − (N2 − 1)δ + (N − N1 − N2)δ = z
+which implies
+N2 = y − z
+2δ
++ N
+2 .
+(43)
+Moreover, at event times the process should be in regions with strictly positive switching rate:
+y + (N1 − 1/2)δ ≥ x0,
+y + (N1 − 1)δ − (N2 − 1/2)δ ≤ −x0.
+These imply respectively
+N1 ≥ x0 − y
+δ
++ 1 =: N1,
+N2 ≥ y + x0
+δ
++ N1.
+Since N2 is determined by (43), we enforce that the second inequality holds:
+y − z
+2δ
++ N
+2 ≥ y + x0
+δ
++ N1
+which implies
+N1 ≤ N
+2 − y + 2x0 + z
+2δ
+=: N1.
+Now to obtain the right dependence on δ, we shall take N such that N1 − N1 is increasing as 1/δ. It
+holds
+N1 − N1 = N − 2
+2
+− 4x0 − y + z
+2δ
+and thus it is sufficient to take
+N = 2 + 4x0 − y + z
+δ
++ 2
+�L
+δ
+�
+for some constant L > 0, as with this choice N1 − N1 = ⌈L/δ⌉.
+
+SPLITTING SCHEMES FOR PDMPS
+47
+Using the results above we find
+P(y,+1)((XN, VN) = (z, +1)) ≥
+N1
+�
+N1=N1
+� N1−1
+�
+n=0
+exp(−δλ(y + (n + 1/2)δ))
+�
+��
+�
+no jumps before N1
+× δλ exp(−δλ)
+�
+��
+�
+a jump at N1
+×
+×
+N2−1
+�
+m=0
+exp(−δλ(y + (N1 − 1 − (m + 1/2))δ))
+�
+��
+�
+no jumps until N2
+× δλ exp(−δλ)
+�
+��
+�
+a jump at N2
+×
+N−N1−N2
+�
+ℓ=0
+exp(−δλ(y + (N1 − 1 − (N2 − 1) + (ℓ + 1/2))δ))
+�
+��
+�
+no jumps after N2
+�
+≥
+N1
+�
+N1=N1
+exp(−δλNδ))δ2λ2 exp(−2δλ)
+=
+�L
+δ
+�
+exp(−λNδ))δ2λ2 exp(−2δλ)
+≥ L exp(−δλN))λ2 exp(−2δ0λ)δ
+≥ 2L exp(−δλN))λ2 exp(−2δ0λ)(2R − δ0)
+�
+ν(+1) × 1
+M
+�
+.
+Similarly to above it is now sufficient to note that Nδ ≤ 4δ0 + 4x0 + 2R + 2L.
+Conclusion. To conclude it is sufficient to observe that the conditions above hold for any choice of
+x, y, z ∈ C since N is as in (41).
+□
+Multidimensional case. To extend to the higher dimensional setting, first observe that it is possible
+to apply the same ideas in the proof of Lemma C.2 to each component, in particular requiring that
+the events happen when all components of the process are outside of the rectangle [−x0, +x0]. This
+implies that Assumption 3.5(a) can be used to lower bound the probability of flipping each component
+of the velocity vector. Hence each coordinate can be controlled independently of the others. It is clear
+that the following minorisation condition is implied: let C = [−R, R]d for R > 0, (x, v) ∈ C × {±1}d,
+and let D(x, v) as in (21); then for all (y, w) ∈ (x, v) it holds that
+P(y,w)((XN, VN) ∈ ·) ≥ bd νd(·),
+where N, b are as in Lemma C.2 and νd is the uniform distribution over states in the grid D(x, v) ∩
+(C × {±1}d).
+C.1.2. Drift condition. Let us first characterise in the following Lemma the law of the jump part of
+the process. This result is then used to prove the wanted drift condition in Lemma C.4 below.
+Lemma C.3. Let ˜V x
+t denote the PDMP corresponding to the generator L2 (for this process x acts as
+a parameter). Suppose that λi(x, v) is independent of vj for j ̸= i. Then for any w ∈ {±1}d we have
+Pv( ˜V x
+t = w) =
+d
+�
+i=1
+λi(x, Fiw) + wi
+vi λi(x, v)e−(λi(x,v)+λi(x,Fiv))t
+λi(x, v) + λi(x, Fiv)
+.
+
+48
+SPLITTING SCHEMES FOR PDMPS
+Proof of Lemma C.3. To simplify notation we will suppress the dependence on x and set Λi(v) =
+λi(v) + λi(−v) = λi(x, v) + λi(x, Fiv). Since ˜V x
+t jumps according to λi which does not depend on vj
+we have that the coordinates of ˜V x
+t are all independent. Hence it is sufficient to show
+Pvi(( ˜V x
+t )i = wi) =
+λi(−wi) + wi
+vi λi(vi)e−Λi(vi)t
+Λi(vi)
+.
+Therefore it is sufficient to consider the setting d = 1. Define for any t ≥ 0, v, w ∈ {1, −1}
+ϕt(v; w) := λ(−w) + w
+v λ(v)e−Λ(v)t
+Λ(v)
+.
+If we show that for all t ≥ 0, v, w ∈ {1, −1}
+∂tϕt(v; w) = L2ϕt(v; w),
+ϕ0(v; w) = 1w(v),
+(44)
+then ϕt coincides with the semigroup applied to 1w and we have the desired result
+ϕt(v; w) = Ev[ϕ0( ˜Vt; w)] = Pv[ ˜Vt = w].
+It is straightforward to confirm the initial condition ϕ0(v; w) = 1w(v) holds. So it remains to show
+that ϕt satisfies the PDE (44). Note that
+∂tϕt(v; w) = −w
+v λ(v)e−Λ(v)t
+L2ϕt(v; w) = λ(v) (ϕt(−v; w) − ϕt(v; w))
+= λ(v)
+�
+λ(−w) − w
+v λ(−v)e−Λ(v)t
+Λ(v)
+− λ(−w) + w
+v λ(v)e−Λ(v)t
+Λ(v)
+�
+= −λ(v)
+� w
+v λ(−v)e−Λ(v)t
+Λ(v)
++
+w
+v λ(v)e−Λ(v)t
+Λ(v)
+�
+= −λ(v)w
+v e−Λ(v)t.
+Therefore we have that (44) holds.
+□
+Lemma C.4. Consider the splitting scheme DBD of ZZS. Let λi(x, v) = (vi∂iψ(x))+ + γi(x) and
+let Assumption 3.5 be verified. Then there exists a function V : R × {±1}d → R, and constants
+ρ ∈ (0, 1), C < ∞ such that for all (x, v) and all t ∈ (0, t0) with t0 < (1 + γ0)−1
+P tV (x, v) = P D
+t P B
+2tP D
+t V (x, v) ≤ (1 − ρt)V (x, v) + Ct.
+(45)
+Proof. For a function g(x, v) conditioning on the event v = w and using Lemma C.3 we have
+P tg(x, v) =
+�
+w∈{±1}d
+g(x + vt + wt, w)
+d
+�
+i=1
+�
+λi(x + vt, Fiw) + wi
+vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+�
+.
+We now construct our Lyapunov function V . Let β ∈ (0, 1/2), define φ(s) = 1
+2sign(s) ln (1 + 2|s|) and
+V (x, v) = exp
+�
+βψ(x) +
+d
+�
+i=1
+φ(vi∂iψ(x))
+�
+.
+This is the same Lyapunov function defined in [8]. For this function we have
+P tV (x, v)
+V (x, v)
+=
+�
+w∈{±1}d
+V (x + vt + wt, w)
+V (x, v)
+d
+�
+i=1
+�
+λi(x + vt, Fiw) + wi
+vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+�
+.
+(46)
+
+SPLITTING SCHEMES FOR PDMPS
+49
+By Taylor’s theorem there exists x1 = x1(x, v, w, t) ∈ B(x, t
+√
+d) such that
+ψ(x + vt + wt) = ψ(x) + t⟨v + w, ∇ψ(x)⟩ + t2
+2 (v + w)T ∇2ψ(x1)(v + w).
+Therefore we can rewrite (46) as
+P tV (x, v)
+V (x, v)
+=
+�
+w∈{±1}d
+e
+t2
+2 (v+w)T ∇2ψ(x1)(v+w)
+d
+�
+i=1
+et(vi+wi)β∂iψ(x)+φ(wi∂iψ(x+vt+wt))−φ(vi∂iψ(x))
+×
+�
+λi(x + vt, Fiw) + wi
+vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+�
+.
+(47)
+Since |φ′(s)| ≤ 1 for all s, by Taylor’s Theorem we have
+φ(wi∂iψ(x + vt + wt)) − φ(vi∂iψ(x)) ≤ |wi∂iψ(x + vt + wt) − wi∂iψ(x)| + φ(wi∂iψ(x)) − φ(vi∂iψ(x)).
+Then we can write
+P tV (x, v)
+V (x, v)
+≤
+�
+w∈{±1}d
+K1
+d
+�
+i=1
+I(i)
+(48)
+with
+K1 = e
+t2
+2 (v+w)T ∇2ψ(x1)(v+w)e
+�d
+i=1|wi∂iψ(x+vt+wt)−wi∂iψ(x)|
+I(i) = et(vi+wi)β∂iψ(x)+φ(wi∂iψ(x))−φ(vi∂iψ(x)) λi(x + vt, Fiw) + wi
+vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+.
+Bound outside of a compact set. We split the product in (48) into four cases: (i) wi = vi and
+vi∂iψ(x + vt) > 0; (ii) wi = vi and vi∂iψ(x + vt) < 0; (iii) wi = −vi and vi∂iψ(x + vt) > 0; (iv)
+wi = −vi and vi∂iψ(x + vt) < 0.
+Consider first case (i). Let i be such that wi = vi and vi∂iψ(x + vt) > 0. Then
+I(i) = e2tviβ∂iψ(x) λi(x + vt, Fiv) + λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+.
+Using the form of λi we can write this as
+I(i) = γi(x + vt)e2βtvi∂iψ(x)(1 − e−(|∂iψ(x+vt)|+2γi(x+vt))t)
+|∂iψ(x + vt)| + 2γi(x + vt)
++ e−(|∂iψ(x+vt)|+2γi(x+vt))t+2tviβ∂iψ(x).
+Using that 1 − e−z ≤ z for all z > 0 (note we make use of this inequality several times in the following
+computations) we find
+I(i) ≤ γi(x + vt)e2βtvi∂iψ(x)t + e−(|∂iψ(x+vt)|+2γi(x+vt))t+2tviβ∂iψ(x).
+By Assumption 3.5(b) we can bound γi for |x| ≥ R with R sufficiently large and we have vi∂iψ(x+vt) ≥
+1 + γi(x + vt) which gives
+I(i) ≤ 1 + (vi∂iψ(x + vt) + γi(x + vt))
+|∂iψ(x + vt)| + 2γi(x + vt)
+e−|∂iψ(x+vt)|t+2tviβ∂iψ(x) ≤ 2e−|∂iψ(x+vt)|t+2tviβ∂iψ(x).
+For case (ii), let i be such that wi = vi and vi∂iψ(x + vt) < 0. Then
+I(i) = e2βtvi∂iψ(x) |∂iψ(x + vt)| + γi(x + vt) + γi(x + vt)e−(|∂iψ(x+vt)|+2γi(x+vt))t
+|∂iψ(x + vt)| + 2γi(x + vt)
+≤ e2βtvi∂iψ(x).
+
+50
+SPLITTING SCHEMES FOR PDMPS
+For case (iii), let i be such that wi = −vi and vi∂iψ(x + vt) > 0. Then
+I(i) = eφ(−vi∂iψ(x))−φ(vi∂iψ(x)) λi(x + vt, v) − λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+.
+For s > 0 it holds that φ(−s) − φ(s) = − ln(1 + 2s) and hence
+I(i) = λi(x + vt, v)
+1 + 2vi∂iψ(x)
+1 − e−(λi(x+vt,v)+λi(x+vt,−v))t
+λi(x + vt, v) + λi(x + vt, Fiv) ≤ λi(x + vt, v)
+1 + 2vi∂iψ(x)t.
+For case (iv), let i be such that wi = −vi and vi∂iψ(x + vt) < 0. Then
+I(i) = eφ(−vi∂iψ(x))−φ(vi∂iψ(x)) γi(x + vt) − γi(x + vt)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+.
+For s < 0 we have φ(−s) − φ(s) = ln(1 + 2|s|), and thus we obtain
+I(i) ≤ γi(x + vt)(1 + 2|∂iψ(x)|)t.
+Combining these estimates we have for |x| ≥ R with R sufficiently large
+P tV (x, v)
+V (x, v)
+≤
+�
+w∈{±1}d
+K1
+�
+i:wi=vi, vi∂iψ>0
+(γi(x + vt)e2βtvi∂iψ(x)t + e−(|∂iψ(x+vt)|+2γi(x+vt))t+2tviβ∂iψ(x))
+×
+�
+i:wi=vi, vi∂iψ<0
+e2βtvi∂iψ(x)
+�
+i:wi=−vi, vi∂iψ>0
+λi(x + vt, v)
+1 + 2vi∂iψ(x)t
+�
+i:wi=−vi, vi∂iψ<0
+γi(x + vt)(1 + 2|∂iψ(x)|)t.
+Now consider K1. By Taylor’s theorem there exists x2 ∈ B(x, 2
+√
+dt) such that
+K1 ≤ exp
+� d
+�
+i=1
+�t2
+2 |((v + w)T ∇2ψ(x1))i| + t|(w∇2ψ(x2))i|
+�
+|vi + wi|
+�
+.
+Using this bound and the four cases above we now obtain
+P tV (x, v)
+V (x, v)
+≤
+�
+i:vi∂iψ>0
+e2t2|(vT ∇2ψ(x1))i|+ 2t
+2 |(v∇2ψ(x2))i| (γi(x + vt)e2βtvi∂iψ(x)t + e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t)
+×
+�
+i:vi∂iψ<0
+e
+�
+t2
+2 (2|vT ∇2ψ(x1))i|+2t|(v∇2ψ(x2))i|
+�
++2βtvi∂iψ(x) +
+�
+w∈{±1}d\{v}
+t|{i:wi̸=vi}|
+×
+�
+i:wi=vi,vi∂iψ>0
+e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)(γi(x + vt)e2βtvi∂iψ(x)t + e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t)
+×
+�
+i:wi=vi,vi∂iψ<0
+e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)e2βtvi∂iψ(x)
+�
+i:wi=−vi,vi∂iψ>0
+λi(x + vt, v)
+1 + 2vi∂iψ(x)
+×
+�
+i:wi=−vi,vi∂iψ<0
+e−t0|∂iψ(x+vt)|(1 + 2|∂iψ(x)|).
+By (23) we have
+P tV (x, v)
+V (x, v)
+≤
+�
+i:vi∂iψ>0
+(γ0t + e2t2|(vT ∇2ψ(x1))i|+ 2t
+2 |(v∇2ψ(x2))i|e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t)
+×
+�
+i:vi∂iψ<0
+e
+�
+t2
+2 (2|vT ∇2ψ(x1))i|+2t|(v∇2ψ(x2))i|
+�
++2βtvi∂iψ(x)
+
+SPLITTING SCHEMES FOR PDMPS
+51
++
+�
+w∈{±1}d\{v}
+t|{i:wi̸=vi}|
+�
+i:wi=vi,vi∂iψ>0
+(γ0t + e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t)
+×
+�
+i:wi=vi,vi∂iψ<0
+e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)e2βtvi∂iψ(x)
+�
+i:wi=−vi,vi∂iψ>0
+λi(x + vt, v)
+1 + 2vi∂iψ(x)
+×
+�
+i:wi=−vi,vi∂iψ<0
+e−t0|∂iψ(x+vt)|(1 + 2|∂iψ(x)|).
+Since β < 1/2, by Assumption 3.5(c) there exists β1 such that for |x| ≥ R with R sufficiently large
+P tV (x, v)
+V (x, v)
+≤
+�
+i:vi∂iψ>0
+(γ0t + e−β1|∂iψ(x+vt)|t)
+�
+i:vi∂iψ<0
+e−β1|∂iψ(x)|t
++
+�
+w∈{±1}d\{v}
+t|{i:wi̸=vi}|
+�
+i:wi=vi,vi∂iψ>0
+(γ0t + e−β1|∂iψ(x+vt)|t)
+�
+i:wi=vi,vi∂iψ<0
+e−β1t|∂iψ(x)|
+×
+�
+i:wi=−vi,vi∂iψ>0
+λi(x + vt, v)
+1 + 2vi∂iψ(x)
+�
+i:wi=−vi,vi∂iψ<0
+e−t0|∂iψ(x+vt)|(1 + 2|∂iψ(x)|).
+For |x| ≥ R with R sufficiently large λi(x + vt, v)/(1 + 2vi∂iψ(x)) ≤ 1 and by (23) we have γi(x)(1 +
+2|∂iψ(x)|) ≤ 1. We also have that |∇ψ(x + vt)| ≥ M for any M > 0 for |x| ≥ R with R sufficiently
+large. Then we have
+P tV (x, v)
+V (x, v)
+≤ (γ0t + e−β1Mt)|{i:vi∂iψ(x+vt)>0}|e−β1Mt|{i:vi∂iψ(x+vt)<0}|
++
+�
+w∈{±1}d\{v}
+t|{i:wi̸=vi}|
+�
+i:wi=vi,vi∂iψ>0
+(γ0t + e−β1Mt)
+�
+i:wi=vi,vi∂iψ<0
+e−β1tM.
+Since e−β1Mt ≤ γ0t + e−β1Mt we obtain
+P tV (x, v)
+V (x, v)
+≤ (γ0t + e−β1Mt)|{i:vi∂iψ(x+vt)>0}|(γ0t + e−β1Mt)|{i:vi∂iψ(x+vt)<0}|
++
+�
+w∈{±1}d\{v}
+t|{i:wi̸=vi}|(γ0t + e−β1Mt)|{i:wi=vi,vi∂iψ(x+vt)>0}|(γ0t + e−β1Mt)|{i:wi=vi,vi∂iψ(x+vt)<0}|.
+Hence
+P tV (x, v)
+V (x, v)
+≤ (γ0t + e−β1Mt)d +
+�
+w∈{±1}d: w̸=v
+t|{i:wi̸=vi}|(γ0t + e−β1Mt)d−|{i:wi̸=vi}|
+=
+�
+w∈{±1}d
+t|{i:wi̸=vi}|(γ0t + e−β1Mt)d−|{i:wi̸=vi}|
+=
+d
+�
+k=0
+�d
+k
+�
+tk (γ0t + e−β1Mt)d−k
+=
+�
+(1 + γ0)t + e−β1Mt�d
+.
+To show that (45) holds for |x| ≥ R it is sufficient to show that (1 + γ0)t + e−β1Mt < 1 − ρt for some
+ρ > 0. Indeed in that case 1 − ρt < 1 and thus
+((1 + γ0)t + e−β1Mt)d < (1 − ρt)d < 1 − ρt.
+
+52
+SPLITTING SCHEMES FOR PDMPS
+Note that for t ≤ t0, with t0 ∈ [0, 1], it holds that e−β1Mt ≤ 1 − ct for c = 1−e−β1Mt0
+t0
+. Then for t ≤ t0
+we have
+(1 + γ0)t + e−β1Mt ≤ 1 − t(c − 1 − γ0).
+Then it is needed that c > 1 + γ0, that is t0 should be such that
+1 − e−β1Mt0
+t0
+> 1 + γ0.
+(49)
+Note we can always increase M by taking R larger. Choose M such that e−β1Mt0 < 1 − t0(1 + γ0),
+which is possible since t0 < (1 + γ0)−1, then (49) holds. Hence (45) holds for |x| ≥ R with ρ =
+(1 − e−β1Mt0)t−1
+0
+− 1 − γ0.
+Bound inside of a compact set. It remains to show that (45) holds for |x| ≤ R. Let C = {x :
+|x| ≤ R} × {±1}d. Recall t < 1 and ψ ∈ C2. We shall use the inequality etr ⩽ 1 + tr + t2r2er/2 ≤
+1 + t(r + e3r/2), which holds for for t ⩽ 1, r > 0.
+First of all we consider the term in the sum
+corresponding to the case w = v. Bounding the probability of this event by 1 we find
+K1
+d
+�
+i=1
+I(i) ≤ et22vT ∇2ψ(x1)v+ 2
+2
+�d
+i=1|∂iψ(x+2vt)−∂iψ(x)|+2tβ⟨v,∇ψ(x)⟩
+≤ 1 + t(A(x, v, t) + e3A(x,v,t)/2),
+where A(x, v, t) = 2vT ∇2ψ(x1)v + (2/2) �d
+i=1|⟨v, ∇∂iψ(x2)⟩| + 2β⟨v, ∇ψ(x)⟩. Taking the maximum of
+A over (x, v, t) ∈ C × {±1}d × (0, 1) we find
+K1
+d
+�
+i=1
+I(i) ≤ 1 + tA.
+Let us now consider the remaining elements in the sum. Here we take advantage that a velocity
+flip is an order t event. Consider for the moment only the i-th component of the velocity vector. The
+probability that this is flipped (i.e. wi = −vi) satisfies
+λi(x + vt, Fiw) − λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t
+λi(x + vt, v) + λi(x + vt, Fiv)
+≤ λi(x + vt, v)(1 − e−t(|∂iψ(x+tv)|+2γi(x+tv)))
+|∂iψ(x + tv)| + 2γi(x + tv)
+≤ tλi(x + vt, v)
+≤ t
+max
+i=1,...,d, (x,v)∈C×{±1}d, t∈(0,1) λi(x + vt, v).
+Here we used that 1 − exp(−z) ≤ z for z ≥ 0. All other probabilities can be bounded by 1 and hence
+�
+w̸=v
+K1
+d
+�
+i=1
+I(i) ≤ t
+max
+i=1,...,d, (x,v)∈C×{±1}d, t∈(0,1) λi(x + vt, v)
+�
+w̸=v
+V (x + vt + wt, w)
+V (x, v)
+.
+Since V is continuous under our assumptions we proved that for every compact set C ×{±1}d there
+exists a constant B > 0 such that for all (x, v) ∈ C × {±1}d it holds
+P tV (x, v) ≤ (1 + tB)V (x, v).
+(50)
+Therefore we have (45) holds for all x ∈ Rd, v ∈ {±1}d.
+□
+
+SPLITTING SCHEMES FOR PDMPS
+53
+C.1.3. Proof of Equation (24). Let us prove (24). Fix a probability measure µ on Rd × {±1}d, and let
+(x, v) be a point in the support of µ. Then we can construct the set D(x, v) corresponding to (x, v)
+and given by (21). By Theorem 3.6 there is a unique invariant measure πx,v
+δ
+for the Markov process
+with kernel P 2
+δ , and by (20) for any probability measures ν, ν′ on D(x, v)
+∥νP 2n
+δ
+− ν′P 2n
+δ ∥V ≤ C
+α ˜κnδ ∥ν − ν′∥V .
+Setting ν = δx,v, ν′ = πx,v
+δ
+and using that πx,v
+δ
+is an invariant measure for the kernel P 2
+δ we have
+∥δx,vP 2n
+δ
+− πx,v
+δ ∥V ≤ C
+α ˜κnδ ∥ν − ν′∥V .
+Then integrating with respect to the probability measure µ we obtain
+∥µP 2n
+δ
+− µπx,v
+δ ∥V = sup
+|g|≤V
+����
+�
+[P 2n
+δ g(x, v) − πx,v
+δ (g)]µ(dx, dv)
+����
+≤
+�
+sup
+|g|≤V
+��P 2n
+δ g(x, v) − πx,v
+δ (g)
+�� µ(dx, dv)
+≤
+� ��δx,vP 2n
+δ
+− πx,v
+δ
+��
+V µ(dx, dv)
+≤ C
+α ˜κnδ
+�
+∥δ(x,v) − πx,v
+δ ∥V µ(dx, dv).
+C.2. Other splitting schemes. In this Section we consider splitting schemes DRBRD and RDBDR
+of ZZS and prove geometric ergodicity in Theorem C.6. The minorisation and drift conditions are
+proved in Sections C.2.1 and C.2.2 respectively. We shall work under the following assumption.
+Assumption C.5. There exists γ0 ∈ (0, ∞) such that the following conditions for the refreshment
+rate hold:
+(a) there exists R > 0 for which for any |x| ≥ R it holds that
+γ(x)
+d
+�
+j=1
+(1 + |∂jψ(x)|) ≤ γ0.
+(b) For |x| > R for some R > 0 it holds that
+sup
+t∈(0,1), y1,y2∈B(x,t
+√
+d),v,w∈{−1,1}d
+γ(x + vt)et|∇ψ(x)|+ t2
+2 (v+w)T ∇2ψ(y1)(v+w)+|∇ψ(y2)|
+d
+�
+i=1
+(1 + 2|∂iψ(x)|) ≤ γ0.
+Theorem C.6. Consider splitting schemes DRBRD and RDBDR of ZZS. Suppose Assumption
+3.5(a), (c) holds for switching rates λi(x, v) = (vi∂iψ(x))+. Suppose moreover that the refreshment
+rate γ satisfies Assumption C.5(a) for scheme RDBDR and Assumption C.5(b) for scheme DRBRD.
+Then statements (1) and (2), as well as Equation (24), hold.
+In particular (2) holds with δ0 <
+2(1 + 2γ0 + γ2
+0)−1 for RDBDR and with δ0 < 2(1 + 2γ0)−1 for DRBRD.
+C.2.1. Minorisation condition.
+Splitting DRBRD. The chain obtained by DRBRD has the same periodic behaviour of DBD. Hence
+this case can be treated in the same way and a minorisation condition follows by the same reasoning
+used in Section C.1.1 for splitting DBD.
+
+54
+SPLITTING SCHEMES FOR PDMPS
+Splitting RDBDR. In this case we give a sketch of the proof. The chain obtained by RDBDR breaks
+the grid behaviour exhibited by DBD because of the two refreshment steps at the beginning and end
+of each step. Indeed, consider the one-dimensional case and recall the definition of the grid D(x, v)
+as in Lemma C.2. Since v = ±1, there are two disjoint grids: D(x, +1) and D(x, −1), with the idea
+being that after even steps of DBD the process lives on the same grid it started from. However, the
+process RDBDR can swap between one grid and the other by having a velocity refreshment. Indeed,
+starting the process at (x, +1) and having a velocity flip due to a refreshment at the end of the first
+step and having no other jumps, we find the state of the process is (X2, V2) = (x, −1). Therefore after
+even steps this process lives on the grid D(x) = {y : y = x + mδ, m ∈ Z}. If the initial and final
+condition are on the same grid D(x, v), then no refreshment is required and one can simply use the
+proof of the scheme DBD. On the other hand, if the two states are on different grids, i.e. one is on
+D(x, +1) and the other on D(x, −1), then a refreshment is required to choose the right grid.
+In order to maintain the right dependence on the step size δ it is required to give the process
+additional ⌈ M
+δ ⌉ iterations, for a constant M > 0.
+Indeed with this modification the probability
+of having a refreshment in the first ⌈ M
+δ ⌉ is constant has a lower bound which is independent of δ,
+assuming δ ≤ δ0 for some δ0 > 0 (see for instance the second case in the proof of Lemma C.2). After
+the first ⌈ M
+δ ⌉ iterations the process is on the right grid and Lemma C.2 can be applied with the further
+constraint that no (more) refreshments take place. Note that this event again has a lower bounded
+probability independent of δ. Since in the first ⌈ M
+δ ⌉ iterations the process can go out of the initial
+compact set C = [−R, R], it follows that the Lemma should be applied with set ˜C = [−R−M, R+M].
+The extension to the multidimensional case follows by applying this same intuition to every com-
+ponent.
+C.2.2. Drift condition. Let us start with an auxiliary result.
+Lemma C.7. Suppose the refreshment rate γ satisfies Assumption C.5(a). Then P R
+t V ≤ (1+γ0t)V +
+Mt, where γ0 is as in Assumption C.5 and M independent of t.
+Proof. Let V be as in Lemma C.4. Applying the transition kernel P R
+t
+to V we find
+P R
+t V (x, v) = V (x, v)e−tγ(x) + 1
+2d (1 − e−tγ(x))
+�
+w̸=v
+V (x, w)
+= V (x, v)e−tγ(x) + (1 − e−tγ(x))V (x, v) 1
+2d
+�
+w̸=v
+V (x, w)
+V (x, v)
+= V (x, v)
+�
+e−tγ(x) + (1 − e−tγ(x)) 1
+2d
+�
+w̸=v
+�
+j:vj̸=wj
+(1 + |∂jψ(x)|)
+�
+≤ V (x, v)
+�
+�e−tγ(x) + tγ(x)
+d
+�
+j=1
+(1 + |∂jψ(x)|)
+�
+� .
+Clearly for x inside of a compact set this implies P R
+t V (x, v) ≤ (1 + Bt)V (x, v) by taking maximum
+over x. Outside of a compact set we use Assumption C.5 to obtain
+P R
+t V (x, v) ≤ V (x, v)(1 + tγ0).
+□
+Lemma C.8. Consider the splitting scheme RDBDR of ZZS. Suppose Assumptions 3.5(c) and C.5(a)
+hold. Then there exist a function V and constants ρ ∈ (0, 1), C > 0 such that for any t ∈ (0, t0) with
+t0 < (1 + 2γ0 + γ2
+0)−1 it holds that
+P R
+t P D
+t P B
+2tP D
+t P R
+t V (x, v) ≤ (1 − ρt)V (x, v) + Ct.
+
+SPLITTING SCHEMES FOR PDMPS
+55
+Proof. Let V be as in Lemma C.4. In the current context the result of the Lemma is that for all
+t ∈ (0, t0) with t0 < 1 it holds that P D
+t P B
+2tP D
+t V (x, v) ≤ (1−ρt)V (x, v)+Bt where ρ = (1−e−Rt0)t−1
+0 −1
+for R sufficiently large such that ρ > 0. Applying Lemmas C.4 and C.7 we obtain
+P R
+t P D
+t P B
+2tP D
+t P R
+t V (x, v) ≤ (1 + tγ0)P R
+t P D
+t P B
+2tP D
+t V (x, v) + Mt
+≤ (1 + tγ0)(1 − ρt)P R
+t V (x, v) + t(M + (1 + γ0)B)
+≤ (1 + tγ0)2(1 − ρt)V (x, v) + t(M(2 + γ0) + (1 + γ0)B).
+It is left to ensure that (1 + tγ0)2(1 − ρt) ≤ (1 − ˜ρt) for ˜ρ > 0. We have
+(1 + tγ0)2(1 − ρt) ≤ (1 − t(ρ − 2γ0 − γ2
+0)).
+Hence it is needed that
+(1 − e−Rt0)
+t0
+− 1 > 2γ0 + γ2
+0
+and thus that e−Rt0 < 1 − t0(1 + 2γ0 + γ2
+0), which is valid as R can be taken as large as needed and
+t0 < (1 + 2γ0 + γ2
+0)−1.
+□
+Lemma C.9. Consider the splitting scheme DRBRD of ZZS. Suppose Assumptions 3.5(c) and C.5
+hold. Then there exist a function V and constants ρ ∈ (0, 1), C > 0 such that for any t ∈ (0, t0) with
+t0 < (1 + 2γ0)−1 it holds that
+P D
+t P R
+t P B
+2tP R
+t P D
+t V (x, v) ≤ (1 − ρt)V (x, v) + Ct.
+Proof. Let V be as in Lemma C.4. Observe that by Lemma C.7 we have that
+P R
+t P D
+t V (x, v) = P R
+t V (x + vt, v) ≤ (1 + γ0t)V (x + vt, v) + Mt
+and thus P R
+t P D
+t V (x, v) ≤ (1 + γ0t)P D
+t V (x, v) + Mt. Then
+P D
+t P R
+t P B
+2tP R
+t P D
+t V (x, v) ≤ (1 + γ0t)P D
+t P R
+t P B
+2tP D
+t V (x, v) + Mt
+and
+P D
+t P R
+t P B
+2tP D
+t V (x, v) = e−tγ(x+vt)P D
+t P B
+2tP D
+t V (x, v)
++ V (x, v)(1 − e−tγ(x+vt))
+�
+w∈{±1}d
+1
+2d
+V (x + vt + wt, w)
+V (x, v)
+.
+The first term corresponds to the case of no refreshments, while in the second term a refreshment
+takes place. For the first term we can directly apply Lemma C.4, which in the current context shows
+that for t < t0 < 1 it holds P D
+t P B
+2tP D
+t V (x, v) ≤ (1 − ρt)V (x, v) + Mt for ρ = (1 − e−β1Mt0)t−1
+0
+− 1.
+The second term can be rewritten as in (47), that is for x1 = x1(x, v, w, t) ∈ B(x, t
+√
+d)
+V (x + vt + wt, w)
+V (x, v)
+= exp
+�
+β(ψ(x + vt + wt) − ψ(x)) +
+d
+�
+i=1
+(φ(wi∂iψ(x + vt + wt)) − φ(vi∂iψ(x)))
+�
+= et|∇ψ(x)|+ t2
+2 (v+w)T ∇2ψ(x1)(v+w)+|∇ψ(x2)|
+d
+�
+i=1
+(1 + 2|∂iψ(x)|)
+Using Assumption C.5(b) we find
+(1 − e−tγ(x+vt))
+�
+w∈{±1}d
+1
+2d
+V (x + vt + wt, w)
+V (x, v)
+≤ tγ(x + vt)
+�
+w∈{±1}d
+1
+2d
+V (x + vt + wt, w)
+V (x, v)
+≤ tγ0.
+
+56
+SPLITTING SCHEMES FOR PDMPS
+Therefore we have shown
+P D
+t P R
+t P B
+2tP R
+t P D
+t V ≤ (1 + γ0t)(1 − ρt + tγ0)V + ˜
+Mt ≤ (1 − t(ρ − 2γ0))V + ˜
+Mt.
+Hence it is sufficient to ensure that ρ > 2γ0, which can be done similarly to the proof of Lemma
+C.8.
+□
+Appendix D. Proof of Proposition 4.5 and related results
+In this section we collect statements and proofs that are not included in Section 4.
+D.1. Proof of Proposition 4.5. In this section we prove Proposition 4.5. We start by focusing on
+the left hand side of (26), i.e. L∗
+BPS(µf2). We find since µ is rotationally invariant in v
+L∗
+BPS(µf2)(x, v) = µ(x, v)
+�
+⟨v, ∇ψ(x)⟩f2(x, v) − ⟨v, ∇xf2(x, v)⟩ + (−⟨v, ∇ψ(x)⟩)+f2(x, R(x)v)
+− ⟨v, ∇ψ(x)⟩+f2(x, v) + λr
+�
+f2(x, y)ν(dy) − λrf2(x, v)
+�
+.
+We shall consider the case of v = ±1, hence ν = (1/2)δ+1 +(1/2)δ−1. In particular this choice satisfies
+Assumption 4.2 below. Introduce the notation f+
+2 (x) = f2(x, 1), f−
+2 (x) = f2(x, −1). We have in the
+1-dimensional setting
+L∗
+BPS(µf2)(x, +1) = −µ(x, +1)
+�
+(f+
+2 )′(x) + ((−ψ′(x))+ + λr/2)f+
+2 (x) − (λr/2 + (−ψ′(x))+)f−
+2 (x)
+�
+,
+L∗
+BPS(µf2)(x, −1) = +µ(x, −1)
+�
+(f−
+2 )′(x) + ((+ψ′(x))+ + λr/2)f+
+2 (x) − (λr/2 + (+ψ′(x))+)f−
+2 (x)
+�
+.
+Define function h such that hµ = L∗
+2µ, and also h+(x) = h(x, +1) and h−(x) = h(x, −1). Therefore
+we wish to solve the following system of ODEs
+�
+(f+
+2 )′(x) = −(λr/2 + (−ψ′(x))+)f+
+2 (x) + (λr/2 + (−ψ′(x))+)f−
+2 (x) − h+(x),
+(f−
+2 )′(x) = −(λr/2 + (+ψ′(x))+)f+
+2 (x) + (λr/2 + (+ψ′(x))+)f−
+2 (x) + h−(x),
+(51)
+with compatibility condition (27), which in this case can be written as
+� ∞
+−∞
+(f+
+2 (x) + f−
+2 (x))π(x)dx = 0
+(52)
+with π(x) = µ(x, 1) + µ(x, −1). Let us find a solution to (51) for a generic (continuous and locally
+lipschitz) function h. Start by subtracting the first line to the second line in (51):
+(f−
+2 )′(x) − (f+
+2 )′(x) = ((ψ′(x))+ − (−ψ′(x))+)(f−
+2 (x) − f+
+2 (x)) + hs(x),
+(53)
+where hs(x) = h+(x) + h−(x). Define g = f−
+2 − f+
+2 and notice that (ψ′(x))+ − (−ψ′(x))+ = ψ′(x).
+Then we can rewrite (53) as
+g′(x) = ψ′(x)g(x) + hs(x).
+Solving this ODE using an integrating factor we find
+g(x) = exp (ψ(x)) lim
+y→−∞ [exp (−ψ(y)) g(y)] + exp (ψ(x))
+� x
+−∞
+hs(y) exp(−ψ(y))dy.
+Recall that g = f− − f+ and f+, f− satisfy (52). In order for f2 to define a proper density we require
+� ∞
+−∞
+g(x)π(x)dx < ∞.
+For this to hold it must be that limy→−∞ exp (−ψ(y)) g(y) = 0 and thus
+g(x) = exp (ψ(x))
+� x
+−∞
+hs(y) exp(−ψ(y))dy.
+(54)
+
+SPLITTING SCHEMES FOR PDMPS
+57
+Since f−
+2 (x) = f+
+2 (x) + g(x) and plugging this in the first equation of (51) we obtain the ODE
+(f+
+2 )′(x) = (λr/2 + (−ψ′(x))+)g(x) − h+(x)
+which can be integrated as
+f+
+2 (x) = f+
+2 (0) +
+� x
+0
+�
+(λr/2 + (−ψ′(y))+)g(y) − h+(y)
+�
+dy.
+(55)
+It follows that
+f−
+2 (x) = f+
+2 (0) +
+� x
+0
+�
+(λr/2 + (−ψ′(y))+)g(y) − h+(y)
+�
+dy
++ exp (ψ(x))
+� x
+−∞
+hs(y) exp(−ψ(y))dy.
+(56)
+Finally we compute f+
+2 (0) enforcing the compatibility condition (52). Plugging (55) and (56) in (52)
+we find the condition
+f+
+2 (0) = −
+� ∞
+−∞
+�
+g(x)/2 +
+� x
+0
+�
+(λr/2 + (−ψ′(y))+)g(y) − h+(y)
+�
+dy
+�
+π(x)dx.
+(57)
+D.2. Proof of Proposition 4.7. Fix x ∈ R, δ > 0 and let G(x, δ) := {(z, v) ∈ R×{±1} : (z −x)/δ ∈
+Z} be the state space of the chain with initial position x. For now, let µδ be any probability measure
+on G(x, δ) such that µδ(y, w) = µδ(y, −w) for all (y, w) ∈ G(x, δ), and let us give a sufficient and
+necessary condition for it to be invariant by the chain. Since such a µδ is invariant by the refreshment
+step, it is invariant for the scheme RDBDR if and only if it is invariant for the scheme R’DBD,
+where R’ is a deterministic flip of the velocity (which, as R, preserves µδ). Besides, from a state
+(y, w) ∈ G(x, δ), one transition of R’DBD can only lead to (y, w) or (y + δw, −w), from which it can
+only stay or come back to the initial (y, w). In other words the pair {(y, w), (y+δw, −w)} is irreducible
+for this chain, and thus µδ is invariant for R’DBD if and only if its restrictions on all these sets for
+(y, w) ∈ G(x, δ) are invariant by this scheme, which by definition reads
+∀(y, w) ∈ G(x, δ) ,
+µδ(y, w)e−δλ(y+wδ/2,w) = µδ(y + δw, −w)e−δλ(y+wδ/2,−w).
+It turns out that this is exactly the skew detailed balance condition (6) for the scheme DBD. Writing
+that µδ(y, w) ∝ exp(−ψδ(y)) for some ψδ and recalling that λ(y, w) − λ(y, −w) = ψ′(y) for all y, w,
+this is equivalent to
+∀(y, w) ∈ G(x, δ),
+ψδ(y + δw) − ψδ(y) = δψ′ (y + δw/2) .
+Up to an additive constant, the only function ψδ which satisfies this is the one given in the statement
+of Proposition 4.7. As a conclusion, we have proven that a probability measure on G(x, δ) which is
+independent from the velocity is invariant for the scheme RDBDR if and only if it is the one given
+in the proposition, which concludes the proof of the first statement.
+Now we focus on the convergence of empirical means, assuming that the conditions of Theorem 3.6
+are met. The reference position x ∈ R is still fixed. The long-time convergence established in Theo-
+rem 3.6 (for P 2
+δ where Pδ is one step of the scheme) is well-known to imply an ergodic Law of Large
+Numbers. In particular, for all initial conditions in G(x, δ) and all bounded f, distinguishing odd and
+even indexes, we see that
+1
+N
+�N
+k=1 f(Ztk) (where (Ztk)k∈N is a trajectory of the scheme) converges
+almost surely as N → ∞ to ˜µδ(f) := (µ′
+δ(f) + µ′′
+δ(f))/2, where µ′
+δ and µ′′
+δ are the unique invariant
+measures of P 2
+δ on each periodic component of the state space. In particular, ˜µδ is an invariant measure
+for Pδ. In dimension 1, the scheme DBD is such that for all y, for all times, the number of visits of
+the points (y, 1) and (y, −1) differ at most by 1, which implies by ergodicity that ˜µδ(y, w) = µδ(y, −w)
+for all (y, w) ∈ G(x, δ), and we conclude thanks to the first part of the proof.
+
+58
+SPLITTING SCHEMES FOR PDMPS
+(a) Refreshment rate λr = 1.0.
+(b) Refreshment rate λr = 3.0.
+Figure 11. Plots of the theoretical invariant measure up to second order for a standard
+Gaussian target, as given by Proposition D.1. The step size is δ = 0.5.
+D.3. Application of Proposition 4.5 to three one-dimensional targets. In this section we give
+the function f2 corresponding to the three cases considered in Figure 3.
+D.3.1. Gaussian target. Let us start with a one-dimensional Gaussian target with mean zero and
+variance σ2 > 0.
+Proposition D.1. Let ψ(x) = x2/(2σ2) for σ2 > 0. Then:
+• For the splitting scheme DBRBD it holds that
+f2(x, +1) = f2(x, −1) = λr
+24
+�
+2
+√
+2
+σ√π − x3
+σ4 sign(x)
+�
+.
+• For the splitting scheme BDRDB it holds that
+f2(x, +1) =
+1
+8σ2 −
+1
+4σ4 x21x<0,
+f2(x, −1) =
+1
+8σ2 −
+1
+4σ4 x21x>0.
+• For the splitting scheme RDBDR it holds that
+f2(x, +1) = f2(x, −1) = 0.
+• For the splitting scheme DRBRD it holds that
+f2(x, +1) = f2(x, −1) = λr
+12
+�
+2
+√
+2
+σ√π − |x|3
+σ4
+�
++ λ2
+r
+16
+�
+1 − x2
+σ2
+�
+.
+Proof of Proposition D.1. Recalling that v ∈ {−1, +1}, for all splitting schemes we can compute the
+functions h = (L∗
+2µ)/µ:
+hDBRBD(x, v) = λr
+8σ4 (x2 + 2vx(−vx)+),
+hBDRDB(x, v) =
+1
+8σ6
+�
+−λrσ2(x2 + 2vx(−vx)+) + 2(−vx)+(x2 − 2σ2)
+�
+,
+hRDBDR(x, v) = 0,
+
+SPLITTING SCHEMES FOR PDMPS
+59
+hDRBRD(x, v) =
+1
+8σ4 λr
+�
+x2 + vx(3(−vx)+ + (vx)+) + λrvxσ2�
+.
+Splitting DBRBD. Observe that hs(x) =
+λr
+4σ4 (x2 + x((−x)+ − (x)+)) = 0. Hence by (54) it holds
+g(x) = 0. Then by (55)
+f+
+2,DBRBD(x) = f+
+2 (0) −
+� x
+0
+λr
+8σ4 (y2 + 2y(−y)+)dy
+= f+
+2 (0) −
+λr
+24σ4 x3(1 − 21x<0).
+Since g = 0 we have f+
+2,DBRBD = f−
+2,DBRBD. To find f+
+2 (0) we enforce (52):
+� +∞
+−∞
+f+
+2 (x)π(x)dx = f+
+2 (0) −
+λr
+24σ4
+� +∞
+−∞
+x3(1 − 21x<0)π(x)dx
+= f+
+2 (0) −
+λr
+12σ4
+� +∞
+0
+x3π(x)dx
+= f+
+2 (0) −
+λr
+12σ4 σ3
+�
+2
+π = 0.
+Clearly this is satisfied for f+
+2 (0) = λr/(6σ
+√
+2π).
+Splitting RDBDR. Clearly in this case hs(x) = 0, hence g(x) = 0 and f+
+2,RDBDR = f−
+2,RDBDR = 0.
+Splitting BDRDB. We have hs(x) =
+1
+4σ6 |x|(x2 − 2σ2). Now inserting this into the expression for g
+we get
+g(x) =
+1
+4σ6 exp(x2/(2σ2))
+� x
+−∞
+|y|(y2 − 2σ2) exp(−x2/(2σ2))dy
+=
+1
+4σ6 exp(x2/(2σ2))(−σ2sign(x) exp(−x2/(2σ2))x2)
+= − 1
+4σ4 x2sign(x).
+We compute f+
+2 by applying (55). First observe that
+� x
+0
+h+(y)dy = − λr
+8σ4
+� x
+0
+y2sign(y)dy +
+1
+4σ6
+� x
+0
+(−y)+(y2 − 2σ2)dy
+and
+� x
+0
+(λr/2 + (−y/σ2)+)g(y)dy = − λr
+8σ4
+� x
+0
+y2sign(y)dy −
+1
+4σ6
+� x
+0
+(−y)+y2sign(y)dy.
+Therefore we obtain
+f+
+2,BDRDB(x) = f+
+2 (0) +
+1
+2σ4
+� x
+0
+(−y)+dy
+= f+
+2 (0) +
+1
+4σ4 x21x<0.
+Enforcing the compatibility condition (57) we obtain f+
+2 (0) = 1/(8σ2).
+
+60
+SPLITTING SCHEMES FOR PDMPS
+Splitting DRBRD. Similarly to the case of DBRBD observe that hs = 0 and thus g(x) = 0. Observe
+that h+
+DRBRD(x) =
+λr
+8σ4 (2x2sign(x) + λrxσ2). Then by (55)
+f+
+2,DRBRD(x) = f+
+2 (0) − λr
+8σ4
+� x
+0
+(2y2sign(y) + σ2λry)dy
+= f+
+2 (0) −
+λr
+12σ4 x3sign(x) −
+λ2
+r
+16σ2 x2.
+To find f+
+2 (0) we enforce (57):
+f+
+2 (0) = λr
+6σ
+�
+2
+π + λ2
+r
+16.
+□
+D.3.2. Non-Lipschitz potential. Now we focus on a target distribution with non-Lipschitz potential.
+Proposition D.2. Let ψ(x) = x4. Then:
+• For the splitting scheme DBRBD it holds that
+f2(x, +1) = f2(x, −1) = λr
+7
+�
+1
+2Γ(5/4) − 2x7sign(x)
+�
++ 1
+2
+�Γ(3/4)
+Γ(1/4) − x2
+�
+.
+• For the splitting scheme BDRDB it holds that
+f2(x, +1) = 5Γ(3/4)
+2Γ(1/4) − x2 − 4x61x<0 ,
+f2(x, −1) = 5Γ(3/4)
+2Γ(1/4) − x2 − 4x61x≥0 .
+• For the splitting scheme RDBDR it holds that
+f2(x, +1) = f2(x, −1) = Γ(3/4)
+2Γ(1/4) − 1
+2x2.
+• For the splitting scheme DRBRD it holds that
+f2(x, +1) = f2(x, −1) = λr
+7
+�
+1
+Γ(5/4) − 4x7sign(x)
+�
++ 1
+2
+�Γ(3/4)
+Γ(1/4) − x2
+�
++ λ2
+r
+8
+�1
+4 − x4
+�
+.
+Proof of Proposition D.2. By Proposition 4.3 we obtain
+hDBRBD(x, v) = +2λr(x6 + 2vx3(−vx3)+) + vx,
+hBDRDB(x, v) = −2λr(x6 + 2vx3(−vx3)+) + 8(−vx3)+(−3x2 + 2x6) − 2vx,
+hRDBDR(x, v) = vx,
+hDRBRD(x, v) = +2λr(x6 + vx3(3(−vx3)+ + (vx3)+)) + vx + (λ2
+rvx3)/2.
+Denote the normalisation constant of the target π(x) by Z = 2Γ(5/4).
+Splitting DBRBD. Since hs(x) = 0 we have
+f+
+2,DBRBD(x) = f−
+2,DBRBD(x) = f+
+2 (0) − 2
+7λrx7sign(x) − 1
+2x2,
+with
+f+
+2 (0) = 4λr
+7
+� ∞
+0
+x7π(x)dx +
+� ∞
+0
+x2π(x)dx =
+λr
+14Γ(5/4) + Γ(3/4)
+2Γ(1/4).
+
+SPLITTING SCHEMES FOR PDMPS
+61
+Splitting BDRDB. In this case hs(x) = 8x2|x3|(2x4 − 3) and thus we find g(x) = −4x6sign(x). It
+follows that
+f+
+2,BDRDB(x) = f+
+2 (0) − 4x61x<0 − x2,
+f−
+2,BDRDB(x) = f+
+2 (0) − 4x61x≥0 − x2,
+where
+f+
+2 (0) = Γ(7/4)
+2Γ(5/4) + Γ(3/4)
+Γ(1/4).
+Splitting RDBDR. Since hs(x) = 0 we have f+
+2,RDBDR(x) = f−
+2,RDBDR(x) = f+
+2 (0) − x2/2 with
+f+
+2 (0) = Γ(3/4)
+2Γ(1/4).
+Splitting DRBRD. Since hs(x) = 0 we have
+f+
+2,DRBRD(x) = f−
+2,DRBRD(x) = f+
+2 (0) − 4
+7λrx7sign(x) − 1
+2x2 − 1
+8λ2
+rx4,
+with
+f+
+2 (0) =
+λr
+7Γ(5/4) + Γ(3/4)
+2Γ(1/4) + 1
+32λ2
+r.
+□
+D.3.3. Heavy tailed target. Finally we consider a Cauchy distribution π(x) = γ/(π(γ2 +x2)) for γ > 0.
+Proposition D.3. Let ψ(x) = ln(γ2 + x2). Then:
+• For the splitting scheme DBRBD it holds that
+f2(x, +1) = f2(x, −1) = λr
+4γ
+�π
+4 − |arctan(x/γ)| +
+γ|x|
+γ2 + x2 − 1
+π
+�
++ 1
+12
+� 1
+4γ2 +
+x2 − γ2
+(γ2 + x2)2
+�
+.
+• For the splitting scheme BDRDB it holds that
+f2(x, v) =
+�
+(x2 − 3γ2)2
+48γ2(x2 + γ2)2
+�
+1xv<0 +
+�x4 − 54x2γ2 + 9γ4
+48γ2(x2 + γ2)2
+�
+1xv≥0.
+• For the splitting scheme RDBDR it holds that
+f2(x, +1) = f2(x, −1) = 1
+12
+� 1
+4γ2 +
+x2 − γ2
+(γ2 + x2)2
+�
+.
+• For the splitting scheme DRBRD it holds that
+f2(x, +1) = f2(x, −1) = λr
+2γ
+�π
+4 − |arctan(x/γ)| +
+γ|x|
+γ2 + x2 − 1
+π
+�
++ 1
+12
+� 1
+4γ2 +
+x2 − γ2
+(γ2 + x2)2
+�
++ λ2
+r
+8
+�
+ln 4 − ln
+�
+1 + x2
+γ2
+��
+.
+Proof of Proposition D.3. By Proposition 4.3 we obtain
+hDBRBD(x, v) =
+λr
+2(γ2 + x2)2 (x2 + 2vx(−vx)+) + 1
+24vψ(3)(x),
+hBDRDB(x, v) = −
+λr
+2(γ2 + x2)2 (x2 + 2vx(−vx)+) +
+2
+(γ2 + x2)3 (−vx)+(−γ2 + 2x2) − 1
+12vψ(3)(x),
+hRDBDR(x, v) = 1
+24vψ(3)(x),
+
+62
+SPLITTING SCHEMES FOR PDMPS
+hDRBRD(x, v) =
+λr
+2(γ2 + x2)2
+�
+x2 + vx(3(−vx)+ + (vx)+)
+�
++ 1
+24vψ(3)(x) + λ2
+r
+xv
+4(γ2 + x2).
+Denote the normalisation constant of the target π(x) by Z = π/γ.
+Splitting DBRBD. Since hs(x) = 0 we have
+f+
+2,DBRBD(x) = f−
+2,DBRBD(x) = f+
+2 (0) − λr
+4 sign(x)
+�arctan(x/γ)
+γ
+−
+x
+γ2 + x2
+�
+− 1
+12
+� γ2 − x2
+(γ2 + x2)2 − 1
+γ2
+�
+,
+with f+
+2 (0) = λr
+4γ
+� π
+4 − 1
+π
+�
+−
+1
+16γ2 .
+Splitting BDRDB. In this case hs(x) = 2(2x2 − γ2)|x|/(γ2 + x2)3 and thus we find
+g(x) = −x2sign(x)/(γ2 + x2)2.
+It follows that f+
+2 (0) =
+3
+16γ2 and
+f+
+2,BDRDB(x) =
+�
+(x2 − 3γ2)2
+48γ2(x2 + γ2)2
+�
+1x<0 +
+�x4 − 54x2γ2 + 9γ4
+48γ2(x2 + γ2)2
+�
+1x≥0,
+f−
+2,BDRDB(x) =
+�
+(x2 − 3γ2)2
+48γ2(x2 + γ2)2
+�
+1x>0 +
+�x4 − 54x2γ2 + 9γ4
+48γ2(x2 + γ2)2
+�
+1x<0.
+Splitting RDBDR. Since hs(x) = 0 we have
+f+
+2,RR(x) = f−
+2,RDBDR(x) =
+1
+12γ2
+�1
+4 − 1
+�
+− 1
+12
+� γ2 − x2
+(γ2 + x2)2 − 1
+γ2
+�
+.
+Splitting DRBRD. Since hs(x) = 0 we have
+f+
+2,DRBRD(x) = f−
+2,DRBRD(x) = f+
+2 (0) − λr
+2 sign(x)
+�arctan(x/γ)
+γ
+−
+x
+γ2 + x2
+�
+− 1
+12
+� γ2 − x2
+(γ2 + x2)2 − 1
+γ2
+�
+− λ2
+r
+8 ln
+�
+1 + x2
+γ2
+�
+,
+with f+
+2 (0) = λr
+2γ
+� π
+4 − 1
+π
+�
++ λ2
+r
+8 ln 4 −
+1
+16γ2 .
+□
+Appendix E. Proof of Proposition 4.3
+In Section E.1 we obtain the first and second order commutators of BPS, while in Section E.2 we
+use the BCH formula and the obtained results to prove Proposition 4.3.
+E.1. Computing the commutators of BPS. In this section we compute the first and second
+order commutators for the various components of the adjoint of the BPS. In Section E.1.1 we write
+down the commutator of the BPS and its decomposition in the three terms that represent the free
+transport, reflection mechanism, and velocity refreshments. In Section E.1.2 we start with first order
+commutators, which are essential to compute second order commutators. The latter are computed in
+Section E.1.3.
+Now let us write the following identities, which form a lemma for convenience. These will be used
+countless times in the computation of the commutators below.
+
+SPLITTING SCHEMES FOR PDMPS
+63
+Lemma E.1. For λ(x, v) = ⟨v, ∇ψ(x)⟩+ it holds that
+λ1(x, R(x)v) − λ1(x, v) = −⟨v, ∇ψ(x)⟩,
+(58)
+λ1(x, R(x)v) + λ1(x, v) = +|⟨v, ∇ψ(x)⟩|.
+(59)
+The proof is trivial.
+E.1.1. The adjoint of BPS. Consider the generator
+Lf(x, v)=⟨v, ∇xf(x, v)⟩ + λ1(x, v)[f(x, R(x)v) − f(x, v)] + λ2
+� �
+f(x, w) − f(x, v)
+�
+ν(dw)
+Then one obtains that the adjoint is given by
+L∗
+BPSg(x, v) = −⟨v, ∇xg(x, v)⟩ + ((gλ1)(x, R(x)v) − (gλ1)(x, v)) + λr
+�
+ν(v)
+�
+g(x, y)dy − g(x, v)
+�
+= (L∗
+D + L∗
+B + L∗
+R)g(x, v),
+where
+L∗
+Dg(x, v) = −⟨v, ∇xg(x, v)⟩,
+L∗
+Bg(x, v) = g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v),
+L∗
+Rg(x, v) = λr
+�
+ν(v)
+�
+g(x, y)dy − g(x, v)
+�
+Here the letters D, B, and R stand for drift, bounce, refreshment. If we take g to be the invariant
+measure of BPS, µ, then
+L∗
+Dµ(x, v) = ⟨v, ∇ψ(x)⟩µ(x, v),
+L∗
+Bµ(x, v) = −⟨v, ∇ψ(x)⟩µ(x, v),
+L∗
+Rµ(x, v) = 0.
+To obtain L∗
+Dµ(x, v) we used the trivial, but useful, identity
+∇xµ(x, v) = −∇ψ(x)µ(x, v).
+E.1.2. First order commutators. Let us start computing the three first order commutators [L∗
+B, L∗
+D],
+[L∗
+R, L∗
+D], and [L∗
+R, L∗
+B], which are essential to compute higher order commutators. This is done below
+respectively in Lemmas E.2, E.4, E.5.
+Lemma E.2. Let g be a suitable function. It holds that
+[L∗
+B, L∗
+D]g(x, v) = −⟨R(x)v, (∇xg)(x, R(x)v)⟩λ1(x, R(x)v) + ⟨v, ∇xg(x, v)⟩λ1(x, v)
++ ⟨v, ∇x
+�
+g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v)
+�
+⟩.
+In particular if g = µ
+[L∗
+B, L∗
+D]µ(x, v) = µ(x, v)
+�
+⟨v, ∇ψ(x)⟩
+�
+⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩|
+�
+− ⟨v, ∇2ψ(x))v⟩
+�
+.
+Note E.3. Alternative ways to write [L∗
+B, L∗
+D]µ(x, v) can be found using the identities in Lemma E.1.
+We find
+[L∗
+B, L∗
+D]µ(x, v) = µ(x, v)
+�
+λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩
+�
+= µ(x, v)
+�
+λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩
+�
+.
+
+64
+SPLITTING SCHEMES FOR PDMPS
+Proof. We have
+[L∗
+B, L∗
+D]g(x, v) = L∗
+B(−⟨v, ∇xg(x, v)⟩) − L∗
+D(g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v))
+= −⟨R(x)v, (∇xg)(x, R(x)v)⟩λ1(x, R(x)v) + ⟨v, ∇xg(x, v)⟩λ1(x, v)
++ ⟨v, ∇x
+�
+g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v)
+�
+⟩
+and hence
+[L∗
+B, L∗
+D]µ(x, v) = −µ(x, v)⟨v, ∇ψ(x)⟩(λ1(x, R(x)v) + λ1(x, v))
++ ⟨v, ∇x
+�
+µ(x, v)(λ1(x, R(x)v) − λ1(x, v))
+�
+⟩
+= µ(x, v)(λ2
+1(x, R(x)v) − λ2
+1(x, v))
+− ⟨v, ∇x
+�
+µ(x, v)⟨v, ∇ψ(x)⟩
+�
+⟩.
+Then note that
+⟨v, ∇x
+�
+µ(x, v)⟨v, ∇ψ(x)⟩
+�
+⟩ = ⟨v, ∇2ψ(x)v − ∇ψ(x)⟨v, ∇ψ(x)⟩ ⟩ µ(x, v).
+and hence
+[L∗
+B, L∗
+D]µ(x, v) = µ(x, v)
+�
+⟨v, ∇ψ(x)
+�
+⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩|
+�
+⟩ − ⟨v, ∇2ψ(x))v⟩
+�
+.
+□
+Lemma E.4. Let g be a suitable function. It holds that
+[L∗
+R, L∗
+D]g(x, v) = λrν(v)
+�
+⟨v,
+�
+∇xg(x, y)dy⟩ −
+�
+⟨y, ∇xg(x, y)⟩dy
+�
+.
+In particular if g = µ
+[L∗
+R, L∗
+D]µ(x, v) = −λr⟨v, ∇ψ(x)⟩µ(x, v).
+Proof. We find
+[L∗
+R, L∗
+D]g(x, v) = −L∗
+R(⟨v, ∇xg(x, v)⟩) − L∗
+D
+�
+λr
+�
+ν(v)
+�
+g(x, y)dy − g(x, v)
+��
+= −λr
+�
+ν(v)
+�
+⟨y, ∇xg(x, y)⟩dy − ⟨v, ∇xg(x, v)⟩
+�
++ λr⟨v, ∇x(ν(v)
+�
+g(x, y)dy − g(x, v))⟩
+= λrν(v)
+�
+⟨v,
+�
+∇xg(x, y)dy⟩ −
+�
+⟨y, ∇xg(x, y)⟩dy
+�
+and thus
+[L∗
+R, L∗
+D]µ(x, v) = L∗
+R(⟨v, ∇ψ(x)⟩µ(x, v))
+= −λr⟨v, ∇ψ(x)⟩µ(x, v)
+□
+Lemma E.5. Let g be a suitable function. It holds that
+[L∗
+R, L∗
+B]g(x, v) = λr
+�
+ν(v)
+�
+(g(x, R(x)y)λ1(x, R(x)y) − g(x, y)λ1(x, y))dy
+
+SPLITTING SCHEMES FOR PDMPS
+65
++
+�
+ν(v)
+�
+g(x, y)dy
+�
+⟨v, ∇ψ(x)⟩
+�
+.
+In particular if g = µ
+[L∗
+R, L∗
+B]µ(x, v) = λr⟨v, ∇ψ(x)⟩µ(x, v).
+Proof. Compute
+[L∗
+R, L∗
+B]g(x, v) = L∗
+R(g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v))
+− L∗
+B
+�
+λr
+�
+ν(v)
+�
+g(x, y)dy − g(x, v)
+��
+= λr
+�
+ν(v)
+�
+(g(x, R(x)y)λ1(x, R(x)y) − g(x, y)λ1(x, y))dy
+− g(x, R(x)v)λ1(x, R(x)v)
+�
+��
+�
+A
++ g(x, v)λ1(x, v)
+�
+��
+�
+B
+�
+− λr
+� �
+�ν(R(x)v)
+�
+g(x, y)dy − g(x, R(x)v)
+�
+��
+�
+A
+�
+� λ1(x, R(x)v)
+−
+�
+�ν(v)
+�
+g(x, y)dy − g(x, v)
+� �� �
+B
+�
+� λ1(x, v)
+�
+.
+It is now sufficient to cancel out the terms denoted by A and B that appear twice with opposite signs
+to obtain the final statement. For g = µ
+[L∗
+R, L∗
+B]µ(x, v) = L∗
+R(−⟨v, ∇ψ(x)µ(x, v))
+= −λr
+�
+ν(v)
+�
+⟨y, ∇ψ(x)⟩p(x, y)dy − ⟨v, ∇ψ(x)⟩µ(x, v)
+�
+which concludes by Assumption 4.2.
+□
+Note E.6. It follows from Lemmas E.4 and E.5 that
+[L∗
+R, L∗
+B + L∗
+D]µ(x, v) = 0.
+(60)
+E.1.3. Higher order commutators. Let us now compute higher order commutators.
+Lemma E.7. It holds that
+[L∗
+B, [L∗
+R, L∗
+D]]µ(x, v) = λrµ(x, v)
+�
+⟨v, ∇ψ(x)⟩
+�
+λ1(x, R(x)v) + λ1(x, v)
+�
++ b tr
+�
+∇ψ(x)(∇ψ(x))T − ∇2ψ(x)
+��
+.
+Proof. Applying Lemma E.4 we obtain
+[L∗
+B, [L∗
+R, L∗
+D]]µ(x, v) = −λrL∗
+B
+�
+⟨v, ∇ψ(x)⟩µ(x, v)
+�
++ [L∗
+R, L∗
+D]
+�
+⟨v, ∇ψ(x)⟩µ(x, v)
+�
+= −λr
+�
+⟨R(x)v, ∇ψ(x)⟩µ(x, v)λ1(x, R(x)v) − ⟨v, ∇ψ(x)⟩µ(x, v)λ1(x, v)
+�
++ λrν(v)
+�
+⟨v, ∇x
+�
+(⟨y, ∇ψ(x)⟩µ(x, y))dy⟩ −
+�
+⟨y, ∇x(⟨y, ∇ψ(x)⟩µ(x, y))⟩dy
+�
+= λrµ(x, v)⟨v, ∇ψ(x)⟩
+�
+λ1(x, R(x)v) + λ1(x, v)
+�
+
+66
+SPLITTING SCHEMES FOR PDMPS
+− λrµ(x, v)
+� �
+(⟨y, ∇2ψ(x)y⟩ − ⟨y, ∇ψ(x)⟩2)ν(dy)
+�
+= λrµ(x, v)
+�
+⟨v, ∇ψ(x)⟩
+�
+λ1(x, R(x)v) + λ1(x, v)
+�
++ b tr
+�
+∇ψ(x)(∇ψ(x))T − ∇2ψ(x)
+��
+.
+Note that in the last line we used that ⟨a, b⟩2 = ⟨a, bbT a⟩ and that
+�
+⟨y, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))y⟩ν(dy) =
+d
+�
+j=1
+d
+�
+ℓ=1
+(∇ψ(x)∇ψ(x)T − ∇2ψ(x))jℓ
+�
+(yjyℓ)ν(dy)
+= b tr
+�
+∇ψ(x)∇ψ(x)T − ∇2ψ(x)
+�
+(61)
+which is a consequence of Assumption 4.2.
+□
+Lemma E.8. It holds that
+[L∗
+R, [L∗
+R, L∗
+B]]µ(x, v) = −λ2
+rµ(x, v)⟨v, ∇ψ(x)⟩.
+Proof. Next we compute [L∗
+R, [L∗
+R, L∗
+B]]. Since L∗
+Rµ(x, v) = 0 we easily find
+[L∗
+R, [L∗
+R, L∗
+B]]µ(x, v) = L∗
+R(λr⟨v, ∇ψ(x)⟩µ(x, v))
+= −λ2
+rµ(x, v)⟨v, ∇ψ(x)⟩.
+□
+Lemma E.9. It holds that
+[L∗
+R, [L∗
+R, L∗
+D]]µ(x, v) = λ2
+rµ(x, v)⟨v, ∇ψ(x)⟩.
+Proof. The result follows from Lemma E.8 and (60).
+□
+Lemma E.10. It holds that
+[L∗
+R, [L∗
+B, L∗
+D]]µ(x, v) = λrµ(x, v)
+�
+b tr
+�
+∇ψ(x)∇ψ(x)T − ∇2ψ(x)
+�
+−
+�
+λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩
+��
+Proof. Taking advantage of Lemma E.2
+[L∗
+R, [L∗
+B, L∗
+D]]µ(x, v) = L∗
+R
+�
+µ(x, v)
+�
+λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩
+��
+= λrµ(x, v)
+� � �
+λ2
+1(x, R(x)y) − λ2
+1(x, y) + ⟨y, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))y⟩
+�
+ν(dy)
+−
+�
+λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩
+��
+.
+Observe that for A = {y : ⟨y, ∇ψ(x)⟩ ≥ 0} we have
+�
+(λ2
+1(x, R(x)y) − λ2
+1(x, y))ν(dy) =
+�
+AC⟨y, ∇ψ(x)⟩2ν(y)dy −
+�
+A
+⟨y, ∇ψ(x)⟩2ν(y)dy
+= 0.
+This can be seen by the change of variables y′ = R(x)y in the first integral. The result then follows
+by using (61).
+□
+
+SPLITTING SCHEMES FOR PDMPS
+67
+Lemma E.11. It holds that
+[L∗
+B, [L∗
+R, L∗
+B]]µ(x, v) = −λrµ(x, v)⟨v, ∇ψ(x)⟩(λ1(x, R(x)v) + λ1(x, v)).
+Proof. Consider now
+[L∗
+B, [L∗
+R, L∗
+B]]µ(x, v) = L∗
+B(λr⟨v, ∇ψ(x)⟩µ(x, v)) + [L∗
+R, L∗
+B](⟨v, ∇ψ(x)⟩µ(x, v))
+= λrµ(x, v)
+�
+⟨R(x)v, ∇ψ(x)⟩λ1(x, R(x)v) − ⟨v, ∇ψ(x)⟩λ1(x, v)
++
+� �
+⟨R(x)y, ∇ψ(x)⟩λ1(x, R(x)y⟩ − ⟨y, ∇ψ(x)⟩λ1(x, y)
+�
+ν(dy)
++
+�
+(⟨y, ∇ψ(x)⟩ν(dy)⟨v, ∇ψ(x)⟩
+�
+.
+The last term equals zero as ν has mean zero. Then observe that by Identity (58)
+� �
+⟨R(x)y, ∇ψ(x)⟩λ1(x, R(x)y) − ⟨y, ∇ψ(x)⟩λ1(x, y)
+�
+ν(dy) =
+= −
+�
+⟨y, ∇ψ(x)⟩(λ1(x, R(x)y) + λ1(x, y))ν(dy)
+= −
+� �
+λ1(x, R(x)y)2ν(dy) −
+�
+λ1(x, y)2ν(dy)
+�
+= 0,
+(62)
+where the last equality follows by invariance under rotation of ν as required in Assumption 4.2. Hence
+we have obtained the statement.
+□
+Lemma E.12. It holds that
+[L∗
+B, [L∗
+B, L∗
+D]]µ(x, v) = 2µ(x, v)λ1(x, R(x)v)
+�
+⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩
+− ⟨R(x)v, ∇ψ2(x)R(x)v⟩
+�
+.
+Proof. By Lemma E.2 we find
+[L∗
+B, [L∗
+B, L∗
+D]]µ(x, v) = L∗
+B
+�
+µ(x, v)(λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩)
+�
+(*)
++ [L∗
+B, L∗
+D]
+�
+⟨v, ∇ψ(x)⟩µ(x, v)
+�
+.
+(**)
+Let us treat the two terms separately, starting with (*). After applying L∗
+B and using that R(x)(R(x)v) =
+v the first term becomes
+(*) = µ(x, v)
+��
+(λ2
+1(x, v) − λ2
+1(x, R(x)v) + ⟨R(x)v, ∇ψ(x)⟩2 − ⟨R(x)v, ∇2ψ(x)R(x)v⟩
+�
+λ1(x, R(x)v)
+−
+�
+λ2
+1(x, R(x)v) − λ2
+1(x, v) + ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩)
+�
+λ1(x, v)
+�
+= µ(x, v)
+�
+(λ2
+1(x, v) − λ2
+1(x, R(x)v))(λ1(x, R(x)v) + λ1(x, v))
++ ⟨v, ∇ψ(x)⟩2(λ1(x, R(x)v) − λ1(x, v))
+− ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩λ1(x, v)
+�
+.
+Using Identity (58) we obtain that
+⟨v, ∇ψ(x)⟩2(λ1(x, R(x)v) − λ1(x, v)) = (λ2
+1(x, v) − λ2
+1(x, R(x)v))(λ1(x, R(x)v) + λ1(x, v))
+
+68
+SPLITTING SCHEMES FOR PDMPS
+and thus cancelling out the corresponding terms in (*) it follows that
+(*) = µ(x, v)
+�
+⟨v, ∇2ψ(x)v⟩λ1(x, v) − ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v)
+�
+.
+Focusing now on (**), we apply Lemma E.2 to find
+(**) = −⟨R(x)v, ∇x
+�
+⟨v, ∇ψ(x)⟩µ(x, v)
+�
+(x, R(x)v)⟩λ1(x, R(x)v)
++ ⟨v, ∇x
+�
+⟨v, ∇ψ(x)⟩µ(x, v)
+�
+⟩λ1(x, v)
++ ⟨v, ∇x
+�
+⟨R(x)v, ∇ψ(x)⟩µ(x, v)λ1(x, R(x)v) − ⟨v, ∇ψ(x)⟩µ(x, v)λ1(x, v)
+�
+⟩.
+Recalling that
+∇x(⟨v, ∇ψ(x)⟩µ(x, v)) = µ(x, v)(∇2ψ(x)v − ∇ψ(x)⟨v, ∇ψ(x)⟩),
+we find
+(**) = µ(x, v)
+� �
+−⟨R(x)v, ∇2ψ(x)R(x)v⟩ + ⟨v, ∇ψ(x)⟩2�
+λ1(x, R(x)v)
++
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨v, ∇ψ(x)⟩2�
+λ1(x, v)
+�
+− ⟨v, ∇x
+�
+⟨v, ∇ψ(x)⟩µ(x, v)|⟨v, ∇ψ(x)⟩|
+�
+⟩.
+In particular we used Lemma E.1 to write the last term more compactly. The derivative in the last
+term can be computed as follows
+− ⟨v, ∇x
+�
+⟨v, ∇ψ(x)⟩µ(x, v)|⟨v, ∇ψ(x)⟩|
+�
+⟩ =
+= −µ(x, v)⟨v, ∇2ψ(x)v|⟨v, ∇ψ(x)⟩| − ∇ψ(x)⟨v, ∇ψ(x)⟩|⟨v, ∇ψ(x)⟩|
++ ⟨v, ∇ψ(x)⟩sign(⟨v, ∇ψ(x)⟩)∇2ψ(x)v⟩
+= −µ(x, v)
+�
+− ⟨v, ∇ψ(x)⟩2|⟨v, ∇ψ(x)⟩| + ⟨v, ∇2ψ(x)v⟩ (|⟨v, ∇ψ(x)⟩| + ⟨v, ∇ψ(x)⟩sign(⟨v, ∇ψ(x)⟩))
+�
+= −µ(x, v)
+�
+− ⟨v, ∇ψ(x)⟩2|⟨v, ∇ψ(x)⟩| + ⟨v, ∇2ψ(x)v⟩2|⟨v, ∇ψ(x)⟩|
+�
+= µ(x, v)|⟨v, ∇ψ(x)⟩|
+�
+⟨v, ∇ψ(x)⟩2 − 2⟨v, ∇2ψ(x)v⟩
+�
+.
+Hence re-applying Lemma E.1 we find
+(**) = µ(x, v)
+�
+− ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v)
++ 2⟨v, ∇ψ(x)⟩2λ1(x, R(x)v) − (2λ1(x, R(x)v) + λ1(x, v))⟨v, ∇2ψ(x)v⟩
+�
+.
+The proof is now concluded by summing (*) and (**).
+□
+Lemma E.13. It holds that
+[L∗
+D, [L∗
+R, L∗
+B]]µ(x, v) = λrµ(x, v)
+�
+⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩
+�
+.
+Proof. Consider now [L∗
+D, [L∗
+R, L∗
+B]]:
+[L∗
+D, [L∗
+R, L∗
+B]]µ(x, v) = L∗
+D(λr⟨v, ∇ψ(x)⟩µ(x, v)) − [L∗
+R, L∗
+B](⟨v, ∇ψ(x)⟩µ(x, v))
+= −⟨v, ∇x(λr⟨v, ∇ψ(x)⟩µ(x, v))⟩
+− λr
+�
+µ(x, v)
+�
+(−⟨y, ∇ψ(x)⟩)(λ1(x, R(x)y) + λ1(x, y))ν(dy)
+�
+= λrµ(x, v)
+�
+⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩
+�
+.
+
+SPLITTING SCHEMES FOR PDMPS
+69
+In particular we used that
+�
+(⟨y, ∇ψ(x)⟩)(λ1(x, R(x)y) + λ1(x, y))ν(dy) =
+�
+λ1(x, y)2ν(dy) −
+�
+λ1(x, R(x)y)2ν(dy) = 0
+which was shown in (62).
+□
+Lemma E.14. It holds that
+[L∗
+D, [L∗
+R, L∗
+D]]µ(x, v) = λrµ(x, v)
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨v, ∇ψ(x)⟩2
++ b tr
+�
+∇2ψ(x) − ∇ψ(x)∇ψ(x)T � �
+Proof. By Lemma E.4
+[L∗
+D, [L∗
+R, L∗
+D]]µ(x, v) = −λrL∗
+D(⟨v, ∇ψ(x)⟩µ(x, v)) − [L∗
+R, L∗
+D](⟨v, ∇ψ(x)⟩µ(x, v))
+= λrµ(x, v)
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨v, ∇ψ(x)⟩2
++
+�
+(⟨y, ∇2ψ(x)y⟩ − ⟨y, ∇ψ(x)⟩2)ν(dy)
+�
+.
+The statement follows by Equation 61.
+□
+Lemma E.15. It holds that
+[L∗
+D, [L∗
+B, L∗
+D]] = µ(x, v)
+�
+− 4⟨v, ∇ψ(x)⟩2λ1(x, R(x)v) + 7⟨v, ∇2ψ(x)v⟩λ1(x, R(x)v)
++ ⟨v, ∇x(⟨v, ∇2ψ(x))v⟩)⟩ + ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v)
+�
+.
+Proof. By Lemma E.2 together with Lemma E.1
+[L∗
+D, [L∗
+B, L∗
+D]]µ(x, v) = L∗
+D
+�
+µ(x, v)
+�
+⟨v, ∇ψ(x)⟩
+�
+⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩|
+�
+− ⟨v, ∇2ψ(x)v⟩
+��
+(†)
+− [L∗
+B, L∗
+D](⟨v, ∇ψ(x)⟩µ(x, v)).
+(††)
+= (†) − (††).
+Consider the two terms separately, starting from the first one:
+(†) = µ(x, v)⟨v, ∇ψ(x)⟩
+�
+⟨v, ∇ψ(x)⟩
+�
+⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩|
+�
+− ⟨v, ∇2ψ(x))v⟩
+�
+− µ(x, v)⟨v, 2⟨v, ∇ψ(x)⟩∇2ψ(x)v − 2∇2ψ(x)v|⟨v, ∇ψ(x)⟩| − ∇x(⟨v, ∇2ψ(x))v⟩)⟩
+= µ(x, v)
+�
+− 2⟨v, ∇ψ(x)⟩2λ1(x, R(x)v) + ⟨v, ∇2ψ(x)⟩(−3⟨v, ∇ψ(x)⟩ + 2|⟨v, ∇ψ(x)⟩|)
++ ⟨v, ∇x(⟨v, ∇2ψ(x))v⟩)⟩
+�
+The second term (††) is the same as term (**) in the proof of Lemma E.12. The statement follows
+taking the difference of the two terms (†) and (††) and using Lemma E.1.
+□
+E.2. Proof of Proposition 4.3. Consider symmetric splitting schemes of the form
+eδLS = e
+δ
+2 LAe
+δ
+2 LBeδLCe
+δ
+2 LBe
+δ
+2 LA.
+We have by the Baker-Campbell-Haussdorff formula
+L∗
+S = L∗ + δ2
+12
+�
+[L∗
+C, [L∗
+C, L∗
+A + L∗
+B]] + [L∗
+B, [L∗
+B, L∗
+A]] + [L∗
+C, [L∗
+B, L∗
+A]] + [L∗
+B, [L∗
+C, L∗
+A]]
+− 1
+2[L∗
+B, [L∗
+B, L∗
+C]] − 1
+2[L∗
+A, [L∗
+A, L∗
+C]] − 1
+2[L∗
+A, [L∗
+A, L∗
+B]]
+�
++ O(δ4)
+
+70
+SPLITTING SCHEMES FOR PDMPS
+= L∗ + δ2L∗
+2 + O(δ4).
+where L∗ = L∗
+A + L∗
+B + L∗
+C. Therefore it is sufficient to use the commutators of Section E.1. Observe
+that L∗
+BPSµ(x, v) = 0. Let us start with L∗
+DBRBD:
+L∗
+DBRBDµ(x, v) = δ2
+12
+�
+[L∗
+R, [L∗
+R, L∗
+D + L∗
+B]] + [L∗
+B, [L∗
+B, L∗
+D]] + [L∗
+R, [L∗
+B, L∗
+D]] + [L∗
+B, [L∗
+R, L∗
+D]]
+− 1
+2[L∗
+B, [L∗
+B, L∗
+R]] − 1
+2[L∗
+D, [L∗
+D, L∗
+R]] − 1
+2[L∗
+D, [L∗
+D, L∗
+B]]
+�
++ O(δ4)
+= δ2
+12µ(x, v)
+�
+3
+2λr
+�
+b tr
+�
+∇ψ(x)∇ψ(x)T − ∇2ψ(x)
+�
++ 2⟨v, ∇ψ(x)⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩
+�
++ 3
+2λ1(x, R(x)v)
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩
+�
++ 1
+2⟨v, ∇x(⟨v, ∇2ψ(x)v⟩)⟩
+�
+.
+Then focus on L∗
+BDRDB:
+L∗
+BDRDBµ(x, v) = δ2
+12
+�
+[L∗
+R, [L∗
+R, L∗
+D + L∗
+B]] + [L∗
+D, [L∗
+D, L∗
+B]] + [L∗
+R, [L∗
+D, L∗
+B]] + [L∗
+D, [L∗
+R, L∗
+B]]
+− 1
+2[L∗
+D, [L∗
+D, L∗
+R]] − 1
+2[L∗
+B, [L∗
+B, L∗
+R]] − 1
+2[L∗
+B, [L∗
+B, L∗
+D]]
+�
++ O(δ4)
+= δ2
+12µ(x, v)
+�
+− 3
+2λr
+�
+b tr
+�
+∇ψ(x)∇ψ(x)T − ∇2ψ(x)
+�
++ 2⟨v, ∇ψ(x)⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩
+�
++ 3λ1(x, R(x)v)
+�
+− 2⟨v, ∇2ψ(x)v⟩ + ⟨v, ∇ψ(x)⟩2�
+− ⟨v, ∇(⟨v, ∇2ψ(x)v⟩)⟩
+�
+.
+Consider L∗
+RDBDR:
+L∗
+RDBDRµ(x, v) = δ2
+12
+�
+[L∗
+B, [L∗
+B, L∗
+R + L∗
+D]] + [L∗
+D, [L∗
+D, L∗
+R]] + [L∗
+B, [L∗
+D, L∗
+R]] + [L∗
+D, [L∗
+B, L∗
+R]]
+− 1
+2[L∗
+D, [L∗
+D, L∗
+B]] − 1
+2[L∗
+R, [L∗
+R, L∗
+B]] − 1
+2[L∗
+R, [L∗
+R, L∗
+D]]
+�
++ O(δ4)
+= δ2
+12µ(x, v)
+�
+3
+2λ1(x, R(x)v)
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩
+�
++ 1
+2⟨v, ∇x(⟨v, ∇2ψ(x)v⟩)⟩
+�
+.
+Finally focus on L∗
+DRBRD:
+L∗
+DRBRDµ(x, v) = δ2
+12
+�
+[L∗
+B, [L∗
+B, L∗
+D + L∗
+R]] + [L∗
+R, [L∗
+R, L∗
+D]] + [L∗
+B, [L∗
+R, L∗
+D]] + [L∗
+R, [L∗
+B, L∗
+D]]
+− 1
+2[L∗
+R, [L∗
+R, L∗
+B]] − 1
+2[L∗
+D, [L∗
+D, L∗
+B]] − 1
+2[L∗
+D, [L∗
+D, L∗
+R]]
+�
++ O(δ4)
+= δ2
+12µ(x, v)
+�
+3
+2λr
+�
+b tr
+�
+∇ψ(x)∇ψ(x)T − ∇2ψ(x)
+�
++ ⟨v, ∇ψ(x)⟩
+�
+3λ1(x, R(x)v) + λ1(x, v)
+�
++ ⟨v, ∇2ψ(x)v⟩
+�
++ 3
+2λ1(x, R(x)v)
+�
+⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩
+�
++ 1
+2⟨v, ∇(⟨v, ∇2ψ(x)v⟩)⟩ + 3
+2λ2
+r⟨v, ∇ψ(x)⟩
+�
+.
+
diff --git a/hNE0T4oBgHgl3EQfpQGK/content/tmp_files/load_file.txt b/hNE0T4oBgHgl3EQfpQGK/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bc68f1ae8428bd5ca4463177d96c15a1f62c09f4
--- /dev/null
+++ b/hNE0T4oBgHgl3EQfpQGK/content/tmp_files/load_file.txt
@@ -0,0 +1,2332 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf,len=2331
+page_content='SPLITTING SCHEMES FOR SECOND ORDER APPROXIMATIONS OF PIECEWISE-DETERMINISTIC MARKOV PROCESSES ANDREA BERTAZZI1, PAUL DOBSON2, AND PIERRE MONMARCH´E3 Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Numerical approximations of piecewise-deterministic Markov processes based on splitting schemes are introduced, together with their Metropolis-adjusted versions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The unadjusted schemes are shown to have a weak error of order two in the step size in a general framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Focusing then on unadjusted schemes based on the Bouncy Particle and Zig-Zag samplers, we provide conditions ensuring ergodicity and consider the expansion of the invariant measure in terms of the step size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The dependency of the leading term in this expansion in terms of the refreshment rate, depending of the splitting order, is analyzed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, we present numerical experiments on Gaussian targets, a Bayesian imaging inverse problem and a system of interacting particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Introduction Piecewise deterministic Markov processes (PDMP) are non-diffusive Markov processes combining a deterministic motion and random jumps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' They appear in a wide range of modelling problems [13, 29, 31] and, over the last decade, have gained considerable interest as Markov Chain Monte Carlo (MCMC) methods [39, 34, 7, 10, 19, 43].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Their dynamics can be described by their infinitesimal generator, which is of the form Lf(z) = ⟨Φ(z), ∇zf(z)⟩ + λ(z) � E (f(y) − f(z))Q(z, dy) , (1) where E is the state space and, in this work, Φ is a smooth and globally Lipschitz vector field, λ : E → [0, ∞) is a continuous function and Q is a probability kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The associated process follows the ordinary differential equation (ODE) ˙z = Φ(z) and, at rate λ(z), jumps to a new position distributed according to Q(z, ·).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We refer to [15, 20] for general considerations on PDMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denoting by ϕt the integral curve of Φ, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the solution to d dtϕt(z) = Φ(ϕt(z)), ϕ0(z) = z, for all t ≥ 0, z ∈ E, which exists since Φ is globally Lipschitz, we assume that ϕt leaves E invariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For T ∼ Exp(1), the random time of the next jump, τ, is given by τ = inf � t > 0 : � t 0 λ(ϕs(z))ds ≥ T � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (2) This work addresses the question of the simulation of a PDMP with generator (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The classical method is to use a Poisson thinning procedure [30, 28] to sample the jump times, and then to solve the ODE exactly if possible, or otherwise by a standard numerical scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly to rejection sampling which requires a good reference measure, an efficient Poisson thinning algorithm requires the knowledge of good bounds for the jump rate λ along the trajectory of the ODE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this work, we 1 Delft Institute of Applied Mathematics, TU Delft 2 University of Edinburgh 3 Sorbonne Universit´e E-mail addresses: a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='bertazzi@tudelft.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='nl, pdobson@ed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='uk, pierre.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='monmarche@sorbonne-universite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='fr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Date: January 9, 2023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='02537v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='PR] 6 Jan 2023 2 SPLITTING SCHEMES FOR PDMPS focus on the case in which such bounds are not available, or are so crude that thinning would not be numerically efficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In that case, the random event times have to be approximated even if the ODE can be solved exactly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This question has recently been addressed in [3, 38, 14] with three different schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here, rather than designing an ad hoc numerical schemes, we work in the general framework of splitting schemes, which are widely used for e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hamiltonian or underdamped Langevin processes [26, 27, 36].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' One of the main interests is that, by design, such schemes have a numerical error which is of order 2 in the step-size, without the need of an approximation of the jump rate along the ODE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, it is a flexible framework and thus such schemes can be easily combined with multi-time- step or factorization methods [25] or integrated in hybrid PDMP/diffusion schemes [37, 35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that, by using a numerical approximation, we lose one of the interest of PDMP for MCMC purpose, which is the exact simulation by thinning, while in our case the invariant measure of the scheme will have a deterministic bias with respect to the true target measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' However, we still benefit from the good long-time convergence properties of the ballistic non-reversible process and, contrary to Hamiltonian-based dynamics, it is still possible to factorize the target measure and define efficient schemes in terms of number of computations of forces (see [37, 35] and Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall also show how the correct stationary distribution can be recovered by means of a non-reversible Metropolis- Hastings acceptance/rejection step (see Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, for classical velocity jump processes used in MCMC, since the norm of the velocity is constant (between possible refreshments which are independent of the potential), these schemes are numerically stable (see the numerical experiments in Section 5 where the step-size of PDMP schemes can be taken larger than for the classical ULA), even for non-globally Lipschitz potentials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The core idea of splitting schemes is first to split the generator in several parts such that a process associated to each part can be simulated exactly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For instance, when the ODE can be solved exactly, one can write L = LD + LJ with LDf(z) = ⟨Φ(z), ∇zf(z)⟩, LJf(z) = λ(z) � E (f(y) − f(z))Q(z, dy) , in which case the process associated to LD is simply the solution of the ODE, hence D stands for drift, while the process associated to LJ is a continuous-time Markov chain, for which the jump rate is constant between two jumps (so that the jump times are simple exponential random variables), hence J stands for jumps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then, one approximates the semigroup of the true process by a Strang splitting Pδ = eδ(LD+LJ) ≈ e δ 2 LDeδLJe δ 2 LD (3) for a small step size δ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore, over one time step the approximation follows LD for time δ/2, then LJ for time δ and finally LD again for time δ/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Given a step size δ, now we illustrate how the (n + 1)-th iteration works.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Starting at time tn = nδ at state Ztn the process first moves deterministically for a half step: Ztn+δ/2 = ϕδ/2(Ztn).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we simulate the pure jump part of the process: we generate an event time τ1 ∼ Exp(λ(Ztn+δ/2)) and, if τ1 < δ, we set Ztn+δ/2 ∼ Q(Ztn+δ/2, ·).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we repeat this step as long as � i τi < δ, though, since we are interested in second order schemes, it is enough to limit ourselves to two jumps per time step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that the rate is updated after every jump and is constant between jumps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We conclude the iteration by a final half step of deterministic motion: Ztn+1 = ϕδ/2(Ztn+δ/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We refer to this scheme as the splitting scheme DJD, where consistently with above D stands for drift and J for jumps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' When the ODE cannot be solved exactly, any second-order numerical scheme SPLITTING SCHEMES FOR PDMPS 3 can be used instead of ϕt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, in some cases (typically for the Hamiltonian dynamics) the generator LD can be further divided in several ODEs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly, for computational purpose, it can be interesting in some cases to split the jump part LJ in several operators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is also possible to keep in LD a combination of ODE and jump, simulated e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' by thinning, while some parts of the jump are treated separately in LJ (it could make sense for instance in the context of [37]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' When there are more than two parts in the splitting of L, a scheme is obtained by starting from (3) and using e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' eδLJ ≈ e δ 2 LAeδLBe δ 2 LA if LJ = LA + LB, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Such splitting schemes can be used to simulate any PDMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For some modelling problems, it can be interesting to have estimates on the trajectorial error between the approximated process and the two process, for instance when dynamical properties (like mean squared displacement or transition rates) are of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' However, in this work, we have mainly in mind the PDMPs which are used for MCMC methods, in particular our recurrent examples will be the Zig-Zag sampler (ZZS) [7, 5] and the Bouncy Particle sampler (BPS) [39, 34, 10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a consequence, we will not discuss trajectorial errors but rather focus on what is relevant for MCMC purpose, namely the long-time convergence of the Markov chain (which should scale properly as the step size vanishes) and the numerical bias on the invariant measure and on empirical averages of the chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Organisation of the paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The article is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We conclude this introduction by presenting the algorithms we focus on in this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 we discuss our two main examples and their approximation with splitting schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Sections 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 and 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we discuss respectively how we can Metropolis-adjust our schemes in a non-reversible fashion and how we can modify the algorithms to do subsampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We conclude our introduction with Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4, where we describe how boundaries can be treated with our splitting schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Section 2 is devoted to the analysis of the weak error for the finite-time empirical averages of the scheme DJD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The main result, Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6, states that for this scheme the weak error is of order 2 in the step-size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The geometric ergodicity of splitting schemes based on our main examples is established in Section 3, with a consistent dependency of the estimates on the step-size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section 4, we provide a formal expansion (in terms of the step-size) of the invariant measure of the scheme based on the so-called Bouncy Particle Sampler depending on the choice of the splitting, in the spirit of [26], with a particular focus in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 on three one-dimensional examples where everything can be made explicit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Numerical experiments are provided in Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, technical proofs are gathered in an Appendix.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Comparison to related works.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The work in this paper can be seen as a continuation of the work that two of the authors started with their coauthors in [3], in which a general framework to approximate PDMPs is introduced and studied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this previous work, the focus is not a specific scheme and thus the results are mostly general and not tailored for particular processes or schemes, though the ZZS and BPS are considered as recurrent examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, the schemes introduced in [3] leave considerable freedom to the user in the choice of some crucial components of the algorithm, namely an approximation of the switching rates or a numerical integrator in place of the exact flow map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' On the other hand, in this paper we follow the philosophy of splitting schemes to describe a simple recipe to approximate PDMPs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that splitting schemes are not considered in [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The main advantage of splitting schemes is the second order of accuracy with one gradient evaluation per iteration, whereas second order algorithms considered in [3] relied on approximations of second order of the switching rates, which can be usually obtained with the expensive computation of the Hessian of the negative log-target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, in this work we describe how to remove the bias introduced by our approximation with a non-reversible Metropolis-Hastings step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Two other works ([38] and [14]) focus on approximate simulation of the Zig-Zag sampler, which is one of our two main examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In [38] the authors suggest to approximate event times by using numerical approximations of the integral of the rates along the dynamics (2), as well as a root finding algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In [14], the authors suggest 4 SPLITTING SCHEMES FOR PDMPS Algorithm 1: Splitting scheme DBD for ZZS Input : Number of iterations N, initial condition (x, v), step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Output: Chain (Xtn, V tn)N n=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = 0, (X0, V 0) = (x, v);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' while n < N do Set Xtn+1 = Xtn + δ 2V tn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set V tn+1 = V tn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for i = 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d do With probability (1 − exp(δλi(Xtn+1, V tn+1))) set V tn+1 = RiV tn+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end Set Xtn+1 = Xtn+1 + δ 2V tn+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = n + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end using a numerical optimisation algorithm at each iteration to obtain a suitable bound that enables the use of Poisson thinning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The first difference is that we mainly consider our approximations as discrete time Markov chains, whereas the processes of [38] and [14] are interpreted in continuous time, although neither resulting process is a Markov process due to the nature of the numerical algorithms that are used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Naturally, one could interpret our algorithms as continuous time processes, which again would not be Markov processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Secondly, without assuming any properties that we do not verify, we derive theoretical justifications of our proposed algorithms, such as bounds on the weak error and existence, uniqueness, and geometric convergence to a stationary distribution under simple conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, we introduce Metropolis adjusted algorithms to eliminate the error introduced by the numerical approximations, while this aspect is not studied in previous works and thus we introduce the first exact PDMP based samplers that can be simulated with only access to the gradient of the negative logarithm of the target distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Main examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us now introduce two examples from the computational statistics liter- ature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this setting we have a target probability measure with density π(x) ∝ exp(−ψ(x)) for x ∈ Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Example 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 (Zig-Zag sampler [5]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let E = Rd × {+1, −1}d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For any z ∈ E, we write z = (x, v) for x ∈ Rd, v ∈ {+1, −1}d, where x is interpreted as the position of the particle and v denotes the corresponding velocity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The deterministic motion of ZZS is determined by Φ(x, v) = (v, 0)T , i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the particle travels with constant velocity v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d we define the jump rates λi(x, v) := (vi∂iψ(x))++γi(x, v), where γi(x) can be any non-negative function and is often chosen to be zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The corresponding (deterministic) jump kernels are given by Qi((x, v), (dy, dw)) = δ(x,Riv)(dy, dw), where δz denotes the Dirac delta measure and Ri is the operator that flips the sign of the i-th component of the vector it is applied to, that is Riv = (v1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , vi−1, −vi, vi+1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , vd).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence the i-th component of the velocity is flipped with rate λi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The ZZS is described by its generator Lf(x, v) = ⟨v, ∇xf(x, v)⟩ + d � i=1 λi(x, v)[f(x, Riv) − f(x, v)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (4) Simulating the event times with rates of this form is in general a very challenging problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 5 Algorithm 2: Splitting scheme RDBDR for BPS Input : Number of iterations N, initial condition (x, v), step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Output: Chain (Xtn, V tn)N n=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = 0, (X0, V 0) = (x, v);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' while n < N do Set V tn+1 = V tn ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With probability (1 − exp(−λr δ 2)) draw V tn+1 ∼ Unif(Sd−1) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set Xtn+1 = Xtn + δ 2V tn+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With probability (1 − exp(δλ1(Xtn+1, V tn+1))) set V tn+1 = R(Xtn+1)V tn+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set Xtn+1 = Xtn+1 + δ 2V tn+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With probability (1 − exp(−λr δ 2)) set V tn+1 ∼ Unif(Sd−1) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = n + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end We can apply the splitting scheme above as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assume the process has canonical rates, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' γi = 0 for all i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we can split the generator as LDf(x, v) = ⟨v, ∇xf(x)⟩, LBf(x, v) = d � i=1 λi(x, v)[f(x, Riv) − f(x, v)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we define the scheme DBD, where B stands for bounces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Given (Xtn, V tn), we start by a half step of deterministic motion: Xtn+ δ 2 = Xtn + δ 2V tn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d we draw τi iid ∼ Exp(λi(Xtn+δ/2, V tn)), which are homogeneous exponential random variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then let τ(1) = min τi and set V tn+1 = � V tn if τ(1) > δ RIV tn if τ(1) ≤ δ where RI = � i∈I Ri and I is the set of indices i for which τi ≤ δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Alternatively to have a second order scheme it is sufficient to flip only the two components with the smallest switching time τi, given that it is before time δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that for canonical rates flipping the sign of a component does not affect the other switching rates, and thus it is not possible to have two flips in the same component when γi = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, set Xtn+1 = Xtn+ δ 2 + δ 2V tn+1, which concludes the iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The procedure is described in pseudo code form in Algorithm 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' An interesting feature of the algorithm is that the jump part of the chain can be computed in parallel, since in that stage a velocity flip in one component does not affect the other components of the process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Example 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 (Bouncy Particle Sampler [10]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let E = Rd×Rd, and for any z ∈ E we write z = (x, v) for x ∈ Rd, v ∈ Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The deterministic motion is the same of ZZS: Φ(x, v) = (v, 0)T .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The BPS has two types of random events: reflections and refreshments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' These respectively have rates λ1(x, v) = (vT ∇xψ(x))+ and λ2(x, v) = λr for λr > 0, and corresponding jump kernels Q1((x, v), (dy, dw)) = δ(x,R(x)v)(y, w), Q2((x, v), (dy, dw)) = δx(dy)ν(dw), 6 SPLITTING SCHEMES FOR PDMPS where ν is a rotation-invariant probability measure on Rd (typically the standard Gaussian measure or the uniform measure on Sd−1), and R(x)v = v − 2⟨v, ∇xψ(x)⟩ |∇xψ(x)|2 ∇xψ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The operator R reflects the velocity v off the hyperplane that is tangent to the contour line of ψ passing though point x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Importantly, the norm of the velocity is unchanged by the application of R, and this corresponds to an elastic collision of the particle on the hyperplane.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The BPS has generator Lf(x, v)=⟨v, ∇xf(x)⟩ + λ1(x, v)[f(x, R(x)v) − f(x, v)] + λ2 � � f(x, w) − f(x, v) � ν(dw).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case we split the generator in three parts: LDf(x, v) = ⟨v, ∇xf(x)⟩, LBf(x, v) = λ1(x, v)[f(x, R(x)v) − f(x, v)], LRf(x, v) = λ2 � � f(x, w) − f(x, v) � ν(dw), We then define the scheme RDBDR, where R stands for refreshments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Starting at time tn = nδ at state (Xtn, V tn) we begin by drawing τ1 ∼ Exp(λr) and setting ˜Vtn+ δ 2 = � V tn if τ1 > δ/2 W1 if τ1 ≤ δ/2 for W1 ∼ ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then the process evolves deterministically for time δ/2: Xtn+ δ 2 = Xtn + δ 2 ˜Vtn+ δ 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' At this point, we check if a reflection takes place by drawing τ2 ∼ Exp(λ1(Xtn+ δ 2 , ˜Vtn+ δ 2 )) and set V tn+ δ 2 = � ˜Vtn+ δ 2 if τ2 > δ R(Xtn+ δ 2 ) ˜Vtn+ δ 2 if τ2 ≤ δ Importantly, λ1(Xtn+ δ 2 , V tn+ δ 2 ) = 0 if a reflection takes place and thus at most one reflection can happen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is a consequence of the fact that ⟨R(x)v, ∇ψ(x)⟩ = −⟨v, ∇ψ(x)⟩ by definition of the reflection operator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' After this we set Xtn+1 = Xtn+ δ 2 + δ 2, and finally conclude the iteration drawing τ3 ∼ Exp(λr) and letting ˜Vtn+1 = � V tn+ δ 2 if τ3 > δ/2 W2 if τ3 ≤ δ/2 where W2 ∼ ν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The pseudo code can be found in Algorithm 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Metropolis adjusted algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Naturally, the use of splitting schemes to approximate a PDMP introduces a discretisation error.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the context of Bayesian statistics, this means that a bias term is introduced in the estimators for statistics of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this section we discuss how to eliminate this bias with the addition of a Metropolis-Hastings (MH) acceptance-rejection step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 we describe the general procedure, which is a non-reversible MH algorithm, and then apply this to ZZS and BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly this can be applied to other kinetic PDMPs used in MCMC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 7 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Non-reversible Metropolis-Hastings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The classical MH algorithm gives a simple procedure to construct a µ invariant Markov chain P by satisfying detailed balance (DB): for all x, y it holds that µ(dx)P(x, dy) = µ(y)P(y, dx).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Integrating DB with respect to x one shows that P is µ-invariant, that is � µ(dx)P(x, dy) = µ(dy).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Given a proposal mechanism Q(x, ·) with density q with respect to some measure ν, MH constructs P by accepting the state y proposed by Q with probability a(x, y) = 1 ∧ µ(y)q(y, x) µ(x)q(x, y).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (5) The resulting chain P is reversible as it satisfies DB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' PDMPs such as BPS and ZZS break DB and are said to be non-reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since there is evidence that this property can lead to a faster converging process (see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [18]), it is reasonable here to Metropolise our splitting schemes in a non-reversible fashion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, as we shall see below, for our chains based on splitting schemes of PDMPs it is not possible to use the standard MH framework, as typically q(x, y) > 0 implies q(y, x) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The reasoning below is an extension of the idea of lifting (for more background on lifting we refer the reader to [42, 45, 24, 33]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall now consider a chain for which the state can be decomposed as z = (x, v), as is the case with the position and velocity parts of the PDMPs we consider.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this context, a sufficient condition other than DB ensuring stationarity is skew detailed balance: for all x, v, y, w µ(dx, dv)P((x, v), (dy, dw)) = µ(dy, dw)P((y, −w), (dx, −dv)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (6) Indeed, integrating both sides wrt x and v it follows � µ(dx, dv)P((x, v), (dy, dw)) = µ(dy, dw), that is µ is a stationary measure for the chain P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As in the context of the standard MH algorithm, assume now we wish to construct our P by accepting or rejecting proposals from a kernel Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denote as a((x, v), (y, w)) the probability of accepting proposal (y, w).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For x ̸= y in order for (6) to hold it should be that a((x, v), (y, w)) a((y, −v), (x, −v)) = µ(dy, dw)Q((y, −w), (dx, −dv)) µ(dx, dv)Q((x, v), (dy, dw)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence the acceptance probability can be taken to be a((x, v), (y, w)) = 1 ∧ µ(dy, dw)Q((y, −w), (dx, −dv)) µ(dx, dv)Q((x, v), (dy, dw)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (7) If the proposal (y, w) is rejected, the new state of the chain becomes (x, −v), in which case (6) is trivially satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This algorithm can be also obtained with a different reasoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose µ(dx, dv) = µ(dx, −dv).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' At every iteration, the algorithm generates (y, w) ∼ Q((x, v), ·) and successively flips the sign of the velocity w.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then the proposal (y, −w) is accepted or rejected with the classical MH step, that is (5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Afterwards, the sign of the velocity is flipped again, hence in case of acceptance the next state is (y, w), while in case of rejection it is (x, −v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This procedure can be justified noting that the MH step ensures that µ is stationary, while changing the sign of velocity preserves µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We remark that, although both the MH step and the velocity reflection are reversible with respect to µ, their composition is not.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Non-reversible Metropolis adjusted ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting DBD of ZZS with initial con- dition (x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let δ > 0 be the step size and x1/2(x, v) = x + vδ/2 (we will drop the dependence on (x, v) when clear).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As explained in Example 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1, after one iteration the algorithm has state ( ˜X, ˜V ) = (x1/2 + δ 2RIv, RIv) 8 SPLITTING SCHEMES FOR PDMPS Algorithm 3: Non-reversible Metropolis adjusted ZZS Input : Number of iterations N, initial condition (x, v), step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Output: Chain (Xtn, V tn)N n=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = 0, (X0, V 0) = (x, v);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' while n < N do Set Xtn+δ/2 = Xtn + δ 2V tn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set ˜V = V tn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for i = 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d do With probability (1 − exp(δλi(Xtn+δ/2, ˜V ))) set ˜V = Ri ˜V ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end Set ˜X = Xtn+δ/2 + δ 2 ˜V ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set (Xtn+1, V tn+1) = ( ˜X, ˜V ) with probability 1 ∧ π( ˜X) π(Xtn) exp � δ d � j=1 � λj(Xtn+δ/2, V tn) − λj(Xtn+δ/2, − ˜V ) � � else set (Xtn+1, V tn+1) = (Xtn, −V tn);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = n + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end with corresponding probability exp � −δ � i/∈I λi(x1/2, v) � � i∈I (1 − exp(−δλi(x1/2, v)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (8) We now want to accept or reject the proposed state with suitable probability to ensure µ-stationarity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note the classical MH scheme (5) is not directly applicable, as typically there is a 0 probability that the process goes from ( ˜X, ˜V ) to (x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence we use the non-reversible MH acceptance probability (7).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For this we need to compute the probability of going from ( ˜X, − ˜V ) to (x, −v) according the transition kernel of DBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This can only be achieved by following the same path of (x, v) → ( ˜X, ˜V ) with reversed time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence the sign of the velocity of the components in α needs to be flipped.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Noticing that x1/2(x, v) = x1/2( ˜X, − ˜V ), we find that the probability of this path is exp � −δ � i/∈I λi(x1/2, − ˜V ) � � i∈I (1 − exp(−δλi(x1/2, − ˜V )), where I is the same set of indices of (8).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that for i ∈ I it holds that ˜Vi = −vi and thus λi(x1/2, v) = λi(x1/2, − ˜V ), while for i /∈ I we have ˜Vi = vi and hence λi(x1/2, v) − λi(x1/2, − ˜V ) = vi∂ψ(x1/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore the acceptance probability (7) simplifies to 1 ∧ π( ˜X) × exp(−δ � i/∈I λi(x1/2, − ˜V )) π(x) × exp(−δ � i/∈I λi(x1/2, v)) = 1 ∧ exp � ψ(x) − ψ( ˜X) + δ � i/∈I vi∂iψ(x1/2) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (9) In case of rejection, the state is set to (x, −v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The procedure is written as pseudo-code in Algorithm 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 9 Algorithm 4: Non-reversible Metropolis adjusted BPS Input : Number of iterations N, initial condition (x, v), step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Output: Chain (Xtn, V tn)N n=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = 0, (X0, V 0) = (x, v);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' while n < N do Set V tn+δ/2 = V tn ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With probability (1 − exp(−λrδ/2)) draw V tn+δ/2 ∼ Unif(Sd−1) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set Xtn+δ/2 = Xtn + δ 2V tn+δ/2;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set ˜V = V tn+δ/2;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With probability (1 − exp(δλ1(Xtn+δ/2, ˜V ))) set ˜V = R(Xtn+δ/2) ˜V ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set ˜X = ˜X + δ 2 ˜V ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set (Xtn+1, V tn+1) = ( ˜X, ˜V ) with probability 1 ∧ π( ˜X) × exp(−δλ(Xtn+δ/2, − ˜V )) π(Xtn) × exp(−δλ(Xtn+δ/2, V tn)) else set (Xtn+1, V tn+1) = (Xtn, −V tn+δ/2) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With probability (1 − exp(−λrδ/2)) set V tn+1 ∼ Unif(Sd−1) ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = n + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end Note 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assuming ψ ∈ C2, the terms ψ(x) and ψ( ˜X) = ψ(x1/2 + RIv δ/2) can be expanded by Taylor’s theorem around x1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This gives that the acceptance probability (9) is 1 ∧ exp �δ2 8 � ⟨v, ∇2ψ(x1)v⟩ − ⟨RIv, ∇2ψ(x2)RIv⟩ �� , (10) where x1 ∈ (x, x1/2) and x2 ∈ (x1/2, x1/2 +RIvδ/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This result shows that the probability of rejecting the proposed state is of order δ2 and gives first evidence that the splitting scheme introduces an error of second order in the invariant measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, it is clear from (10) that the acceptance probability equals 1 for instance when ∇2ψ is a constant diagonal matrix, as is the case in a d-dimensional independent Gaussian vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this setting the splitting scheme DBD has the correct stationary distribution µ and does not need a Metropolis correction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, observe that the rejection probability is of order δ3 if ∇2ψ is diagonal but not constant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Non-reversible Metropolis adjusted BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we consider scheme RDBDR of BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The re- freshment does not alter the stationary distribution of the process, thus we focus first on the DBD part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denote x1/2(x, v) = x + δv/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' According to DBD, the process moves from an initial condition (x, v) to ( ˜X, ˜V ) = � (x1/2 + δ 2R(x1/2)v, R(x1/2)v) with probability 1 − exp(−δλ(x1/2, v)), (x + δv, v) with probability exp(−δλ(x1/2, v)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (11) Observe that for both states in (11) it holds x1/2(x, v) = x1/2( ˜X, − ˜V ) = x1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We now focus on computing the acceptance probability (7) in the two cases in (11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider first the case in which a reflection took place, which corresponds to the first line of (11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we need to compute the probability that the process goes from ( ˜X, − ˜V ) back to (x, −v) using scheme DBD, which is equal to the probability that the process has a reflection at x1/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By definition 10 SPLITTING SCHEMES FOR PDMPS of the reflection rate λ, it holds that λ(x1/2, v) = λ(x1/2, −R(x1/2)v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore in this case the probability that the process goes from ( ˜X, − ˜V ) to (x, −v) is the same as that of going from (x, v) to ( ˜X, ˜V ) and thus the acceptance probability (7) is 1 ∧ π(x1/2 + δ 2R(x1/2)v) π(x) = 1 ∧ exp � ψ(x) − ψ � x1/2 + δR(x1/2)v/2 � � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (12) Observe that moves that decrease ψ are accepted with probability 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider now the second case in (11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The probability that the process goes from (x + δv, −v) to (x, −v) is exp(−δλ(x1/2, −v)), while the probability of going from (x, v) to (x+δv, v) is exp(−δλ(x1/2, v)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observing that λ(x1/2, v) − λ(x1/2, −v) = ⟨v, ∇ψ(x1/2)⟩ we find that in this case the MH acceptance probability is 1 ∧ π(x + δv) × exp(−δλ(x1/2, −v)) π(x) × exp(−δλ(x1/2, v)) = 1 ∧ exp � ψ(x) − ψ(x + vδ) + δ⟨v, ∇ψ(x1/2)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (13) Hence we have shown that the acceptance probability can in general be written as 1 ∧ π( ˜X) × exp(−δλ(x1/2, − ˜V )) π(x) × exp(−δλ(x1/2, v)) = 1 ∧ exp � ψ(x) − ψ( ˜X) + δ(λ(x1/2, v) − λ(x1/2, − ˜V )) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In case of rejection, the state is set to (x, −v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Two refreshments half-steps, to be executed before and after the scheme DBD, are necessary to ensure irreducibility of the Markov chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The described procedure is written in pseudo code form in Algorithm 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us Taylor expand the acceptance probabilities similarly to Note 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed for (12) we expand both terms around x1/2 and for x1 ∈ (x, x1/2) and x2 ∈ (x1/2, x + vδ) we obtain ψ(x) − ψ � x1/2 + δ 2R(x1/2)v � = −δ 2 � ⟨v, ∇ψ(x1/2)⟩ + ⟨R(x1/2)v, ∇ψ(x1/2)⟩ � + δ2 8 � ⟨v, ∇2ψ(x1)v⟩ − ⟨R(x1/2)v, ∇2ψ(x2)R(x1/2)v⟩ � = δ2 8 � ⟨v, ∇2ψ(x1)v⟩ − ⟨R(x1/2)v, ∇2ψ(x2)R(x1/2)v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the last line we used that ⟨R(x1/2)v, ∇ψ(x1/2)⟩ = −⟨v, ∇ψ(x1/2)⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly, in (13) we expand terms ψ(x) and ψ(x + vδ) around x1/2 to find ψ(x) − ψ(x + vδ) + δ⟨v, ∇ψ(x1/2)⟩ =δ2 8 � ⟨v, ∇2ψ(x1)v⟩ − ⟨v, ∇2ψ(x2)v⟩ � for x1 ∈ (x, x1/2) and x2 ∈ (x1/2, x + vδ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If the Hessian of ψ is constant, as for instance in the Gaussian case, then proposals of this type are accepted with probability 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Overall, this means the acceptance probability has the form 1 ∧ exp �δ2 8 � ⟨v, ∇2ψ(x1)v⟩ − ⟨ ˜V , ∇2ψ(x2) ˜V ⟩ �� , which means that the probability of rejecting the proposed state is of order δ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular if π is a d-dimensional Gaussian with covariance Σ = cId, then the probability of accepting the proposed state in the MH step is equal to 1, as ∥ ˜V ∥ = ∥v∥ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence in this case the splitting scheme RDBDR has the correct stationary distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 11 Algorithm 5: Splitting scheme DBD for ZZS with subsampling Input : Number of iterations N, initial condition (x, v), step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Output: Chain (Xtn, V tn)N n=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = 0, (X0, V 0) = (x, v);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' while n < N do Set Xtn+1 = Xtn + δ 2V tn;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Draw J ∼ Unif({1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , N});' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for i = 1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d do Obtain (V tn+1)i by simulating a pure jump process with kernel Ri and rate v �→ (v∂iψj(Xtn+1))+ with initial velocity (V tn)i and time horizon δ ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end Set Xtn+1 = Xtn+1 + δ 2V tn+1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set n = n + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' end 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Algorithms with subsampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' One of the attractive features of ZZS and BPS is exact sub- sampling, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the possibility when the potential is of the form ψ(x) = 1 N �N j=1 ψj(x) of using only a randomly chosen ψj to simulate the next event time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The clearest application of this technique is Bayesian statistics, where ψ(x) is the posterior distribution, x is the parameter of the chosen statistical model and, when the data points are independent realisations, ψj can be chosen to depend only on the j-th batch of data points and not on the rest of the dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore, this technique can greatly reduce the computational cost per event time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Naturally, Bayesian statistics is not the only area where this structure of ψ arises.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' An example from molecular dynamics whit this type of potential is considered in Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we define a splitting scheme of ZZS with this feature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' With the same ideas it is possible to define a splitting scheme with subsampling based on BPS, but we do not give the details here for the sake of brevity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us briefly explain the basic idea in the case of ZZS, as given in [5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assume the target distri- bution is of the form ψ(x) = 1 N �N j=1 ψj(x) and define the switching rates λj i(x, v) = (vi∂iψj(x))+ for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d and j = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assuming we have a tractable M such that λj i(x + vt, v) ≤ M(t) for all j = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , N, one can use Poisson thinning to obtain a proposal τ for the next event time distributed as Exp(M(t)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This proposal is then accepted with probability λJ i (x + vτ, v)/M(τ), where J ∼ Unif({1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , N}) independently of the rest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This procedure defines a ZZS with switching rates λi(x, v) = 1 N �N j=1 λj i(x, v), which are larger than the canonical rates, but keep π stationary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Clearly the bottleneck of this procedure is that a sharp bound M needs to be available.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Algorithm 5 defines an approximation of this process with a similar idea as [3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' At each iteration the algorithm draws J ∼ Unif({1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , N}) independently of the rest and uses the corresponding ψJ to update the process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since the rates are now larger than the canonical rates, that is λi(x, v) > (vi∂iψ(x))+, there can be more than one jump per component at each iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Nonetheless, the algorithm requires only one gradient computation per iteration since the position is not updated during the jump part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, in this case obtaining the gradient ∇ψJ is an order 1 computation as opposed to the usual order N needed to compute the full gradient ∇ψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' PDMPs with boundaries.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Another interesting feature of PDMPs such as BPS and ZZS is that, thanks to the simple deterministic dynamics, boundary conditions can be included and hitting times of the boundary can be easily computed (see [16] or [12] for a discussion of PDMPs with boundaries).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we illustrate how to simply adapt splitting schemes to these settings by adding the boundary behaviour to the D part of the scheme.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 12 SPLITTING SCHEMES FOR PDMPS Boundary terms appear for instance when the target distribution π is defined on a restricted domain, in which case a boundary jump kernel can be introduced as considered in [4].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case, Algorithms 1 and 2 can be easily modified by incorporating the boundary term in part D of the splitting scheme, as typically the boundary can be hit only if there is deterministic motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence, the continuous deterministic dynamics are applied as in the exact process, while other jumps are performed in the B steps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Another example of this setting is when π is a mixture of a continuous density and a discrete distribution on finitely many states, as in Bayesian variable selection when a spike and slab prior is chosen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Sticky PDMPs were introduced in [6] to target a distribution of the form µ(dx) ∝ exp(−ψ(x)) d � i=1 (dxi + 1 ci δ0(dxi)), which assigns strictly positive mass to events {xi = 0}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The sticky ZZS of [6] is obtained following the usual dynamics of the standard ZZS and in addition freezing the i-th component for a time τ ∼ Exp(ci) when xi hits zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The simulation of this process is challenging for the same reasons of the standard ZZS, since the two processes have the same switching rates λi for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The i-th component is either frozen, which is denoted by (xi, vi) ∈ Ai, or it evolves as given by the usual dynamics of ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The generator can then be decomposed as L = LD + LB where LD = �d i=1 LD,i and LB = �d i=1 LB,i, LD,if(x, v) = vi ∂ ∂xi f(x, v)1AC i (xi, vi) + ci(f(Ti(x, v)) − f(x, v))1Ai(xi, vi), LB,if(x, v) = λi(x, v)[f(x, Riv) − f(x, v)]1AC i (xi, vi), and Ti(x, v) corresponds to unfreezing the i-th component (we refer to [6] for a detailed description).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' An iteration of the scheme DBD in this case proceeds by a first half step of D, which is identical to the continuous sticky ZZS but with λi temporarily set to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence frozen components are unfrozen with rate ci and then start moving again, or unfrozen components move with their corresponding velocity vi and become frozen for a random time with rate ci if they hit xi = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then a full step of the usual bounce kernel B is done for the components which are not frozen, while for the frozen components, that is (xi, vi) ∈ Ai, the generator LB,i does nothing and so the velocity cannot be flipped.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' So unfreezing is not possible in this step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The iteration ends with another half step of D in a similar fashion to the previous one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' These ideas are more general than the two specific examples we considered and do not introduce further difficulties for our schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We observe that in these cases it might be useful to consider the process obtained with the splitting schemes as continuous time processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, notice that a Metropolis correction can be added following Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, and subsampling is possible following Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Convergence of the splitting scheme In this section we prove that under suitable conditions the splitting scheme DJD described in Section 1 is indeed a second order approximation of the original PDMP (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that in this section we have a PDMP defined on some arbitrary space E therefore it is not clear what it means to have a derivative, indeed we will typically be interested in the setting E = Rd × V for some set V which may be a discrete set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Instead of working with a full derivative we will define the directional derivative, DΦ, in the direction Φ as DΦg(z) = lim t→0 d dtg(ϕt(z)) SPLITTING SCHEMES FOR PDMPS 13 for any g ∈ C(E) for which t �→ g(ϕt(z)) is continuously differentiable in t for every z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note if E is a subset of Rd for some d and g is continuously differentiable then DΦg(z) = Φ(z)T ∇g(z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We extend this definition to multi-dimensional valued functions G : E → Rm by defining DΦG(z) = (DΦGi(z))m i=1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We define the space Ck,m Φ to be the set of all functions g : E → R which are k times continuously differentiable in the direction Φ with all derivatives Dℓ Φg(z) up to order k bounded by a polynomial of order m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We endow this space with the norm ∥g∥Ck,m Φ := sup z∈E |g(z)| + �k ℓ=1|Dℓ Φg(z)| 1 + |z|m .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us make the following assumptions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let Φ be a globally Lipschitz vector field defined on E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We assume that the directional derivative in the direction Φ is well-defined and that Φ be continuous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The switching rate λ : E → [0, ∞) is twice continuously differentiable in the direction Φ and λ, DΦλ, D2 Φλ grow at most polynomially.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We denote by mλ a constant such that ∥λ∥C 2,mλ Φ < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let Q be a probability kernel defined on E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall consider the operator Q : Cb(E) → Cb(E) defined by Qg(z) = � g(˜z)Q(z, d˜z), for any g ∈ Cb(E).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (14) Moreover we assume that Q has moments of all orders and Qg has at most polynomial growth of order m whenever g has at most polynomial growth of order m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For any m ∈ N, and g ∈ C1,m Φ we assume the following distribution is well-defined: (DΦQ)g(z) = DΦ(Qg)(z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (15) As an abuse of notation we shall write DΦQ also as a kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We assume for any m ∈ N, and g ∈ C1,m Φ |Qg(ϕs(z)) − Qg(z)| ≤ Cs(1 + |z|m)∥g∥C1,m Φ , (16) and also that there exists a constant C such that for any g ∈ C2,m Φ |Qg(ϕs(z)) − Qg(z) − sDΦQg(z)| ≤ Cs2(1 + |z|m)∥g∥C2,m Φ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (17) Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The closure (L, D(L)) of the operator (L, C1 c (E)) in L2 µ generates a C0-semigroup Pt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If g ∈ C2,0 Φ then we assume that Ptg is also twice continuously differentiable in the direction Φ and LPtg is continuously differentiable in the direction Φ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover we assume D2 ΦPtg and DΦLPtg are both polynomially bounded for finite t and for some C > 0, R ∈ R, mP ∈ N |DΦPtg(z)| + |D2 ΦPtg(z)| + |DΦLPtg(z)| ≤ C(1 + |z|mP)eRt∥g∥C2,0 Φ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let Ztk denote the approximation obtained by the splitting scheme DJD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assume that for each k, Ztk has moments of all orders and moreover for every M ∈ N there exists some GM such that sup m≤M Ez[|Ztk|m] ≤ GM(z).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 14 SPLITTING SCHEMES FOR PDMPS Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let Zt be a PDMP corresponding to the generator (1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assume that Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 to Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then there exist constants C, R such that for any g ∈ C2,0 Φ ∩ D(L) we have for some M ∈ N sup k≤n |E[g(Ztk)] − E[g(Ztk)]| ≤ CeRtnGM(z)δ3n∥g∥C2,0 Φ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof is adapted from [3, Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='24] and can be found in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Example 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7 (ZZS continued).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall the Zig-Zag sampler from Example 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 let us verify the Assump- tion 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 to 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 in this case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order to have a smooth switching rate we replace λi(x, v) by λi(x, v) = log (1 + exp(vi∂iψ(x))) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is shown to be a valid switching rate in [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We will assume that ψ ∈ C2 with bounded second and third derivatives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us now consider each assumption in turn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1: In this case Φ(x, v) = (v, 0)T which is clearly smooth and globally Lipschitz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2: Since λi is the composition of smooth maps and ψ we have that λi has the same smoothness in x as ψ and hence x �→ λi(x, v) is C2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As s �→ log(1+es) grows at most linearly, has first and second derivatives bounded by 1 we have that λi, ∇xλi and ∇2 xλi are all polynomially bounded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3: The proof of this can be found in Section A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4: By [1] we have that Pt is a strongly continuous semigroup on L2 µ with generator (L, D(L)) given as the closure of (L, C1 c (E)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover we have that the assumptions of [20, Theorem 17] are satisfied and hence Ptg(x, v) is differentiable in x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Following the proof of [20, Theorem 17] one also has |∇xPtg| ≤ C(1 + |x|m)eRt∥g∥C1,0 Φ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note here since DΦg(x, v) = vT ∇xg we have that Ck,0 Φ coincides with the space of continuous functions which are k-times continuously differentiable in the variable x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By the same arguments one can also obtain |∇2 xPtg| ≤ C(1 + |x|m)eRt∥g∥C2,0 Φ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5: This will be established in Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Ergodicity of splitting schemes of BPS and ZZS We shall now focus on results on ergodicity of splitting schemes of BPS and ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular we show existence of an invariant distribution, characterise the set of all invariant distributions, and establish convergence of the law of the process to such distributions with geometric rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order to prove this we rely on the following classical result, due to Meyn and Tweedie [32] (here the specific statement is based on [23, Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2], see also [19, Theorem S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7] for the explicit constants).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall the definition of V -norm: ∥µ∥V := sup|g|≤V |µ(g)|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider a Markov chain with transition kernel P on a set E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose that there exist constants ρ ∈ [0, 1), C, α > 0, a function V : E → [1, +∞) and a probability measure ν on E such that the two following conditions are verified: (1) Drift condition: for all x ∈ E, PV (x) ≤ ρV (x) + C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (18) (2) Local Dobelin condition: for all x ∈ E with V (x) ⩽ 4C/(1 − ρ), δxP ⩾ αν .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 15 Then, for all probability measures µ, µ′ on E and all n ∈ N, ∥µP n − µ′P n∥V ⩽ C α κn∥µ − µ′∥V (19) where κ = max(1 − α/2, (3 + ρ)/4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover P admits a unique stationary distribution µ∗ satisfying µ∗(V ) < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Under the Drift condition (18) alone, following the proof of [23, Theorem 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2] in the case α = 0 we get that for all probability measures µ, µ′ on E, ∥µP − µ′P∥V ⩽ (ρ + 2C)∥µ − µ′∥V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall now consider our splitting schemes and prove geometric ergodicity under suitable conditions by showing that the assumptions of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 are satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The splitting schemes of BPS and ZZS are respectively addressed in Theorems 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 below and, in both cases, the dependence of all constants in (19) on the step size is made explicit (statements with more details are postponed to Appendixes B and C).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' More precisely, in both cases, we obtain a local Doeblin (or minorisation) condition with constant α after n∗ = ⌈t∗/δ⌉ steps, where t∗ > 0 plays the role of physical time and n∗ is the number of steps needed to travel for an equivalent time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here t∗, α are independent of δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' On the other hand, we show that the drift condition holds for one step of the kernel with constants ρ = 1 − bδ and C = Dδ, where b, D and the Lyapunov function V are independent of δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This implies that for any s > 0 and any δ ∈ (0, δ0] (Pδ)⌈s/δ⌉V ⩽ (1 − bδ)⌈s/δ⌉ V + Dδ ⌈s/δ⌉−1 � k=0 (1 − bδ)k ⩽ e−bsV + D b .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Applying Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1, we get for P n∗ δ a long-time convergence estimate which is uniform over δ ∈ (0, δ0], that is for all δ ∈ (0, δ0] and n ≥ 1 we find ∥µ(P n∗ δ )n − µ′(P n∗ δ )n∥V ⩽ C′ α κn ∥µ − µ′∥V , where C′ = D/b and κ = max(1 − α/2, (3 + e−bt∗)/4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that the rhs does not depend on δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using the observation in Note 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, we can get convergence in V -norm for P n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed for n = mn∗ + r with r < n∗ we have ∥µP n δ − µ′P n δ ∥V = ∥µP mn∗+r δ − µ′P mn∗+r δ ∥V ⩽ C′ α κm∥µP r δ − µ′P r δ ∥V ⩽ C′ α � 1 + 2C′� κm∥µ − µ′∥V ⩽ C′′˜κnδ∥µ − µ′∥V , (20) where ˜κ = κ1/(t∗+δ0) ∈ (0, 1) and C′′ = C′ (1 + 2C′) /(ακ) are independent from δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we used that with computations identical to above we get the drift condition P r δ V ≤ (1−bδ)V +D(1−(1−bδ)r)/b ≤ V +C′, which is enough for the current purpose.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a conclusion, the estimates given in Theorems 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 below (or in Appendixes B and C for more details) give the expected dependency in δ for the convergence rate of the process toward equilibrium.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For splitting schemes of the BPS, we work under the following condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 16 SPLITTING SCHEMES FOR PDMPS Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The dimension is d ≥ 2, the velocity equilibrium ν is the uniform measure on Sd−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' There exists C > 0 such that 1 C |x|2 − C ⩽ ψ(x) ⩽ C|x|2 + C , 1 C |x| − C ⩽ |∇ψ(x)| ⩽ C|x| + C for all x ∈ Rd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, ∥∇2ψ∥∞ < ∞ and, without loss of generality, inf ψ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Notice that, when d = 1, the BPS and the ZZS coincide, in which case we refer to Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Our result of ergodicity for splitting schemes of the BPS is the following.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider any scheme of the BPS based on the decomposition D,R,B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Under Assump- tion 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3, there exist δ0, a, C′′ > 0, ˜κ ∈ (0, 1) and V : Rd × Sd−1 → [1, +∞) satisfying for all x ∈ Rd, v ∈ Sd−1, e|x|/a/a ⩽ V (x, v) ⩽ aea|x| such that, for all δ ∈ (0, δ0], Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 is applicable and (20) holds with these C′′, ˜κ, V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof can be found in Appendix B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ More care is required for the DBD scheme of the ZZS since this Markov chain has periodicity and is not irreducible, which is reminiscent of the discrete-space Zig-Zag chain studied in [35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us illustrate this behaviour by considering the one dimensional setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let (x, v) be the initial condition of the process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since v has magnitude 1, the position component x can only vary by multiples of the step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Thus for a fixed initial condition (x, v) the process remains on a grid (x + δZ) × {−1, 1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, after a single step of the scheme there are two possible outcomes: either the velocity does not change, in which case x moves to x + δv, or the velocity is flipped and the position remains the same.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This means that the change in the position (by amounts of δ) plus half the difference in the velocity always changes by ±1 each step and hence is equal to the number of steps in the scheme up to multiples of two, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Xnδ − x δ + 1 2(V nδ − v) ∈ n + 2Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a consequence, the chain lives on two disjoint sets depending on whether n is even or odd, which means that it is periodic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To overcome this issue, we consider the chain with one step transition kernel given by P 2 δ = PδPδ, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' we restrict to the case of an even number of steps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Markov chain with kernel P 2 δ is aperiodic, but it is not irreducible on Rd and hence has (infinitely) many invariant measures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order to characterise the invariant measures we restrict to the set in which the Markov kernel P 2 δ is irreducible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For fixed (x, v) ∈ Rd × {−1, 1}d we construct the grid which contains (x, v) as follows: D(x, v) := {(y, w) ∈ C × {±1}d : (yi, wi) ∈ D1(xi, vi) for all i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d}, (21) where D1(xi, vi) := D+(xi, vi) ∪ D−(xi, vi), with D+(xi, vi) := {(yi, wi) : wi = vi, yi = xi + mδ, m ∈ 2Z}, D−(xi, vi) := {(yi, wi) : wi = −vi, y = xi + mδ, m ∈ 2Z + 1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case we show in Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 that the Markov chain with transition kernel P 2 δ is irreducible on D(x, v), has a unique invariant measure, πx,v δ , and is geometrically ergodic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now we can characterise all the invariant measures of the Markov chain with transition kernel P 2 δ defined on Rd × {−1, 1}d as the closed convex hull of the set {πx,v δ : x ∈ Rd, v ∈ {−1, 1}d}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now consider the Markov chain with transition kernel P 2 δ on Rd × {−1, 1}d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For any initial distribution µ we have convergence of µP 2n δ to some measure πµ δ as n tends to ∞ and πµ δ is given by πµ δ (ϕ) = (µπx,v δ )(ϕ) := � Rd×{−1,1}d � Rd×{−1,1}d ϕ(y, w)πx,v δ (dy, dw)µ(dx, dv).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (22) SPLITTING SCHEMES FOR PDMPS 17 We use the next assumption to verify that Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 applies for initial conditions drawn from probability distributions with support on D(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider switching rates λi(x, v) = (vi∂iψ(x))++γi(x) for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ψ ∈ C2(Rd) and the following conditions hold: (a) The switching rates λi(x, v) are such that there exist x0 ≥ 0 such that for all x1 > x0 λ(x1) := min i=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=',d min (x,v): xivi∈[x0,x1], |xj|∈[x0,x1] for all j̸=i λi(x, v) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) For |x| ≥ R for some R > 0 sup t∈(0,1),y1,y2∈B(x,t √ d),v,w∈{−1,1}d e(t2(|(v+w)T ∇2ψ(y1))i|+2t|(w∇2ψ(y2))i|)γi(x + vt)etvi∂iψ(x) ≤ γ0 < 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (23) (c) Denote as B(x, δ √ d) the ball with centre at x and radius δ √ d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then lim ∥x∥→∞ sup y1,y2∈B(x,δ √ d) max{1, ∥∇2ψ(y1)∥} |∂iψ(y2)| = 0 for all 0 ≤ δ ≤ δ0, i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d, where δ0 = 2(1 + γ0)−1, for γ0 as in part (b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Part (a) in Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 is inspired by [7, Assumption 3] and is used to show that a minorisation condition holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This condition is either a consequence of properties of the target, or else can be enforced by taking a non-negative excess switching rate, in which case γi(x) can be chosen to be a continuous function γi : Rd → (0, ∞).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In principle one could prove a minorisation condition using the techniques of [8], but this is beyond the scope of this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Part (b) is a condition on the decay of the refreshment rate, while Part (c) is similar to Growth Condition 3 in [8] and is satisfied for instance if ψ is strongly convex with globally Lipschitz gradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' These two conditions are used to show that a drift condition holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting scheme DBD for ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then there exist C′′, δ0 > 0, ˜κ ∈ (0, 1) and V : Rd × {−1, 1}d → [1, ∞) satisfying for all (x, v) ∈ Rd × {−1, 1}d , d � i=1 (1 + 2|∂iψ(x)|)− 1 2 ≤ V (x, v) exp(βψ(x)) ≤ d � i=1 (1 + 2|∂iψ(x)|) 1 2 for all β ∈ (0, 1/2) such that, for all δ ∈ (0, δ0], the following holds: (1) Fix (x, v) ∈ Rd × {−1, 1}d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 is applicable to P 2 δ = PδPδ seen as a transition kernel on D(x, v), and the inequality (20) holds (with Pδ replaced by P 2 δ ) with these C′′, ˜κ, V for any µ, µ′ having support on D(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (2) For any probability measure µ on Rd × {−1, 1}d with µ(V ) < ∞, we have that µP 2n δ converges as n → ∞ to the measure πµ δ given by (22) where πx,v δ is the unique invariant measure of P 2 δ on D(x, v) and we have ∥µP 2n δ − µπx,v δ ∥V ≤ C′′˜κnδ � ∥δ(x,v) − πx,v δ ∥V µ(dx, dv) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (24) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof can be found in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Under similar assumptions we establish geometric ergodicity of schemes DRBRD, RDBDR of ZZS, where the switching rates in the B part are λi(x, v) = (vi∂iψ(x))+, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the canonical rates, while refreshments in the R part are independent draws from Unif({±1}d) with rate γ(x) : Rd → [0, ∞).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The rigorous statement of this result, Theorem C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6, and its proof can be found in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 18 SPLITTING SCHEMES FOR PDMPS Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Empirical error for the radius statistic t(x) = x2 with a one-dimensional standard Gaussian target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The step size is set to δ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0, the number of iterations is N = 105, and the experiment is repeated 250 times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The schemes BDB (left) and DBD (right) correspond to including the refreshment part in B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In schemes B DR B (left) and DR B DR (right) we denote by B the standard bounce part, by DR the transition kernel which corresponds to having refreshments and deterministic motion together, and we use underscores to divide these two kernels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here ν is the uniform distribution on {±1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Expansion of the invariant measure of splitting schemes for BPS In this section we investigate the bias in the invariant measure of different splittings of BPS and draw conclusions on which schemes perform best.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Motivated by Theorems 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4, we assume that the processes corresponding to our splitting schemes have an invariant distribution with density µδ(x, v) = µ(x, v)(1 − δ2f2(x, v) + O(δ4)), (25) where µ(x, v) = ν(v)π(x), π is the target and ν is a distribution satisfying Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 below, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the uniform distribution on the unit sphere or the standard Gaussian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is then our goal to compute and compare f2 for different schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' There are several splitting schemes that could be compared, and thus we make a selection of the ones it is worth focusing on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The numerical simulations shown in Figure 1 give an idea of the relative performance of the schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The plots show that the schemes that have DBD as their limit as the refreshment rate goes to zero have a smaller bias in the x component compared to those that converge to BDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Naturally the difference between the two schemes is expected to vanish as δ → 0 and also appears to be diminishing as the dimension increases (see Figure 4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Based on this result we decide to concentrate on schemes RDBDR, DBRBD, DRBRD, as well as BDRDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that all these schemes have the same cost of one gradient computation per iteration (in BDRDB it is sufficient to keep track of the gradient at the previous iteration).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Following the approach of [26] , we will show in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 (more precisely in Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5) that the second order of the bias f2 can be computed analytically for one-dimensional targets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We then focus on the dependence of f2 on the refreshment rate, which is the only parameter of the algorithm (outside of δ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As we will see, and as already hinted by Figure 1, some splittings like RDBDR and BDRDB are robust to poor choices of the refreshment rate, while others like DBRBD and DRBRD have linear or quadratic dependence on λr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The numerical experiments of Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 confirm the theoretical results of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and suggest that the bias behaves similarly in higher dimensions, where obtaining f2 analytically is very challenging.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, in Figure 3 we show that, in the cases we consider, splitting RDBDR is the scheme that shows the best overall behaviour.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 19 This scheme was shown to be unbiased for standard Gaussian targets in Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, and is confirmed to have f2 = 0 in such cases in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, we fully characterise the invariant distribution of RDBDR in one dimension in Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section 3 we will see cases where a splitting scheme may admit more than one invariant measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In such cases it is not immediately clear what the expansion (25) means.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order to make (25) consistent as δ → 0, in those cases we consider µδ as the limit of the law of the splitting scheme as the number of steps tends to infinity when the process is started according to µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Computing f2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us discuss briefly how to find f2 with the approach of [26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using the Baker-Campbell-Hausdorff (BCH) formula (see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [9]) we can find L2 such that Ex,v[f(Xδ, V δ)] = f(x, v) + δLf(x, v) + δ3L2f(x, v) + O(δ4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here L is the infinitesimal generator of the continuous time process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Integrating both sides with respect to µδ and using that µδ is an invariant measure for the splitting scheme we obtain � f(x, v)µδ(x, v)dxdv = � f(x, v)µδ(x, v)dxdv + δ � Lf(x, v)µδ(x, v)dxdv + δ3 � L2f(x, v)µδ(x, v)dxdv + O(δ4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Substituting for µδ with the expansion (25) we have 0 = δ � Lf(x, v)µ(x, v)dxdv − δ3 � Lf(x, v)µ(x, v)f2(x, v)dxdv + δ3 � L2f(x, v)µ(x, v)dxdv + O(δ4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since µ is an invariant measure for BPS we have � Lfp dxdv = 0 which gives the equation L∗(µf2) = L∗ 2µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (26) Here L∗ and L∗ 2 are the adjoints on L and L2 in L2 with respect to Lebesgue measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since there is not a unique solution to the equation (26) we need to impose a compatibility condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since both ˆµ and µ are probability densities, integrating (25) gives the requirement � f2(x, v)µ(x, v)dxdv = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (27) It is then the goal of this section to solve (26) and compare the solutions corresponding to the different splitting schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We start by computing the term L∗ 2 using the BCH formula.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall that the adjoint of the generator of BPS is given by L∗g(x, v) = −⟨v, ∇xg(x, v)⟩ + ((gλ1)(x, R(x)v) − (gλ1)(x, v)) + λr � ν(v) � g(x, y)dy − g(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We now compute L∗ 2 for the splitting schemes DBRBD, RDBDR, DRBRD, BDRDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us start with an assumption on the invariant distribution of the velocity vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The invariant measure for the velocity component ν satisfies the following condi- tions: (1) Invariance under rotations: ν(w) = ν(v) for any v, w such that |v| = |w|;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (2) Mean zero: Eν[V ] = 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (3) Isotropic: for some b > 0 it holds that Covν(V ) = bI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' These properties hold for instance if ν is the standard Gaussian distribution, as well as if ν is the uniform on the unit sphere (in that case b = 1/d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 20 SPLITTING SCHEMES FOR PDMPS Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 hold and define A(x, v) = 3 2λr � b tr � ∇ψ(x)∇ψ(x)T − ∇2ψ(x) � + 2⟨v, ∇ψ(x)⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩ � , B(x, v) = 3 2λ1(x, R(x)v) � ⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩ � + 1 2⟨v, ∇x(⟨v, ∇2ψ(x)v⟩)⟩, C(x, v) = 3λ1(x, R(x)v) � − 2⟨v, ∇2ψ(x)v⟩ + ⟨v, ∇ψ(x)⟩2� − ⟨v, ∇(⟨v, ∇2ψ(x)v⟩)⟩, D(x, v) = 3 2λr � b tr � ∇ψ(x)∇ψ(x)T −∇2ψ(x) � +⟨v, ∇2ψ(x)v⟩ + ⟨v, ∇ψ(x)⟩ � 3λ1(x, R(x)v) + λ1(x, v) �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The splitting scheme DBRBD satisfies L∗ 2µ(x, v) = µ(x, v) 12 � A(x, v) + B(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The splitting scheme RDBDR satisfies L∗ 2µ(x, v) = µ(x, v) 12 B(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The splitting scheme DRBRD satisfies L∗ 2µ(x, v) = µ(x, v) 12 � D(x, v) + B(x, v) + 3 2λ2 r⟨v, ∇ψ(x)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The splitting scheme BDRDB satisfies L∗ 2µ(x, v) = µ(x, v) 12 � − A(x, v) + C(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof can be found in Appendix E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Note 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Clearly, if L∗ 2µ = 0 then f2 must be a constant that satisfies (27) and hence it must be that f2 = 0, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the second order term in µδ is zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is the case for instance for scheme RDBDR when the target is a multidimensional standard Gaussian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed in Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 we proved that RDBDR is unbiased for standard Gaussian targets, thus this is a consistent result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the same setting, we observe that f2 = 0 for schemes DBRBD and DRBRD when the refreshment rate is λr = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is an expected result, as when λr = 0 these schemes coincide with RDBDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Figure 2 we confirm that, for a one-dimensional standard Gaussian and when λr = 0, the scheme DBD is unbiased, while the scheme BDB is of second order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Equation (26) is in general hard to solve, as the adjoint of BPS contains both derivatives and integrals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Nonetheless, we are able to solve (26) and find f2 in the one-dimensional case, as stated in the next Proposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the one-dimensional setting with state space R × {±1} and target distri- bution µ(x, v) = π(x)ν(v) with π ∝ exp(−ψ) and ν = Unif({±1}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let λr ≥ 0 be the refreshment rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then the function f2 that solves (26) is f2(x, +1) = f+ 2 (0) + � x 0 ��λr 2 + (−∂ψ(y))+ � g(y) − L∗ 2µ(y, +1) µ(y, +1) � dy, f2(x, −1) = f2(x, +1) + g(x), where g(x) = exp (ψ(x)) � x −∞ �L∗ 2µ(y, +1) µ(y, +1) + L∗ 2µ(y, −1) µ(y, −1) � exp(−ψ(y))dy, f+ 2 (0) = − � ∞ −∞ �g(x) 2 + � x 0 ��λr 2 + (−∂ψ(y))+ � g(y) − L∗ 2µ(y, +1) µ(y, +1) � dy � π(x)dx.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 21 Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Error for the radius statistic for a one-dimensional standard Gaussian tar- get.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here λr = 0 for both schemes DBD and BDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The dashed, blue line corresponds to second order convergence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The time horizon is fixed to T = 105 and the number of iterations is N = T/δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The experiment is repeated 250 times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof can be found in Appendix D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Note 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' An immediate consequence of Propositions 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 is that in the one dimensional case the second order term of the bias of scheme RDBDR is always independent of the refreshment rate and of v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed applying the propositions we find f2(x, v) = f+ 2 (0) − 1 24 � x 0 ψ(3)(y)dy (28) with f+ 2 (0) = 1 24 � ∞ −∞ � x 0 ψ(3)(y)dyπ(dx).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In fact, in 1D, for the scheme RDBDR, we can get an explicit expression for the invariant measure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the scheme RDBDR for BPS in one dimension, where the velocity is refreshed from ν = Unif({±1}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then for a fixed initial condition x ∈ R and step size δ the distribution with support on {y ∈ R : y = x + nδ, n ∈ Z} × {±1} given by µδ(y, v) ∝ e−ψδ(y) where ψδ(x) = ψ(x) and for y = x + nvδ, n ∈ N ψδ(y) = ψ(x) + δ n � ℓ=1 ψ′(x + (ℓ − 1/2)vδ) is stationary for the process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, under the conditions of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 we obtain that µδ is ergodic, in the sense that for all bounded functions lim N→∞ 1 N N � n=1 f(Xtn, V tn) = µδ(f) Px,v − a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof can be found in Appendix D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Once again it is clear that the scheme is unbiased in the Gaussian case ψ(x) = x2/(2σ2) (in the sense that ψδ(y) = ψ(y) for all y = x + nδv, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the BPS is ergodic with respect to the restriction of the true Gaussian target to the grid, and moreover the target measure is invariant for the scheme).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 22 SPLITTING SCHEMES FOR PDMPS More generally, for y = x + vnδ with n ∈ N we get ψ(y) = ψ(x) + n � ℓ=1 � δ/2 −δ/2 ψ′(x + v(ℓ − 1/2)δ + u)du = ψδ(y) + 1 2 n � ℓ=1 � δ/2 −δ/2 u2ψ(3)(x + v(ℓ − 1/2)δ)du + O(nδ5) = ψδ(y) + δ2 24 � y x ψ(3)(u)du + O(δ4|x − y|) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Setting x = 0 this gives ψδ = ψ + δ2f2 + O(δ4) with f2(y) = � y 0 ψ(3)(u)du, which is the same of Equation (28).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed the term f+ 2 (0) in (28) was introduced to make exp(−ψ)(1+δ2f2) a probability distribution and would appear also in the present context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence Propositions 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7 agree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Application to three one-dimensional target distributions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this section we compare the splitting schemes by applying Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 to three one-dimensional target distributions: a centred Gaussian distribution, a distribution with non-Lipschitz potential ψ(x) = x4, and a Cauchy distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The formal statements can be found in the Appendix D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3, correspondingly in Propositions D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here instead of giving the complicated analytic expressions for f2 in all cases, we give plots of the TV distance between µ and µδ as a function of λr as given by Propositions D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The results, both according to the theory and numerical simulations, are shown in Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us briefly explain how the TV distance is derived from the analytic expression of f2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall focus on the position part of µδ, which we denote as πδ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By marginalising and recalling in this context ν = Unif({±1}) we obtain πδ(x) = π(x) � 1 − δ2 2 (f2(x, +1) + f2(x, −1)) � + O(δ4), (29) Using (29) we can express the TV distance between π and πδ as ∥π − πδ∥TV = δ2 2 sup A ���� � A (f2(x, +1) + f2(x, −1))π(x)dx ���� + O(δ4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (30) The δ2 contribution of the rhs can be computed by plugging in the expressions for f2 found in Propo- sitions D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We neglect higher order terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us comment on these results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First of all, the theoretical results are consistent with the numerical experiments of Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed, it is clear that the schemes RDBDR and BDRDB have a bias that is independent of the refreshment rate, while DBRBD and DRBRD have respectively linear and quadratic dependence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the one-dimensional case, the plots show that it is best to choose λr = 0, which is possible as in this case BPS is irreducible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' However, in higher dimensional settings it is necessary to take λr > 0 as shown in Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since choosing a good value of λr is difficult and depends on the target distribution, it is desirable to use schemes that have good performance for most values of λr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, it is clear from Figure 3 that RDBDR is indeed unbiased in the Gaussian case, and also has the smallest bias out of all the considered splittings with the exception of the Cauchy target, where the difference in performance between RDBDR and BDRDB is almost negligible and seems to slightly favour the latter in experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case, we also see a small dependence on λr for RDBDR and BDRDB, which could be due to higher order terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The experiments in Figure 4 suggest that the findings of the one-dimensional case extend to multi- dimensional targets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, RDBDR has either a better performance than other splittings or behaves very similarly to BDRDB both on an independent as well as a correlated Gaussian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, the independence on λr of the bias of schemes DBRBD and DRBRD is confirmed also when d > 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 23 (a) TV distance according to Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 (b) Absolute value of the error for the radius statistic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (c) TV distance according to Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (d) Absolute value of the error for the radius statistic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (e) TV distance according to Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (f) Absolute value of the error for min{4, x2}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Total variation distance to the true target as given by the second order term in (30) (left) and numerical simulations (right) for the various splittings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The top row is obtained with standard Gaussian target, the middle row with ψ(x) = x4, and the bottom row with a one dimensional Cauchy target with γ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In all plots δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and the number of iterations is N = 2 · 105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the Gaussian and Cauchy cases we initialise the processes at µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 24 SPLITTING SCHEMES FOR PDMPS (a) d = 2, ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) d = 10, ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (c) d = 2, ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (d) d = 10, ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Error of estimators of the radius with splitting schemes for BPS with a Gaussian target with covariance Σii = 1, Σij = ρ for i ̸= j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The step size is δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and the number of iterations is N = 2 · 105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The processes are initialised with a draw from µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In conclusion, we have conducted a detailed analysis of the bias in the invariant measure, both theoretical in Propositions D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3, and empirical in Figures 1, 2, 3, 4, and the evidence suggests that RDBDR is the best candidate out of the pool of splitting schemes that are available.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The closest competitor BDRDB shows similar performance in some settings, but a larger bias in others in which RDBDR enjoys desirable properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Numerical experiments In this section we discuss some numerical simulations for the proposed samplers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The codes for all these experiments can be found at https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='com/andreabertazzi/splittingschemes_PDMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Gaussian target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we study the behaviour of the proposed algorithms on two types of Gaussian targets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The first type is a correlated Gaussian, for which the covariance matrix has unitary variances and correlation ρ between all components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The second type is an independent Gaussian, where components i ≥ 2 have unitary variance, while the first component has (small) variance σ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We study the performance of our algorithms as a function of ρ and σ2, as well as of the step size δ and the dimension of the target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular we first focus on the number of rejections in the Metropolised algorithms, that is Algorithms 3 and 4, and then we focus on the error in the estimation of the expected radius for all our algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 25 (a) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (c) Here ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (d) Here σ2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (e) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (f) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and σ2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Fraction of rejected proposals in the Metropolis step for Algorithms 3 and 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The plots on the left are obtained running the algorithms with a Gaussian target with covariance Σii = 1, Σij = ρ for j ̸= i, while the plots on the right with diagonal covariance Σ11 = σ2, Σii = 1 for i ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In all experiments the refreshment rate for BPS is λr = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 5 shows the number of rejections in adjusted algorithms, with the left part of the plot showing the first type of Gaussian target and the right part showing the second.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This experiment allows us to understand the efficiency of the Metropolis adjusted algorithms, as a larger fraction of rejections corresponds to more computations required to obtain an accepted state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' What we observe is that the adjusted ZZS defined in Algorithm 3 is exact for targets with diagonal covariance as expected, but the number of rejections increases with the correlation between components of the target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is well known 26 SPLITTING SCHEMES FOR PDMPS (a) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (c) Here ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (d) Here σ2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 and d = 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (e) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and ρ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (f) Here δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 and σ2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Error for the radius statistic for Algorithms 1, 2, 3, 4, as well as the continuous ZZS and BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The plots on the left are obtained running the algorithms with a Gaussian target with covariance Σii = 1, Σij = ρ for j ̸= i, while the plots on the right with diagonal covariance Σ11 = σ2, Σii = 1 for i ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In all experiments the refreshment rate for BPS is λr = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The processes are started from a draw of the target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The time horizon is T = 103 and the number of iterations is N = ⌈T/δ⌉.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The radius is estimated with the usual Monte Carlo averages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 27 that the continuous time ZZS has lower efficiency for correlated targets (see [2]), and in the case of Algorithm 3 this is seen as a large number of reflections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' On the other hand, the adjusted BPS given by Algorithm 4 appears to suffer when σ2 is small, while the number of rejections remains controlled for large correlation ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 6 shows the error in the estimation of the expected radius for the adjusted and unadjusted algorithms, as well as for the continuous BPS and ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As expected, ZZS is sensitive to high correlation between components and its error increases with ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is possible to improve in these cases by applying the adaptive schemes proposed in [2], which learn the covariance structure of the target and use this information to tune the set of velocities of the ZZS suitably.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It seems also clear that the schemes based on ZZS are more robust when the target is very narrow in some components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is a reasonable behaviour, as DBD schemes for ZZS essentially decompose the target in one dimensional problems, hence the chain can explore efficiently some components while being stuck in others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a consequence, in the second type of Gaussian target the chain will rarely move in the component with small variance, but it can freely move in the other components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' On the other hand, in BPS the switching rate and reflection operator are dominated by the component with small variance, thus the whole chain is affected by settings with e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' small variances of some components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We observe that the adjusted BPS given by Algorithm 4 is more robust than its unadjusted counterpart.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For these reasons, Algorithms 1, 3, and 4 are to be preferred in case of stiff targets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Image deconvolution using a total variation prior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this section we test Algorithm 3 on an imaging inverse problem, which we solve with a Bayesian approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the following we shall refer to an image either as a N × N matrix or as a vector of length N2, which is obtained by placing each column of the matrix below the previous one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We set d = N2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In both cases each entry corresponds to a pixel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now denote as x ∈ Rd the image we are interested in estimating and y ∈ Rd the observation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The observation is related to x via the statistical model y = Ax + ξ, where A is a d × d-dimensional matrix which may be degenerate and ill-conditioned and ξ a d- dimensional Gaussian random variable with mean zero and variance σ2Id.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The forward problem we consider is given by a blurring operator, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' A acts by a discrete convolution with a kernel h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In our examples h will be a uniform blur operator with blur length either 9 or 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The likelihood of y given x is given by p(y|x) ∝ e−fy(x), fy(x) = 1 2σ2 ∥Ax − y∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the Bayesian approach one then has to place a prior distribution on x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we choose the total variation prior: p(x) ∝ e−g(x), where θ > 0 and g(x) = θ∥x∥TV is the total variation of the image x (see [41]) and is given by ∥x∥TV := N � i,j=1 (|xi+1,j − xi,j| + |xi,j+1 − xi,j|).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The total variation prior corresponds to the ℓ1-norm of the discrete gradient of the image and therefore promotes piecewise constant reconstructions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that this prior is not smooth and hence we cannot directly apply the gradient based algorithms such as the unadjusted Langevin algorithm (ULA), BPS or ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore we approximate g with a Moreau-Yosida envelope gλ(x) = min z∈Rd � g(z) + 1 2λ∥x − z∥2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 28 SPLITTING SCHEMES FOR PDMPS By [40, Proposition 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='19] we have that gλ is Lipschitz differentiable with Lipschitz constant λ−1 and ∇gλ(x) = 1 λ(x − proxλ g(x)), proxλ g(x) = arg min z∈Rd � g(z) + 1 2λ∥x − z∥2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using Bayes theorem, we have the posterior distribution π(x) := p(x|y) ∝ e−fy(x)−θgλ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (31) We select the optimal θ by using the SAPG algorithm [44, 17] and we choose λ based on the guidelines given in [22], which set λ = 1/Lf where Lf is the Lipschitz constant of fy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Sampling from this model using MCMC schemes is difficult because x is usually very high dimensional and the problem is ill- conditioned.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case the unadjusted Langevin algorithm can be very expensive to run since the step size is limited by 2/L, where L = Lf + λ−1 is the Lipschitz constant of ∇ log π.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that we do not consider an unadjusted underdamped Langevin algorithm since this algorithm scales poorly (see [11, 21]) with the conditioning number which is very large in these examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We are now interested in drawing samples from the posterior (31), and in particular we compare the unadjusted ZZS (Algorithm 1, abbreviated as UZZS in the plots)), the unadjusted Langevin algorithm (ULA), as well as the continuous ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed, we can compute the Lipschitz constant of the gradient of the negative log-posterior, L, and thus we can implement the exact ZZS using the Poisson thinning technique based on the simple bound λi(x + vt, v) ≤ tL √ d + λi(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (32) In order to compare the computational cost of the continuous ZZS to the unadjusted ZZS and to ULA we count each proposal for an event time obtained by Poisson thinning as a gradient evaluation and thus as an iteration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed, an update of the computational bounds requires the evaluation of λi(x, v) for all i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , d and thus the full gradient has to be computed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To estimate the posterior mean for the continuous ZZS we compute the time average T −1 � T 0 Xtdt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Figures 7 and 9 we show the original images, the observed images after blurring and adding noise, and the estimated posterior mean using the different samplers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 8 shows the mean square error (MSE) between the true image and the estimated posterior mean as a function of the number of iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The MSE is computed for two images x, y ∈ [0, 1]N×N as MSE(x, y) = 1 N2 N � i=1 N � j=1 (xij − yij)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (33) It is clear in both Figures 7 and 9 that the unadjusted ZZS shows the fastest convergence to the posterior mean.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is clear from visual inspection of the reconstructed images, as the reconstruction of ZZS are less noisy after just a few thousand iterations, and even more evident from the MSE shown in Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the case of Figure 7 it appears ZZS has essentially converged after 105 iterations, while it takes around 4 × 105 iterations for ULA to obtain a comparable approximation of the posterior mean.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is likely due to the fact that the step size of ULA must be very small or otherwise the process tends to infinity, while for ZZS larger step sizes can be selected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For instance in the context of Figure 7 the step size for ULA is approximately 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='9×10−6, which is the largest available value without going over the stability barrier, while for ZZS the step size is 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='9 × 10−3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This constitutes a major difference because every iteration is very computationally intensive, as the target distribution e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for the context of Figure 9 is of dimension 65536.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Notably each iteration involves solving an optimisation problem, which is solved by the SAPG algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' A similar behaviour is observed for the cameraman image shown in Figure 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case the unadjusted ZZS needs around 6×104 gradient evaluations to converge, while ULA still has not achieved the same accuracy after 2 × 105 iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This difference SPLITTING SCHEMES FOR PDMPS 29 Figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Results for the reconstruction of one of the MNIST handwritten digits using a TV prior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The observed image is obtained with a blur length of 9 pixels and then adding Gaussian noise with standard deviation σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The step size for ULA is 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='98/L, for ZZS is 3000/L, where L ≈ 1042855.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Mean after 2×103 iterations (second column) and after 106 iterations (third column) of the samplers based on the states at iterations n × 103 for n = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , 100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Mean square errors as defined in (33) for the setting of Figure 7 (left) and of Figure 9 (right).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' can also be seen from the reconstructions after 2 × 103 iterations shown in Figure 9, as indeed the unadjusted ZZS gives a clearly better estimate for the posterior mean.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, let us compare the unadjusted ZZS with the continuous time ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is clear from our experiments that ZZS performs poorly compared to its discretisation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The reason is twofold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First, the major drawback of Poisson thinning using the bounds (32) is that a considerable proportion of the proposed event times are rejected (in our examples the rejection rate is around 70 − 80%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Moreover, the rates λi are very large in the current framework and the process can have even 109 switches per continuous time unit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This means that many gradient computations are required to travel a decent distance and thus the 30 SPLITTING SCHEMES FOR PDMPS Figure 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Results for the reconstruction of full 256 × 256 pixels cameraman image using a TV prior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The observed image is obtained with a blur length of 25 pixels and then adding Gaussian noise with standard deviation σ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The reconstructions show the estimated mean after 2 × 103 iterations (top row) and after 2 × 105 iterations (bottom row) of the samplers based on the states at iterations n×103 for n = 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , 200.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The step size for ULA is 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='98/L, for ZZS is 1000/L, where L ≈ 518349.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' process itself is expensive to run.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The combination of these two phenomena implies an important loss of efficiency, which explains the results of our simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Chain of interacting particles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, let us consider a problem which will serve as an illustration of a typical context where ZZS is favored with respect to other samplers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is a toy model that presents in a simpler form features which are similar to the molecular system considered in [37], where splitting schemes involving velocity bounces have proven efficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We consider a chain of N particles in 1D, labeled from 1 to N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The particles interact through two potentials: a chain interaction, where the particle i interacts with the particles i − 1 and i + 1;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' and a mean-field interaction, where each particle interacts with all the others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For x ∈ RN, the potential is thus of the form ψ(x) = N−1 � i=1 V (xi − xi+1) + a N N � i,j=1 W(xi − xj) , where a > 0 measures the strength of the mean-field interaction, V is the chain potential and W is the mean-field potential.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the following we take V (s) = s4, W(s) = − � 1 + s2 , for s ∈ R, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' the chain interaction is an anharmonic quartic potential which constrains two consecutive particles in the chain to stay close, while the mean-field interaction induces a repulsion from the rest SPLITTING SCHEMES FOR PDMPS 31 of the system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Although this specific ψ is an academic example meant for illustration purpose, its general form is classical in statistical physics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Notice that ψ is invariant by translation of the whole system, so that e−ψ is not integrable on RN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' However we are not interested in the behavior of the barycentre ¯x = 1 N �N i=1 xi, so we consider e−ψ as a probability density on the subspace {x ∈ RN, ¯x = 0}, which amounts to looking at the system of particles from its center of mass.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Anyway, in practice, we run particles in RN without constraining their barycentre to zero, which does not change the output as long as we estimate the expectations of translation-invariant functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Specifically, here, we consider the empirical variance of the system v(x) = 1 N2 N � i,j=1 (xi − xj)2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The important points concerning this model (which are typically met in real molecular dynamics as in [37]) are the following.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The forces ∇ψ can be decomposed in two parts, one of which (the chain interaction) is unbounded and not globally Lipschitz but is relatively cheap to compute (with a complexity O(N)), while the second part (the mean-field interaction) is bounded but numerically expensive (with a complexity O(N2)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If this decomposition is not taken into account, so that we simply run a classical MCMC sampler based on the computation of ∇ψ, then the step size has to be very small because of the non-Lipschitz part of the forces, and then each step is very costly because of the mean-field force.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Besides, due to the non-Lipschitz part, sampling a continuous-time PDMP via thinning would not be very efficient (in fact in this specific simple case it could be possible to design a suitable thinning procedure with some effort, but this would be more difficult with 3D particles and singular potentials such as the Lennard-Jones one [37]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now, as was already discussed in Section 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 for subsampling, PDMPs and their splitting schemes can be used with a splitting of the forces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the present case, we consider a ZZS where the switching rate of the i-th velocity is given by λi(x, v) = � vi(V ′(xi − xi+1) − V ′(xi−1 − xi)) � + + a N � j̸=i � viW ′(xi − xj) � + , where, for the particles 1 and N, we set x0 = x1 and xN+1 = xN to cancel out the corresponding terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The corresponding continuous-time ZZS has the correct invariant measure (once centered).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We consider the DBD splitting to approximate this ZZS (although several other choices are possible, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' including the Poisson thinning part in the D step and having only jumps according to the potential V in the B part).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To sample the jump times of the i-th velocity, using that |W ′(s)| ⩽ 1 for all s ∈ R, we sample two jump times with rates respectively (vi(V ′(xi − xi+1) − V ′(xi−1 − xi)))+ and a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If both times are larger than the step size δ, then the velocity is not flipped.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Else, if the time corresponding to the first rate is smaller than δ and than that corresponding to the second, then we flip the i-th velocity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Alternatively, if the second time is smaller than δ and than the first, we draw J ∼ Unif({1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , N}) and we flip the sign of the i-th velocity with probability (viW ′(xi−xJ))+ a (note that if J = i then this probability is indeed 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since in this case the rates are not canonical due to the splitting of forces, this procedure is repeated until there are no events before the end of the time step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This results in O(1) computations per particle on average, hence O(N) for the whole system.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We compare this scheme with an HMC sampler implemented in the Julia package [46].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This contains state of the art techniques and implementation of HMC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The results of our simulations are shown in Figure 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Initially, particles are i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' standard Gaussian variables.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This gives a configuration which is far from the modes of the target distribution, where particles are organized so that i �→ xi is close to the monotonous profile which minimizes ψ (by the symmetry i ←→ N + 1 − i there are two such modes, but the empirical variance is unchanged by this symmetry so that it is sufficient to see one of them).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For both ZZS and HMC, the convergence 32 SPLITTING SCHEMES FOR PDMPS (a) N = 25, a = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='01.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) N = 25, a = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (c) N = 25, a = 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (d) N = 50, a = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (e) N = 100, a = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Empirical variance (on the y-axis) and runtime in seconds (on the x- axis) for various values of N and a in the setting of Section 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The green line corresponds to the unadjusted ZZS, the red line corresponds to HMC, and the dashed black line corresponds to the estimated empirical variance with a long run of HMC (when computationally feasible).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' of the estimator is thus essentially driven by a deterministic motion from the biased initial condition towards a mode.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is clear that our algorithm gives considerably cheaper yet accurate estimates of the empirical variance for all values of a and N considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is the result of the subsampling procedure, which reduces the cost per iteration from O(N2) to O(N), whereas in each iteration of HMC the full mean-field interaction needs to be computed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Notably, as expected the gain in performance increases with the number of particles N, which makes the required runtime of HMC prohibitive for large values of N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Funding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' AB acknowledges funding from the Dutch Research Council (NWO) as part of the research programme ‘Zigzagging through computational barriers’ with project number 016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='Vidi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='189.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='043.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' PD acknowledges funding from the Engineering and Physical Sciences Research Council (EPSRC) grant EP/V006177/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' PM acknowledges funding from the French ANR grant SWIDIMS (ANR-20-CE40- 0022) and from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement No 810367), project EMC2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' References [1] Christophe Andrieu and Samuel Livingstone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Peskun–tierney ordering for markovian monte carlo: beyond the reversible scenario.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Annals of Statistics, 49(4):1958–1981, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [2] Andrea Bertazzi and Joris Bierkens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Adaptive schemes for piecewise deterministic Monte Carlo algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Bernoulli, 28(4):2404 – 2430, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [3] Andrea Bertazzi, Joris Bierkens, and Paul Dobson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Approximations of piecewise deterministic markov processes and their convergence properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Stochastic Processes and their Applications, 154:91–153, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [4] Joris Bierkens, Alexandre Bouchard-Cˆot´e, Arnaud Doucet, Andrew B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Duncan, Paul Fearnhead, Thibaut Lienart, Gareth Roberts, and Sebastian J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Vollmer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Piecewise deterministic markov processes for scalable monte carlo on restricted domains.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Statistics & Probability Letters, 136:148–154, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The role of Statistics in the era of big data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 33 [5] Joris Bierkens, Paul Fearnhead, and Gareth Roberts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The zig-zag process and super-efficient sampling for bayesian analysis of big data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Annals of Statistics, 47, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [6] Joris Bierkens, Sebastiano Grazzi, Frank van der Meulen, and Moritz Schauer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Sticky pdmp samplers for sparse and local inference problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Statistics and Computing, 33(1):8, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [7] Joris Bierkens and Gareth Roberts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' A piecewise deterministic scaling limit of lifted metropolis–hastings in the curie–weiss model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Annals of Applied Probability, 27(2):846–882, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [8] Joris Bierkens, Gareth O Roberts, and Pierre-Andr´e Zitt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Ergodicity of the zigzag process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Annals of Applied Probability, 29(4):2266–2301, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [9] Andrea Bonfiglioli and Roberta Fulci.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Topics in noncommutative Algebra: The Theorem of Campbell, Baker, Haus- dorff and Dynkin, volume 2034.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Springer, 01 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [10] Alexandre Bouchard-Cˆot´e, Sebastian J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Vollmer, and Arnaud Doucet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The bouncy particle sampler: A nonreversible rejection-free markov chain monte carlo method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Journal of the American Statistical Association, 113(522):855–867, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [11] Xiang Cheng, Niladri S Chatterji, Peter L Bartlett, and Michael I Jordan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Underdamped langevin mcmc: A non- asymptotic analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Conference on learning theory, pages 300–323.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' PMLR, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [12] Augustin Chevallier, Sam Power, Andi Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Wang, and Paul Fearnhead.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Pdmp monte carlo methods for piecewise- smooth densities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' arXiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='05859, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [13] Cloez, Bertrand, Dessalles, Renaud, Genadot, Alexandre, Malrieu, Florent, Marguet, Aline, and Yvinec, Romain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Probabilistic and piecewise deterministic models in biology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ESAIM: Procs, 60:225–245, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [14] Alice Corbella, Simon E F Spencer, and Gareth O Roberts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Automatic zig-zag sampling in practice.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' arxiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='11410, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [15] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Davis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Piecewise-Deterministic Markov Processes: A General Class of Non-Diffusion Stochastic Models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Journal of the Royal Statistical Society.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Series B (Methodological), 46(3):353–388, 1984.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [16] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Davis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Markov Models & Optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Chapman & Hall/CRC Monographs on Statistics & Applied Prob- ability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Taylor & Francis, 1993.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [17] Valentin De Bortoli, Alain Durmus, Marcelo Pereyra, and Ana Fernandez Vidal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Maximum likelihood estimation of regularization parameters in high-dimensional inverse problems: an empirical bayesian approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' part ii: Theoretical analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SIAM Journal on Imaging Sciences, 13(4):1990–2028, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [18] Persi Diaconis, Susan Holmes, and Radford M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Neal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Analysis of a nonreversible Markov chain sampler.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Annals of Applied Probability, 10(3):726 – 752, 2000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [19] Alain Durmus, Arnaud Guillin, and Pierre Monmarch´e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Geometric ergodicity of the bouncy particle sampler.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Annals of Applied Probability, 30(5):2069–2098, 10 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [20] Alain Durmus, Arnaud Guillin, and Pierre Monmarch´e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Piecewise deterministic Markov processes and their invariant measures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Annales de l’Institut Henri Poincar´e, Probabilit´es et Statistiques, 57(3):1442 – 1475, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [21] Alain Durmus and Eric Moulines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Sampling from strongly log-concave distributions with the unadjusted langevin algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' arXiv preprint arXiv:1605.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='01559, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [22] Alain Durmus, ´Eric Moulines, and Marcelo Pereyra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Efficient bayesian computation by proximal markov chain monte carlo: When langevin meets moreau.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SIAM Journal on Imaging Sciences, 11(1):473–506, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [23] Martin Hairer and Jonathan C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Mattingly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Yet another look at Harris’ ergodic theorem for Markov chains.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Seminar on Stochastic Analysis, Random Fields and Applications VI, volume 63 of Progr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Probab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=', pages 109–117.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Birkh¨auser/Springer Basel AG, Basel, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [24] K Hukushima and Y Sakai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' An irreversible markov-chain monte carlo method with skew detailed balance conditions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Journal of Physics: Conference Series, 473:012012, dec 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [25] Ben Leimkuhler, Daniel T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Margul, and Mark E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Tuckerman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Stochastic, resonance-free multiple time-step algorithm for molecular dynamics with very large time steps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Molecular Physics, 111(22-23):3579–3594, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [26] Benedict Leimkuhler and Charles Matthews.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Rational Construction of Stochastic Numerical Methods for Molecular Sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Applied Mathematics Research eXpress, 2013(1):34–56, 06 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [27] Benedict Leimkuhler, Charles Matthews, and Gabriel Stoltz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The computation of averages from equilibrium and nonequilibrium Langevin molecular dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' IMA J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Numer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Anal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=', 36(1):13–79, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [28] Vincent Lemaire, Mich`ele Thieullen, and Nicolas Thomas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Exact simulation of the jump times of a class of piecewise deterministic markov processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Journal of Scientific Computing, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [29] Vincent Lemaire, Mich´ele Thieullen, and Nicolas Thomas.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Thinning and multilevel monte carlo methods for piece- wise deterministic (markov) processes with an application to a stochastic morris–lecar model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Advances in Applied Probability, 52(1):138–172, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [30] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lewis and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Shedler.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Simulation of nonhomogeneous poisson processes by thinning, 1978.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [31] Eva L¨ocherbach and Pierre Monmarch´e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Metastability for systems of interacting neurons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Annales de l’Institut Henri Poincar´e, Probabilit´es et Statistiques, 58(1):343 – 378, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 34 SPLITTING SCHEMES FOR PDMPS [32] Sean P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Meyn and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Tweedie.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Stability of markovian processes iii: Foster–lyapunov criteria for continuous-time processes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Advances in Applied Probability, 25(3):518–548, 1993.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [33] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Michel, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Kapfer, and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Krauth.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Generalized event-chain Monte Carlo: Constructing rejection-free global- balance algorithms from infinitesimal steps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Journal of Chemical Physics, 140(5), 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [34] Pierre Monmarch´e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Piecewise deterministic simulated annealing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ALEA, 13(1):357–398, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [35] Pierre Monmarch´e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Kinetic walks for sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ALEA Lat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Am.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Probab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Stat.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=', 17:491–530, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [36] Pierre Monmarch´e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' High-dimensional MCMC with a standard splitting scheme for the underdamped Langevin diffusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Electronic Journal of Statistics, 15(2):4117 – 4166, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [37] Pierre Monmarch´e, Jeremy Weisman, Louis Lagard`ere, and Jean-Philip Piquemal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Velocity jump processes: An alternative to multi-timestep methods for faster and accurate molecular dynamics simulations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The Journal of Chemical Physics, 153(2):024101, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [38] Filippo Pagani, Augustin Chevallier, Sam Power, Thomas House, and Simon Cotter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Nuzz: numerical zig-zag sampling for general models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' arXiv:2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='03636, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [39] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Peters and G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' De With.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Rejection-free Monte Carlo sampling for general potentials.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Physical Review E Statistical, Nonlinear, and Soft Matter Physics, 85(2):1–5, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [40] R Tyrrell Rockafellar and Roger J-B Wets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Variational analysis, volume 317.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Springer Science & Business Media, 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [41] Leonid I Rudin, Stanley Osher, and Emad Fatemi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Nonlinear total variation based noise removal algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Physica D: nonlinear phenomena, 60(1-4):259–268, 1992.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [42] Konstantin S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Turitsyn, Michael Chertkov, and Marija Vucelja.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Irreversible monte carlo algorithms for efficient sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Physica D: Nonlinear Phenomena, 240(4):410–414, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [43] Paul Vanetti, Alexandre Bouchard-Cˆot´e, George Deligiannidis, and Arnaud Doucet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Piecewise-deterministic markov chain monte carlo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' arXiv:1707.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='05296, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [44] Ana Fernandez Vidal, Valentin De Bortoli, Marcelo Pereyra, and Alain Durmus.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Maximum likelihood estimation of regularization parameters in high-dimensional inverse problems: An empirical bayesian approach part i: Methodol- ogy and experiments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SIAM Journal on Imaging Sciences, 13(4):1945–1989, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [45] Marija Vucelja.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lifting – a nonreversible markov chain monte carlo algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' American Journal of Physics, 84, 12 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [46] Kai Xu, Hong Ge, Will Tebbutt, Mohamed Tarek, Martin Trapp, and Zoubin Ghahramani.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Advancedhmc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' jl: A ro- bust, modular and efficient implementation of advanced hmc algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Symposium on Advances in Approximate Bayesian Inference, pages 1–10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' PMLR, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proofs of Section 2 A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Fix g ∈ C2,0 Φ ∩ D(L).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By a telescoping sum we have Ez[g(Ztn)] − Ez[g(Ztn)] = n−1 � k=0 (Ez[Ptn−tk+1g(Ztk+1)] − Ez[Ptn−tkg(Ztk)]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For each k ∈ {0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' , n − 1}, set fk(y, s) = Ptn−tk−sg(y) then we have Ez[g(Ztn)] − Ez[g(Ztn)] = n−1 � k=0 Ez[fk(Ztk+1, δ) − fk(Ztk, 0)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By conditioning on Ztk it is sufficient to prove that |Ez[fk(Zδ, δ)] − fk(z, 0)| ≤ R(1 + |z|M)∥g∥C2,0 Φ δ3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (34) Indeed if we have that (34) holds then by Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 we have |Ez[g(Ztn)] − Ez[g(Ztn)]| ≤ Cδ3 n−1 � k=0 eR(tn−tk)Ez[G(Ztk)] ≤ C∥g∥C2,0 Φ eRtnδ3nGM(z), which gives the desired result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It remains to show that (34) holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 35 As done in [3] we rewrite the lhs as Ez[fk(Zδ, δ)] − fk(z, 0) = Ez[fk(Zδ, δ)] − fk(ϕδ(z), δ) + fk(ϕδ(z), δ) − fk(z, 0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (35) In particular, with identical steps to [3] we can rewrite the last two terms on the left hand side of (35) using the fundamental theorem of calculus and that ∂sfk(z, s) = −Lfk(z, s): fk(ϕδ(z), δ) − fk(z, 0) = − � δ 0 λ(ϕr(z))[Q(fk(·, r))(ϕr(z)) − fk(ϕr(z), r)]dr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we compute the expectation in the right hand side of (35), collecting a term for the case of no jumps, a single jump and the case of multiple jumps Ez[fk(Zδ, δ)] − fk(z, δ) = = � � δ 0 Q(ϕδ/2(z), d˜z) � fk(ϕδ/2(˜z), δ) − fk(ϕδ(z), δ) � λ(ϕδ/2(z))e−sλ(ϕδ/2(z))e−(δ−s)λ(˜z)ds (†) + ∞ � ℓ=2 Ez[(fk(Zδ, δ) − fk(ϕδ(z), δ))1{ℓ events}] (‡) − � δ 0 λ(ϕr(z))[Q(fk(·, r))(ϕr(z)) − fk(ϕr(z), r)]dr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (‡†) Observe that the sum in the second term (‡) can be truncated from ℓ = 3 onward as we only wish to get an order δ3 local error.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' we have |fk| ≤ ∥g∥∞ and hence ����� ∞ � ℓ=3 Ez[(fk(Zδ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' δ) − fk(ϕδ(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' δ))1{ℓ events}] ����� ≤ 2∥g∥∞Pz(ℓ ≥ 3 events) ≤ 2∥g∥∞ � � δ 0 � δ−s1 0 � δ−s1−s2 0 λ(ϕδ/2(z))e−s1λ(ϕδ/2(z))λ(z1)e−s2λ(z1)λ(z2)e−s3λ(z2) Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz1)Q(z1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz2)Q(z2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz3) ds1ds2ds3 ≤ 2∥g∥∞ � (1 − e−δλ(ϕδ/2(z)))(1 − e−δλ(z1))(1 − e−δλ(z2))Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz1)Q(z1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz2) ≤ 2δ3∥g∥∞ � λ(ϕδ/2(z))λ(z1)λ(z2)Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz1)Q(z1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz2) ≤ 2δ3∥g∥∞ � λ(ϕδ/2(z))λ(z1)Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz1)Qλ(z1) ≤ 2δ3∥g∥∞ λ(ϕδ/2(z))Q(λ(·)Qλ(·))(ϕδ/2(z)) where we used that 1 − exp(−z) ≤ z for z ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we have that λQ(λQλ) is polynomially bounded and therefore we can bound (‡) by (‡) = � Q(ϕδ/2(z), dz1)Q(z1, dz2) � fk(ϕδ/2(z2), δ) − fk(ϕδ(z), δ) � � δ 0 � δ−s1 0 λ(ϕδ/2(z))e−s1λ(ϕδ/2(z))λ(z1)e−s2λ(z1)e−(δ−s1−s2)λ(z2)ds2 ds1 + O(∥g∥∞(1 + |z|3mλ)δ3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here and throughout we understand F(z, δ, g) = O(∥g∥C2 Φ(1 + |z|m)δn) to mean that lim sup δ→0 sup z∈E sup g |F(z, δ, g)| ∥g∥C2 Φδn(1 + |z|m) ≤ C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 36 SPLITTING SCHEMES FOR PDMPS We Taylor expand several terms in order to verify that the local error is of order δ3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We use repeatedly the following expansions: λ(ϕs(z)) = λ(z) + sDΦλ(z) + s2R(z, ˜s;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' λ), fk(ϕs(z), δ) = fk(z, δ) + sDΦfk(z, δ) + s2R(z, ˜s;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' fk), R(z, ˜s;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' g) = D2 Φg(ϕ˜s(z))/2, fk(z, s) = fk(z, 0) − sLfk(z, 0) + s2L2fk(z, ˜s)/2, for some ˜s ∈ [0, s] (note that ˜s may vary with each term so when we use these expansions we include an index to distinguish different incidents of ˜s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that by Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 we have ∥fk∥C2 Φ ≤ CeR(tn−tk)(1+|z|mP)∥g∥C2,0 Φ which gives us a bound on the remainder terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Applying the expansions above to (†) we obtain (†) = � Q(ϕδ/2(z), d˜z) � fk(˜z, 0) − δLfk(˜z, 0) + δ2 2 L2fk(˜z, ˜s2) + δ 2DΦfk(˜z, δ) + 1 8δ2R(˜z, ˜s2;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' fk) − fk(z, 0) + δLfk(z, 0) − 1 2δ2L2fk(z, ˜s3) − δDΦfk(z, δ) − 1 2δ2R(z, ˜s3;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' fk) � � δ 0 � λ(z) + δ 2DΦλ(z) + 1 8δ2R(z, ˜s4;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' λ) � � 1 − sλ(ϕδ/2(z)) − (δ − s)λ(˜z) + 1 2(sλ(ϕδ/2(z)) + (δ − s)λ(˜z))2e−η� ds = � Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z) � fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � � δ 0 � λ(z) + δ 2DΦλ(z) �� 1 − sλ(z) − (δ − s)λ(˜z) � ds + δ � Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z) � − Lfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + 1 2DΦfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + Lfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − DΦfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � � δ 0 � λ(z) + δ 2DΦλ(z) �� 1 − sλ(z) − (δ − s)λ(˜z) � ds + eR(tn−tk)O(∥g∥C2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0 Φ (1 + |z|M)δ3) where we used η ∈ [0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' sλ(ϕδ/2(z)) + (δ − s)λ(˜z)] in the first equality and further Taylor expansions to obtain the second equality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here M = 3mλ+mP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now using Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we can expand the Q term (†) = � � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z) + δ 2DΦQ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z) �� fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � � δ 0 � λ(z) + δ 2DΦλ(z) �� 1 − sλ(z) − (δ − s)λ(˜z) � ds + δ � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z) � − Lfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + 1 2DΦfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + Lfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − DΦfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � � δ 0 � λ(z) + δ 2DΦλ(z) �� 1 − sλ(z) − (δ − s)λ(˜z) � ds + eR(tn−tk)O(∥g∥C2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0 Φ (1 + |z|M)δ3) Term (‡) can be expanded as (‡) = � Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)Q(z1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz2) � fk(z2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � � δ 0 � δ−s1 0 � λ(z) + δDΦ(λ)(ϕ˜s4(z)) � λ(z1) � 1 + (−s1λ(ϕδ/2(z)) − s2λ(z1) − (δ − s1 − s2)λ(z2))e−ξ� ds2 ds1 + eR(tn−tk)O(∥g∥C1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0 Φ (1 + |z|mP+3mλ)δ3) =δ2 2 � Q(ϕδ/2(z),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)Q(z1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz2) � fk(z2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � λ(z)λ(z1) + eR(tn−tk)O(∥g∥C1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0 Φ (1 + |z|mP+3mλ)δ3) SPLITTING SCHEMES FOR PDMPS 37 =δ2 2 � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz1)Q(z1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' dz2) � fk(z2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � λ(z)λ(z1) + eR(tn−tk)O(∥g∥C1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0 Φ (1 + |z|mP+3mλ)δ3),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' where ξ ∈ [0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' s1λ(ϕδ/2(z)) + s2λ(z1) + (δ − s1 − s2)λ(z2)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we can expand the term (‡†) as follows: (‡†) = − � δ 0 λ(ϕr(z)) � Qfk(·, r)(z) + r � DΦQ(z, d˜z)fk(˜z, r) − fk(ϕr(z), r) � dr + eR(tn−tk)O(δ3∥g∥C2,0 Φ (1 + |z|mλ)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Taylor’s theorem (‡†) = − � δ 0 λ(ϕr(z)) � Qfk(·, 0)(z) − r � Q(z, d˜z)Lfk(˜z, 0) + r � DΦQ(z, d˜z)(fk(˜z, 0) − rLfk(˜z, ˜r)) − (fk(z, 0) + rDΦfk(z, 0) − rLfk(z, 0))) dr + eR(tn−tk)O(δ3∥g∥C2,0 Φ (1 + |z|mλ+mP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that � DΦQ(z, d˜z)Lfk(˜z, ˜r) = Q(DΦLfk(·, ˜r))(z) = eR(tn−tk)O((1 + |z|mP)∥g∥C2,0 Φ ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using this and Taylor expanding λ(ϕr(z)) we have (‡†) = − � δ 0 λ(z) � Qfk(·, 0)(z) − r � Q(z, d˜z)Lfk(˜z, 0) + r � DΦQ(z, d˜z)fk(˜z, 0) − (fk(z, 0) + rDΦfk(z, 0) − rLfk(z, 0)) � dr − � δ 0 rDΦλ(z) (Qfk(·, 0)(z) − fk(z, 0)) dr + eR(tn−tk)O(δ3∥g∥C2,0 Φ (1 + |z|mλ+mP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Evaluating the integral over r (‡†) = − λ(z) � Qfk(·, 0)(z)δ − 1 2δ2 � Q(z, d˜z)Lfk(˜z, 0) + 1 2δ2 � DΦQ(z, d˜z)fk(˜z, 0) − � fk(z, 0)δ + 1 2δ2DΦfk(z, 0) − 1 2δ2Lfk(z, 0) �� − 1 2δ2DΦλ(z) (Qfk(·, 0)(z) − fk(z, 0)) + eR(tn−tk)O(δ3∥g∥C2,0 Φ (1 + |z|mλ+mP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First order terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Terms of order δ appear only in (†) and (‡†) and clearly they cancel out.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Second order terms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In (†) we can further expand terms of the form DΦ(f)(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' δ) and rearrange as Order δ2 of (†) =δ2 � � λ(z) � − Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)Lfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)1 2DΦfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + 1 2DΦQ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)(fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)) � + λ(z)(Lfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − DΦfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)⟩) + � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)(fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)) � − 1 2λ(z)(λ(˜z) + λ(z)) + 1 2DΦλ(z) �� =δ2 � � λ(z) � − Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)Lfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) + Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)1 2DΦfk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)⟩ 38 SPLITTING SCHEMES FOR PDMPS + 1 2DΦQ(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)(fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)) � �� � Term A � + λ(z) � Lfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − DΦfk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)) − 1 2λ(z) � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)(fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)) � � �� � Term B + 1 2DΦλ(z) � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)(fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0)) � �� � Term C − 1 2 � Q(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' d˜z)(fk(˜z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) − fk(z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 0) � �� � Term D )λ(z)λ(˜z) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For term (‡) we have Order δ2 of (‡) = 1 2δ2 � Q(z, dz1)Q(z1, dz2)(fk(z2, 0) − fk(z, 0) � �� � Term D )λ(z)λ(z1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly for (‡†) we have Order δ2 of (‡†) = − δ2 2 � � DΦ(Q)(z)λ(z)(fk(˜z, 0) − fk(z, 0)) � �� � Term A + � Q(z, d˜z)DΦ(λ)(z)(fk(˜z, 0) − fk(z, 0)) � �� � Term C + � Q(z, d˜z)λ(z)(−Lfk(˜z, 0) + Lfk(z, 0) − DΦ(fk)(z, 0) � �� � Term B � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' After cancellations we obtain Order δ2 of (†) + (‡) + (‡†) = δ2λ(z) � � � − Q(z, d˜z)Lfk(˜z, 0) + Q(z, d˜z)1 2DΦ(fk)(˜z, 0) � − 1 2 � Q(z, d˜z)fk(˜z, 0)λ(˜z) + 1 2 � Q(z, d˜z)Q(˜z, dz2)fk(z2, 0)λ(˜z) + 1 2 � Q(z, d˜z)Lfk(˜z, 0) � = 1 2δ2λ(z) � � Q(z, d˜z) � − Lfk(˜z, 0) + DΦ(fk)(˜z, 0) + λ(˜z) � Q(˜z, dz2)[fk(z2, 0) − fk(˜z, 0)] �� = 0, where the last equality follows by the definition of the generator of the PDMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore we have shown that second order terms cancel out.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proofs for Example 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us verify that Assumption 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that (DΦQ)(g)(x, v) = vT d � i=1 ∇x �λi(x, v) λ(x, v) g(x, Fiv) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 39 Therefore by Taylor’s theorem we have for some η between x and x + sv Qg(x + sv, v) − Qg(x, v) − (DΦQ)(g)(x, v) = 1 2s2 d � i=1 vT ∇2 x �λi(x, v) λ(x, v) g(x, Fiv) � ���� x=η v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It remains to show that λi λ , ∇x � λi(x,v) λ(x,v) � , ∇2 x � λi(x,v) λ(x,v) � are bounded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is clear that 0 < λi/λ ≤ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us consider the first derivative.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Set Ξ(s) = log(1 + es) so that λi(x, v) = Ξ(vi∂iψ(x)) and note that 0 ≤ Ξ′(s) Ξ(s) ≤ 1, 0 ≤ Ξ′′(s) Ξ(s) ≤ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now ����∇x �λi(x, v) λ(x, v) ����� = ���� ∇xλi(x, v) λ(x, v) − λi(x, v)∇xλ(x, v) λ(x, v)2 ���� ≤ ���� ∇xλi(x, v) λ(x, v) ���� + d � j=1 ���� ∇xλj(x, v) λ(x, v) ���� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' So it remains to show that ∇xλi/λ is bounded.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using the bounds on Ξ we have ���� ∇xλi(x, v) λ(x, v) ���� ≤ ���� ∇xλi(x, v) λi(x, v) ���� = ���� Ξ′(vi∂iψ(x)) Ξ(vi∂iψ(x)) vi∇x∂iψ(x) ���� ≤ |∇x∂iψ(x)| .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is bounded by our assumptions on ψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us now consider ∇2 x � λi(x,v) λ(x,v) � : ����∇2 x �λi(x, v) λ(x, v) ����� = ����� ∇2 xλi(x, v) λ(x, v) − 2∇xλi(x, v)∇xλ(x, v)T λ(x, v)2 + 2λi(x, v)∇xλ(x, v)∇xλ(x, v)T λ(x, v)3 − λi(x, v)∇2 xλ(x, v) λ(x, v)2 �����.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using the bound on ∇xλi/λ we can bound all the terms aside from the one involving the second derivative of λ so it suffices to consider ���� ∇2 xλi(x, v) λ(x, v) ���� = ����� Ξ′′(vi∂iψ(x)) � j Ξ(vj∂jψ(x))∇x∂iψ(x)∇x∂iψ(x)T + Ξ′(vi∂iψ(x)) � j Ξ(vj∂jψ(x))vi∇2 x∂iψ(x) ����� ≤ ��∇x∂iψ(x)∇x∂iψ(x)T �� + ��∇2 x∂iψ(x) �� This is bounded since ψ has bound second and third order derivatives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Appendix B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proofs of ergodicity for splitting schemes of the BPS To fix ideas, we consider a splitting scheme for a BPS with unitary velocity with generator L decomposed as L = L1 + L2 + L3 with L1f(x, v) = vT ∇xf(x, v), L2f(x, v) = � vT ∇ψ(x) � + (f(x, R(x)v) − f(x, v)) , L3f(x, v) = λr � Sd−1 (f(x, w) − f(x, v)) dw .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Write P j t = etLj the associated semi-groups for j ∈ �1, 3�.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall show the following statement, which implies Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider any scheme of the BPS based on the decomposition D,R,B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Under Assump- tion 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3, the following hold: 40 SPLITTING SCHEMES FOR PDMPS (1) There exist a, b, C, δ0 > 0 and a function V (with a, b, C, δ0, V depending only on ψ and λr, but not on δ) such that, for all x, v, e|x|/a/a ⩽ V (x, v) ⩽ aea|x| and for all δ ∈ (0, δ0] and all x, v, PδV (x, v) ⩽ (1 − bδ) V (x, v) + Cδ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (2) For all R > 0, there exist c, δ0 > 0 and a probability measure ν on E such that for all x, v with |x| ⩽ R and all δ ∈ (0, δ0], setting n∗ = ⌈4R/δ⌉, δx,vP n∗ δ ⩾ cν .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lyapunov function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Under Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3, let W(x) = � ψ(x) so that ∇W is bounded and, as |x| → ∞, lim inf |∇W(x)| > 0 and ∇2W(x) → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let cW > 0 be such that |∇W(x)| ⩾ cW for |x| large enough.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let q = � Sd−1 1w1⩽−1/2dw > 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let ϕ ∈ C2(R) be a non-decreasing function such that ϕ′(0) > 0 and ϕ(θ) � � � = 1 for θ ⩽ − 1 2cV ⩾ 2 − q 2 for θ ⩾ − 1 4cV = 2 for θ ⩾ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let κ = 4λr cW , M = κ ϕ � qλr 4κ � − ϕ � − qλr 4κ � (where we used that ϕ′(0) > 0 so that the denominator is positive).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, let ε = λrq 16∥ϕ′∥∞ and let R > 0 be such that for all x ∈ Rd with |x| ⩾ R, |∇W(x)| ⩾ cW , 2W(x) ⩾ M, and ∥∇2W(x)∥ ⩽ ε .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Preliminary computation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the following we denote by the same letter C various constants (inde- pendent from t) and we assume that t ∈ (0, δ0] with δ0 = 1/(∥∇W∥∞κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We consider a Lyapunov function V (x, v) = eκW(x)ϕ(θ(x, v)) , θ(x, v) = vT ∇W(x) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Notice that |θ| is bounded by ∥∇W∥∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' More generally,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for a non-negative C2 function g,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' we bound P 1 t � eκW g ◦ θ � (x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = eκW(x+tv)g(vT ∇W(x + tv)),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' using that etr ⩽ 1 + tr + r2t2erδ0/2 for t ⩽ δ0 as eκW(x+tv) ⩽ eκ(W(x)+tvT ∇W(x)+ 1 2 t2∥∇2W∥∞) ⩽ eκW(x) � 1 + tκvT ∇W(x) + Ct2� and,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' similarly,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' g(v · ∇W(x + tv)) ⩽ g(θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) + tvT ∇2W(x)vg′(θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) + Ct2 ⩽ g(θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) + � ε∥g′∥∞ + C1|x|⩽R � t + Ct2 SPLITTING SCHEMES FOR PDMPS 41 (here and below the constants C implicitly involve the suprema of g,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' |g′| and |g′′| over [−∥∇W∥∞,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∥∇W∥∞]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Combining these two bounds and using that |tκθ| ⩽ 1, we get P 1 t � eκW g ◦ θ � ⩽ eκW � (1 + tκθ)g ◦ θ + ε∥g′∥∞t + Ct2� + Ct .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (36) Second, for any function g, P 2 t � eκW g ◦ θ � (x, v)=eκW(x) � e−t(vT ∇ψ(x))+g � vT ∇W(x) � + � 1 − e−t(vT ∇ψ(x))+� g � (R(x)v)T ∇W(x) �� =eκW(x) � g(θ(x, v)) + � 1 − e−2tW(x)θ+(x,v)� (g (−θ(x, v)) − g (θ(x, v))) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the second equality we used (vT ∇ψ(x))+ = 2W(x)(vT ∇W(x))+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, if g is non-decreasing, P 2 t � eκW g ◦ θ � (x, v) ⩽ eκW(x) � g(θ(x, v)) + � 1 − e−tMθ+(x,v)� (g (−θ(x, v)) − g (θ(x, v))) + tC1|x|⩽R � ⩽ eκV (x) � g(θ(x, v)) + tMθ+(x, v) (g (−θ(x, v)) − g (θ(x, v))) + Ct2� + tC1|x|⩽R .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (37) Third,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for any function g,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' similarly,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' P 3 t � eκW g ◦ θ � (x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = eκW(x) � g (θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) + � 1 − e−λrt� �� Sd−1 g(wT ∇W(x))dw − g (θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) �� ⩽ eκW(x) � g (θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) + λrt �� Sd−1 g(w · ∇W(x))dw − g (θ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) � + Ct2 � (38) In particular,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for g = ϕ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' using the rotation-invariance of the uniform measure on the sphere and the fact that ϕ is increasing and with values in [1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 2],' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' we bound � Sd−1 ϕ(w · ∇W(x))dw = � Sd−1 ϕ(w1|∇W(x)|)dw ⩽ 1|x|⩽R + A with A = sup |x|⩾R � Sd−1 ϕ(w1|∇W(x)|)dw .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Combining the computations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now, to fix ideas, let us start with the DRBRD case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' From our previous computations, using first (36) and then (38) (keeping track only of the terms of order 0 or 1 in t, the higher order ones being bounded and gathered in the term t2C), P 1 t P 3 t P 2 2tP 3 t P 1 t V ⩽ P 1 t P 3 t P 2 2tP 3 t � eκW � (1 + tκθ)ϕ(θ) + tε∥ϕ′∥∞ + t2C �� + tC ⩽ P 1 t P 3 t P 2 2t � eκW � (1 + tκθ)ϕ(θ) + λrt(A − ϕ(θ)) + tε∥ϕ′∥∞ + t2C �� + tC := P 1 t P 3 t P 2 2t � eκW Ψ(θ) � + tC .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assume that t ⩽ 1/(2λr + 2∥W∥∞κ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Under this condition Ψ(θ) is non-decreasing in θ, since ϕ is non-decreasing in θ and (1 + t(κθ − λr)) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Thus,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' applying (37) and then (38) and (36) again,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' P 1 t P 3 t P 2 2tP 3 t P 1 t V ⩽ P 1 t P 3 t � eκW � Ψ(θ) + 2tMθ+ (Ψ(−θ) − Ψ(θ)) + t2C �� + tC ⩽ P 1 t P 3 t � eκW � Ψ(θ) + 2tMθ+ (ϕ(−θ) − ϕ(θ)) + t2C �� + tC ⩽ P 1 t � eκW � (1 + tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tMθ+ (ϕ(−θ) − ϕ(θ)) + tε∥ϕ′∥∞ + t2C �� + tC ⩽ eκW � (1 + 2tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tMθ+ (ϕ(−θ) − ϕ(θ)) + 2tε∥ϕ′∥∞ + t2C0 � + tC ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 42 SPLITTING SCHEMES FOR PDMPS for some C0,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C > 0 (in the last inequality,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' we use that t is small enough so that the function which multiplies eκW is non-negative in order to apply (36)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The choice of ε ensures that 4ε∥ϕ′∥∞ ⩽ λrq/4 =: η.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' So, if we have that, for all r ∈ R, κr + λ(A − ϕ(r)) + Mr+ (ϕ(−r) − ϕ(r)) ⩽ −η , (39) then, using that ϕ(θ) ∈ [1, 2], we will conclude that, for t ⩽ min(η/C0, 1/(λr + ∥W∥∞κ))/2, P 1 t P 3 t P 2 2tP 3 t P 1 t V ⩽ eκW � ϕ(θ) − 2ηt + 1 2ηt + 1 2ηt � + tC ⩽ � 1 − tη 2 � V + tC , which will conclude the proof of the first part of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 for the scheme DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us check what happens for different splitting orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The computations are similar, in particular an important point is to check that the bound for P 1 t (resp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' P 2 t ) is always used for a function g which is non-negative (resp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' non-decreasing in θ), which is ensured each time by the assumption that t is small enough, as in the previous DRBRD case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For instance,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for BDRDB,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' we get,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' for t small enough,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='P 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2tP 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t V ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='⩽ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='P 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2tP 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='eκW � ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='ϕ(θ) + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+ tC ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='⩽ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='P 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 3 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2t ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='eκW � ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='(1 + tκθ)ϕ(θ) + tε∥ϕ′∥∞ + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+ tC ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='⩽ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='P 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t P 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='eκW � ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='(1 + tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + tε∥ϕ′∥∞ + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+ tC ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='⩽ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='P 2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='t ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='eκW � ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='(1 + 2tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tε∥ϕ′∥∞ + tMθ+(ϕ(−θ) − ϕ(θ)) + t2C ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+ tC ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='⩽ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='eκW � ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='(1 + 2tκθ)ϕ(θ) + 2λrt(A − ϕ(θ)) + 2tε∥ϕ′∥∞ + 2tMθ+(ϕ(−θ) − ϕ(θ)) + t2C ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+ tC ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' which is exactly the same bound as in the DRBRD case,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' as expected since we only keep track of the first order terms in t,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' which coincide for all splitting orders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The other cases are similar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Conclusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us check that (39) holds with our choices of parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using that ϕ(r) ⩽ 2 for all r ∈ R and ϕ(r) = 1 for all r ⩽ −cV /2 and that |∇V (x)| ⩾ cV if |x| ⩾ R, A ⩽ q × 1 + (1 − q) × 2 = 2 − q .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then, for r ⩾ −cV /4, since ϕ(r) ⩾ 2 − q/2, A − ϕ(r) ⩽ −q/2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The choice of κ ensures that, for all r ⩽ −cV /4, κr + λr(A − ϕ(r)) ⩽ κr + λr(1 − q) ⩽ −λrq .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For r ∈ [cV /4, qλr/(4κ)], κr + λr(A − ϕ(r)) ⩽ κr − λrq/2 ⩽ −λrq/4 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally, for r ⩾ qλr/(4κ), the choice of M ensures that κr + λr(A − ϕ(r)) + Mr (ϕ(−r) − ϕ(r)) ⩽ � κ − M � ϕ �qλr 4κ � − ϕ � −qλr 4κ ��� r − λrq/2 ⩽ −λrq/2 , which concludes the proof of the first part of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 43 B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Doeblin condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof is an adaptation from [34, Lemma 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2] (see also [19, Lemma 11]), with the additional difficulty that time is discrete.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Fix R > 0 and consider an initial condition (x, v) ∈ Rd × Sd−1 with |x| ⩽ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We want to prove that the law of the process at time n∗ = ⌈4R/δ⌉ is bounded below by a positive constant (independent from δ small enough) times the Lebesgue measure on some domain, uniformly in x with |x| ⩽ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since the process moves at unitary speed, for all n ∈ N, |Xn| ⩽ R + nδ, and thus P (no bounce in the n first steps) ⩾ e−h(nδ) with h(t) = sup{∥∇ψ(x)|, |x| ⩽ R + t}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the absence of bounces, depending on the splitting order, the chain behaves either as the chain given by the splitting DRD or RDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To fix ideas, we only consider the DRD case in the following (corresponding either to DRBRD, DBRBD or BDRDB), the proof being similar in the other case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For k ⩾ 1, denote by Tk the number of transitions of the chain between the (k − 1)th refreshment and the kth one, so that (Tk)k∈N is an i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' sequence of geometric random variables with parameter 1−e−δλr, independent from the random variables used to define the bounces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly, let (Vk)k∈N be the i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' sequence of random variables uniformly distributed on the sphere used to define the velocities at the refreshment times.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For a small parameter η > 0 to be fixed later on, consider the events A1 = {T1 + T2 + T3 ⩽ n∗, |δTj − 2R| ⩽ ηR for j = 2, 3} A2 = {δ(T4 − 1) > 2ηR, no bounce in the n∗ first steps} .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, under A1 ∩ A2, exactly three refreshments occur during the n∗ first transitions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then, for all Borel set B of Rd × Sd−1, P � Zn∗ ∈ B � ⩾ P � Zn∗ ∈ B, A1, A2 � = P � ( ˜X, V3) ∈ B, A1, A2 � = P(A2)P � ( ˜X, V3) ∈ B, A1 � with, in the DRD case for instance, ˜X = x + δ �� T1 − 1 2 � v + T2V1 + T3V2 + � n∗ − T1 − T2 − T3 + 1 2 � V3 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (Indeed, the DRD scheme starts and ends with a half step of transport).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Notice that P(A2) is lower bounded uniformly in δ ∈ (0, 1] as P(A2) ⩾ e−h(4R+1)e−λr(2ηR+1) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' On the other hand, writing I = {(t1, t2, t3) ∈ N3, t1 + t2 + t3 ⩽ n∗, |δtj − 2R| ⩽ ηR for j = 2, 3} , we get P � ( ˜X, V3) ∈ B, A1 � = � I (1 − e−δλr)3e−δλr(t1+t2+t3−3) � (Sd−1)3 1B � x + δ �� t1 − 1 2 � v + t2v1 + t3v2 + � n∗ − t1 − t2 − t3 + 1 2 � v3 � , v3 � dv1dv2dv3 ⩾ � I (1 − e−δλr)3e−δλr(t1+t2+t3−3) inf |x′|⩽R(1+2η) � (Sd−1)3 1B � x′ + δ(t2v1 + t3v2), v3 � dv1dv2dv3 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 44 SPLITTING SCHEMES FOR PDMPS By the rotation invariance of the uniform measure on the sphere, for fixed t, s > 0, sV1 + tV2 has the same distribution as V1|sw+tV2| where w is a fixed unitary vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then, |sw+tV2|2 = s2+t2+2stV T 2 w and V T 2 w has on [−1, 1] the probability density proportional to u �→ (1 − u2)d/2−1 (this is here we use that d ≥ 2), which is lower bounded by a positive constant on [−1 + ε, 1 − ε] for all ε > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assuming that η ⩽ 1/4 and considering for y ∈ Rd the ring R(y) = {x ∈ Rd, 4ηR ⩽ |x − y| ⩽ 4R(1 − 2η)}, we get that sV2 + tV3 has a density which is lower bounded on R(0) by a constant α > 0 which is independent from t, s ∈ [R(2 − η), R(2 + η)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a consequence, for (t1, t2, t3) ∈ I and x′ ∈ Rd with |x′| ⩽ R(1 + 2η), the law of (x′ + δ(t2V1 + t3V2), V3) has a density lower bounded by α on R(x′) × Sd−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assuming that η ⩽ 1/16, let R′ = {y ∈ Rd, R(1 + 6η) ⩽ |y| ⩽ R(3 − 10η)} (which has a non-zero Lebesgue measure).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The triangle inequality implies that R′ ⊂ R(x′) if |x′| ⩽ R(1 + 2η).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a consequence, P � ( ˜X, V3) ∈ B, A1 � ⩾ P (A1) α � R′×Sd−1 1B(y, v)dydv , which concludes since P(A1) converges to a positive constant as δ → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Ergodicity for splitting schemes of ZZS C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this Section we focus on splitting scheme DBD for ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall prove the following, which implies Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting scheme DBD for ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then the following hold: (1) There exists a function V : Rd × {±1}d → R, and constants b ∈ (0, 1), C < ∞ such that for all (x, v) and all δ ∈ (0, δ0) with δ0 = 2(1 + γ0)−1, where γ0 is as in Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(b), it holds that PδV (x, v) ≤ (1 − bδ)V (x, v) + Cδ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (40) (2) For any R > 0 consider a set C = [−R, R]d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For some L > 0 let n∗ = 2 + 4x0 + 2R δ + 2 �L δ � ∈ 2N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For (x, v) ∈ C × {±1}d define the set D(x, v) given by (21).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then for any (y, w) ∈ D(x, v) ∩ (C × {±1}d) and δ ∈ (0, δ0] for δ0 > 0 it holds that δy,wP n∗ δ ⩾ cν .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' where c is independent of δ and ν is uniform over D(x, v) ∩ (C × {±1}d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 we prove the minorisation condition, in Section C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 we prove the drift condition, while in Section C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we prove Equation (24).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Minorisation condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We now prove a minorisation condition for splitting scheme DBD of ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the following Lemma we consider the one-dimensional setting, for which the reasoning is similar to that of the proof of a minorisation condition for the continuous ZZS done in [2, Lemma B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting scheme DBD of ZZS with step size δ ≤ δ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a) holds for some x0 ≥ 0 and consider a set C = [−R, R] for R > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For L > 0 let N = 2 + 4x0 + 2R δ + 2 �L δ � ∈ 2N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (41) SPLITTING SCHEMES FOR PDMPS 45 For (x, v) ∈ C × {±1} define the set D(x, v) := D+(x, v) ∪ D−(x, v), where D+(x, v) := {(y, w) : w = v, y = x + mδ, m ∈ 2Z}, D−(x, v) := {(y, w) : w = −v, y = x + mδ, m ∈ 2Z + 1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (42) Then for any (y, w) ∈ D(x, v) ∩ (C × {±1}) it holds that P(y,w)((XN, VN) ∈ ·) ≥ bν(·) where b is independent of δ and ν is uniform over D(x, v) ∩ (C × {±1}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let C = [−R, R] for a fixed R > 0 and let x ∈ C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall consider only the case of v = +1, as the same arguments extend to the symmetric case v = −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular observe that if the process is started in set D(x, +1) (respectively D(x, −1)), then after an even number of iterations it will again be in D(x, +1) (D(x, −1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This means that the process lives on D(x, +1) (respectively on D(x, −1)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To shorten the notation we denote by D+, D− the sets D+(x, +1), D−(x, +1) as defined in (42).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Below we focus on the case of an initial condition in D+, while the case of D− follows with an identical reasoning and obvious changes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Fix N ∈ 2N and define λ := max x∈C max y:|y−x|≤Nδ,v=±1 λ(y, v) which is the largest switching rate that can be reached within N iterations starting in C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that taking N as in (41) implies that Nδ is upper bounded by a constant as δ ≤ δ0 and thus λ can be chosen independently of the step size δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall λ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' From here on we shall denote the initial condition as (y, w) ∈ D+, and without loss of generality we shall assume x0 = x+ℓδ for some ℓ ∈ N, where x0 is as in Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We want to lower bound the probability that after N iterations the process is in measurable sets B ⊂ D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We consider two cases: in the first one the final state of the process is of the form (XN, VN) = (z, −1) ∈ D− ∩ (C × {±1}), while in the second case (XN, VN) = (z, +1) ∈ D+ ∩ (C × {±1}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the case in which the final state has negative velocity, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' VN = −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To lower bound the probability of reaching this state, we consider the case in which only one switching event takes place.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let z = y + mδ with m ∈ N odd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then in order to have (XN, VN) = (z, −1) with exactly one event taking place at time N1 it must be that y + (N1 − 1)δ − (N − N1)δ = z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Thus we find that the event should take place at N1 = z − y 2δ + N + 1 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order to guarantee the switching rate is strictly positive it must also be that XN1 ≥ x0, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' y + (N1 − 1)δ ≥ x0 and thus N1 ≥ 1 + (x0 − y)/δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note N1 < N, where N is as in (41).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denote the position at the time of the switching event by ˜x = y + δ(N1 − 1/2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then probability of exactly one event taking place at iteration N1 is given by � δ 0 λ(˜x, 1) exp(−sλ(˜x, 1)) exp(−(δ − s)λ(˜x, −1))ds ≥ δλ exp(−δλ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The probability of this path is simple to lower bound,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' since upper bounding the switching rates gives a smaller probability: P(y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+1)((XN,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' VN) = (z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' −1)) ≥ N1−1 � n=0 exp(−δλ(y + (n + 1/2)δ)) � �� � no jumps before N1 × δλ exp(−δλ) � �� � a jump at N1 × 46 SPLITTING SCHEMES FOR PDMPS × N−N1 � n=0 exp(−δλ(y + (N1 − 1 − n)δ)) � �� � no jumps after N1 ≥ exp(−(N1 − 1)δλ) δλ exp(−δλ) exp(−(N − N1)δλ) ≥ 2 exp(−(N − 1)δλ) λ exp(−δ0λ) × �1 2 δM M � ≥ 2 exp(−(N − 1)δλ) λ exp(−δ0λ)(2R − δ) � ν(−1) × 1 M � ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' where M ∈ N is the number of points in D+ ∩ (C × {±1}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the last line we used that δM ≥ 2R − δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall that δ ≤ δ0 and that N is given by (41).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This concludes as (N − 1)δ ≤ 4x0 + R + 2L + 3δ0 and 2R − δ ≥ 2R − δ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Second case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We now focus on the case in which VN = +1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall find an appropriate lower bound by restricting to the case in which exactly two switching events take place.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denoting the times of the two events as N1, N2, if the final position is z it must be y + (N1 − 1)δ − (N2 − 1)δ + (N − N1 − N2)δ = z which implies N2 = y − z 2δ + N 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (43) Moreover, at event times the process should be in regions with strictly positive switching rate: y + (N1 − 1/2)δ ≥ x0, y + (N1 − 1)δ − (N2 − 1/2)δ ≤ −x0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' These imply respectively N1 ≥ x0 − y δ + 1 =: N1, N2 ≥ y + x0 δ + N1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since N2 is determined by (43), we enforce that the second inequality holds: y − z 2δ + N 2 ≥ y + x0 δ + N1 which implies N1 ≤ N 2 − y + 2x0 + z 2δ =: N1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now to obtain the right dependence on δ, we shall take N such that N1 − N1 is increasing as 1/δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds N1 − N1 = N − 2 2 − 4x0 − y + z 2δ and thus it is sufficient to take N = 2 + 4x0 − y + z δ + 2 �L δ � for some constant L > 0, as with this choice N1 − N1 = ⌈L/δ⌉.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 47 Using the results above we find P(y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='+1)((XN,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' VN) = (z,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' +1)) ≥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='N1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='N1=N1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� N1−1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='n=0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='exp(−δλ(y + (n + 1/2)δ)) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='no jumps before N1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='× δλ exp(−δλ) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='a jump at N1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='× ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='× ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='N2−1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='m=0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='exp(−δλ(y + (N1 − 1 − (m + 1/2))δ)) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='no jumps until N2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='× δλ exp(−δλ) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='a jump at N2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='× ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='N−N1−N2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='ℓ=0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='exp(−δλ(y + (N1 − 1 − (N2 − 1) + (ℓ + 1/2))δ)) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='no jumps after N2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='≥ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='N1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='N1=N1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='exp(−δλNδ))δ2λ2 exp(−2δλ) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='= ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='�L ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='exp(−λNδ))δ2λ2 exp(−2δλ) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='≥ L exp(−δλN))λ2 exp(−2δ0λ)δ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='≥ 2L exp(−δλN))λ2 exp(−2δ0λ)(2R − δ0) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='ν(+1) × 1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='M ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='� ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly to above it is now sufficient to note that Nδ ≤ 4δ0 + 4x0 + 2R + 2L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Conclusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To conclude it is sufficient to observe that the conditions above hold for any choice of x, y, z ∈ C since N is as in (41).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Multidimensional case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To extend to the higher dimensional setting, first observe that it is possible to apply the same ideas in the proof of Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 to each component, in particular requiring that the events happen when all components of the process are outside of the rectangle [−x0, +x0].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This implies that Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a) can be used to lower bound the probability of flipping each component of the velocity vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence each coordinate can be controlled independently of the others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is clear that the following minorisation condition is implied: let C = [−R, R]d for R > 0, (x, v) ∈ C × {±1}d, and let D(x, v) as in (21);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' then for all (y, w) ∈ (x, v) it holds that P(y,w)((XN, VN) ∈ ·) ≥ bd νd(·), where N, b are as in Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 and νd is the uniform distribution over states in the grid D(x, v) ∩ (C × {±1}d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Drift condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us first characterise in the following Lemma the law of the jump part of the process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This result is then used to prove the wanted drift condition in Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let ˜V x t denote the PDMP corresponding to the generator L2 (for this process x acts as a parameter).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose that λi(x, v) is independent of vj for j ̸= i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then for any w ∈ {±1}d we have Pv( ˜V x t = w) = d � i=1 λi(x, Fiw) + wi vi λi(x, v)e−(λi(x,v)+λi(x,Fiv))t λi(x, v) + λi(x, Fiv) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 48 SPLITTING SCHEMES FOR PDMPS Proof of Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To simplify notation we will suppress the dependence on x and set Λi(v) = λi(v) + λi(−v) = λi(x, v) + λi(x, Fiv).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since ˜V x t jumps according to λi which does not depend on vj we have that the coordinates of ˜V x t are all independent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence it is sufficient to show Pvi(( ˜V x t )i = wi) = λi(−wi) + wi vi λi(vi)e−Λi(vi)t Λi(vi) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore it is sufficient to consider the setting d = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Define for any t ≥ 0, v, w ∈ {1, −1} ϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) := λ(−w) + w v λ(v)e−Λ(v)t Λ(v) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If we show that for all t ≥ 0, v, w ∈ {1, −1} ∂tϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) = L2ϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w), ϕ0(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) = 1w(v), (44) then ϕt coincides with the semigroup applied to 1w and we have the desired result ϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) = Ev[ϕ0( ˜Vt;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w)] = Pv[ ˜Vt = w].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is straightforward to confirm the initial condition ϕ0(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) = 1w(v) holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' So it remains to show that ϕt satisfies the PDE (44).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that ∂tϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) = −w v λ(v)e−Λ(v)t L2ϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) = λ(v) (ϕt(−v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) − ϕt(v;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w)) = λ(v) � λ(−w) − w v λ(−v)e−Λ(v)t Λ(v) − λ(−w) + w v λ(v)e−Λ(v)t Λ(v) � = −λ(v) � w v λ(−v)e−Λ(v)t Λ(v) + w v λ(v)e−Λ(v)t Λ(v) � = −λ(v)w v e−Λ(v)t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore we have that (44) holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting scheme DBD of ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let λi(x, v) = (vi∂iψ(x))+ + γi(x) and let Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 be verified.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then there exists a function V : R × {±1}d → R, and constants ρ ∈ (0, 1), C < ∞ such that for all (x, v) and all t ∈ (0, t0) with t0 < (1 + γ0)−1 P tV (x, v) = P D t P B 2tP D t V (x, v) ≤ (1 − ρt)V (x, v) + Ct.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (45) Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For a function g(x, v) conditioning on the event v = w and using Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we have P tg(x, v) = � w∈{±1}d g(x + vt + wt, w) d � i=1 � λi(x + vt, Fiw) + wi vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We now construct our Lyapunov function V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let β ∈ (0, 1/2), define φ(s) = 1 2sign(s) ln (1 + 2|s|) and V (x, v) = exp � βψ(x) + d � i=1 φ(vi∂iψ(x)) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is the same Lyapunov function defined in [8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For this function we have P tV (x, v) V (x, v) = � w∈{±1}d V (x + vt + wt, w) V (x, v) d � i=1 � λi(x + vt, Fiw) + wi vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (46) SPLITTING SCHEMES FOR PDMPS 49 By Taylor’s theorem there exists x1 = x1(x, v, w, t) ∈ B(x, t √ d) such that ψ(x + vt + wt) = ψ(x) + t⟨v + w, ∇ψ(x)⟩ + t2 2 (v + w)T ∇2ψ(x1)(v + w).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore we can rewrite (46) as P tV (x, v) V (x, v) = � w∈{±1}d e t2 2 (v+w)T ∇2ψ(x1)(v+w) d � i=1 et(vi+wi)β∂iψ(x)+φ(wi∂iψ(x+vt+wt))−φ(vi∂iψ(x)) × � λi(x + vt, Fiw) + wi vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (47) Since |φ′(s)| ≤ 1 for all s, by Taylor’s Theorem we have φ(wi∂iψ(x + vt + wt)) − φ(vi∂iψ(x)) ≤ |wi∂iψ(x + vt + wt) − wi∂iψ(x)| + φ(wi∂iψ(x)) − φ(vi∂iψ(x)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we can write P tV (x, v) V (x, v) ≤ � w∈{±1}d K1 d � i=1 I(i) (48) with K1 = e t2 2 (v+w)T ∇2ψ(x1)(v+w)e �d i=1|wi∂iψ(x+vt+wt)−wi∂iψ(x)| I(i) = et(vi+wi)β∂iψ(x)+φ(wi∂iψ(x))−φ(vi∂iψ(x)) λi(x + vt, Fiw) + wi vi λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Bound outside of a compact set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We split the product in (48) into four cases: (i) wi = vi and vi∂iψ(x + vt) > 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (ii) wi = vi and vi∂iψ(x + vt) < 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (iii) wi = −vi and vi∂iψ(x + vt) > 0;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (iv) wi = −vi and vi∂iψ(x + vt) < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider first case (i).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let i be such that wi = vi and vi∂iψ(x + vt) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then I(i) = e2tviβ∂iψ(x) λi(x + vt, Fiv) + λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using the form of λi we can write this as I(i) = γi(x + vt)e2βtvi∂iψ(x)(1 − e−(|∂iψ(x+vt)|+2γi(x+vt))t) |∂iψ(x + vt)| + 2γi(x + vt) + e−(|∂iψ(x+vt)|+2γi(x+vt))t+2tviβ∂iψ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using that 1 − e−z ≤ z for all z > 0 (note we make use of this inequality several times in the following computations) we find I(i) ≤ γi(x + vt)e2βtvi∂iψ(x)t + e−(|∂iψ(x+vt)|+2γi(x+vt))t+2tviβ∂iψ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(b) we can bound γi for |x| ≥ R with R sufficiently large and we have vi∂iψ(x+vt) ≥ 1 + γi(x + vt) which gives I(i) ≤ 1 + (vi∂iψ(x + vt) + γi(x + vt)) |∂iψ(x + vt)| + 2γi(x + vt) e−|∂iψ(x+vt)|t+2tviβ∂iψ(x) ≤ 2e−|∂iψ(x+vt)|t+2tviβ∂iψ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For case (ii), let i be such that wi = vi and vi∂iψ(x + vt) < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then I(i) = e2βtvi∂iψ(x) |∂iψ(x + vt)| + γi(x + vt) + γi(x + vt)e−(|∂iψ(x+vt)|+2γi(x+vt))t |∂iψ(x + vt)| + 2γi(x + vt) ≤ e2βtvi∂iψ(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 50 SPLITTING SCHEMES FOR PDMPS For case (iii), let i be such that wi = −vi and vi∂iψ(x + vt) > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then I(i) = eφ(−vi∂iψ(x))−φ(vi∂iψ(x)) λi(x + vt, v) − λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For s > 0 it holds that φ(−s) − φ(s) = − ln(1 + 2s) and hence I(i) = λi(x + vt, v) 1 + 2vi∂iψ(x) 1 − e−(λi(x+vt,v)+λi(x+vt,−v))t λi(x + vt, v) + λi(x + vt, Fiv) ≤ λi(x + vt, v) 1 + 2vi∂iψ(x)t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For case (iv), let i be such that wi = −vi and vi∂iψ(x + vt) < 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then I(i) = eφ(−vi∂iψ(x))−φ(vi∂iψ(x)) γi(x + vt) − γi(x + vt)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For s < 0 we have φ(−s) − φ(s) = ln(1 + 2|s|), and thus we obtain I(i) ≤ γi(x + vt)(1 + 2|∂iψ(x)|)t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Combining these estimates we have for |x| ≥ R with R sufficiently large P tV (x, v) V (x, v) ≤ � w∈{±1}d K1 � i:wi=vi, vi∂iψ>0 (γi(x + vt)e2βtvi∂iψ(x)t + e−(|∂iψ(x+vt)|+2γi(x+vt))t+2tviβ∂iψ(x)) × � i:wi=vi, vi∂iψ<0 e2βtvi∂iψ(x) � i:wi=−vi, vi∂iψ>0 λi(x + vt, v) 1 + 2vi∂iψ(x)t � i:wi=−vi, vi∂iψ<0 γi(x + vt)(1 + 2|∂iψ(x)|)t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now consider K1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Taylor’s theorem there exists x2 ∈ B(x, 2 √ dt) such that K1 ≤ exp � d � i=1 �t2 2 |((v + w)T ∇2ψ(x1))i| + t|(w∇2ψ(x2))i| � |vi + wi| � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using this bound and the four cases above we now obtain P tV (x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) V (x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) ≤ � i:vi∂iψ>0 e2t2|(vT ∇2ψ(x1))i|+ 2t 2 |(v∇2ψ(x2))i| (γi(x + vt)e2βtvi∂iψ(x)t + e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t) × � i:vi∂iψ<0 e � t2 2 (2|vT ∇2ψ(x1))i|+2t|(v∇2ψ(x2))i| � +2βtvi∂iψ(x) + � w∈{±1}d\\{v} t|{i:wi̸=vi}| × � i:wi=vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ>0 e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)(γi(x + vt)e2βtvi∂iψ(x)t + e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t) × � i:wi=vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ<0 e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)e2βtvi∂iψ(x) � i:wi=−vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ>0 λi(x + vt,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) 1 + 2vi∂iψ(x) × � i:wi=−vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ<0 e−t0|∂iψ(x+vt)|(1 + 2|∂iψ(x)|).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By (23) we have P tV (x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) V (x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) ≤ � i:vi∂iψ>0 (γ0t + e2t2|(vT ∇2ψ(x1))i|+ 2t 2 |(v∇2ψ(x2))i|e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t) × � i:vi∂iψ<0 e � t2 2 (2|vT ∇2ψ(x1))i|+2t|(v∇2ψ(x2))i| � +2βtvi∂iψ(x) SPLITTING SCHEMES FOR PDMPS 51 + � w∈{±1}d\\{v} t|{i:wi̸=vi}| � i:wi=vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ>0 (γ0t + e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)e−(1−2β)|∂iψ(x+vt)|t−2γi(x+vt)t) × � i:wi=vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ<0 e(t2(|(v+w)T ∇2ψ(x1))i|+2t|(w∇2ψ(x2))i|)e2βtvi∂iψ(x) � i:wi=−vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ>0 λi(x + vt,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) 1 + 2vi∂iψ(x) × � i:wi=−vi,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='vi∂iψ<0 e−t0|∂iψ(x+vt)|(1 + 2|∂iψ(x)|).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since β < 1/2, by Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(c) there exists β1 such that for |x| ≥ R with R sufficiently large P tV (x, v) V (x, v) ≤ � i:vi∂iψ>0 (γ0t + e−β1|∂iψ(x+vt)|t) � i:vi∂iψ<0 e−β1|∂iψ(x)|t + � w∈{±1}d\\{v} t|{i:wi̸=vi}| � i:wi=vi,vi∂iψ>0 (γ0t + e−β1|∂iψ(x+vt)|t) � i:wi=vi,vi∂iψ<0 e−β1t|∂iψ(x)| × � i:wi=−vi,vi∂iψ>0 λi(x + vt, v) 1 + 2vi∂iψ(x) � i:wi=−vi,vi∂iψ<0 e−t0|∂iψ(x+vt)|(1 + 2|∂iψ(x)|).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For |x| ≥ R with R sufficiently large λi(x + vt, v)/(1 + 2vi∂iψ(x)) ≤ 1 and by (23) we have γi(x)(1 + 2|∂iψ(x)|) ≤ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We also have that |∇ψ(x + vt)| ≥ M for any M > 0 for |x| ≥ R with R sufficiently large.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we have P tV (x, v) V (x, v) ≤ (γ0t + e−β1Mt)|{i:vi∂iψ(x+vt)>0}|e−β1Mt|{i:vi∂iψ(x+vt)<0}| + � w∈{±1}d\\{v} t|{i:wi̸=vi}| � i:wi=vi,vi∂iψ>0 (γ0t + e−β1Mt) � i:wi=vi,vi∂iψ<0 e−β1tM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since e−β1Mt ≤ γ0t + e−β1Mt we obtain P tV (x, v) V (x, v) ≤ (γ0t + e−β1Mt)|{i:vi∂iψ(x+vt)>0}|(γ0t + e−β1Mt)|{i:vi∂iψ(x+vt)<0}| + � w∈{±1}d\\{v} t|{i:wi̸=vi}|(γ0t + e−β1Mt)|{i:wi=vi,vi∂iψ(x+vt)>0}|(γ0t + e−β1Mt)|{i:wi=vi,vi∂iψ(x+vt)<0}|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence P tV (x, v) V (x, v) ≤ (γ0t + e−β1Mt)d + � w∈{±1}d: w̸=v t|{i:wi̸=vi}|(γ0t + e−β1Mt)d−|{i:wi̸=vi}| = � w∈{±1}d t|{i:wi̸=vi}|(γ0t + e−β1Mt)d−|{i:wi̸=vi}| = d � k=0 �d k � tk (γ0t + e−β1Mt)d−k = � (1 + γ0)t + e−β1Mt�d .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To show that (45) holds for |x| ≥ R it is sufficient to show that (1 + γ0)t + e−β1Mt < 1 − ρt for some ρ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed in that case 1 − ρt < 1 and thus ((1 + γ0)t + e−β1Mt)d < (1 − ρt)d < 1 − ρt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 52 SPLITTING SCHEMES FOR PDMPS Note that for t ≤ t0, with t0 ∈ [0, 1], it holds that e−β1Mt ≤ 1 − ct for c = 1−e−β1Mt0 t0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then for t ≤ t0 we have (1 + γ0)t + e−β1Mt ≤ 1 − t(c − 1 − γ0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then it is needed that c > 1 + γ0, that is t0 should be such that 1 − e−β1Mt0 t0 > 1 + γ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (49) Note we can always increase M by taking R larger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Choose M such that e−β1Mt0 < 1 − t0(1 + γ0), which is possible since t0 < (1 + γ0)−1, then (49) holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence (45) holds for |x| ≥ R with ρ = (1 − e−β1Mt0)t−1 0 − 1 − γ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Bound inside of a compact set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It remains to show that (45) holds for |x| ≤ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let C = {x : |x| ≤ R} × {±1}d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall t < 1 and ψ ∈ C2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall use the inequality etr ⩽ 1 + tr + t2r2er/2 ≤ 1 + t(r + e3r/2), which holds for for t ⩽ 1, r > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First of all we consider the term in the sum corresponding to the case w = v.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Bounding the probability of this event by 1 we find K1 d � i=1 I(i) ≤ et22vT ∇2ψ(x1)v+ 2 2 �d i=1|∂iψ(x+2vt)−∂iψ(x)|+2tβ⟨v,∇ψ(x)⟩ ≤ 1 + t(A(x, v, t) + e3A(x,v,t)/2), where A(x, v, t) = 2vT ∇2ψ(x1)v + (2/2) �d i=1|⟨v, ∇∂iψ(x2)⟩| + 2β⟨v, ∇ψ(x)⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Taking the maximum of A over (x, v, t) ∈ C × {±1}d × (0, 1) we find K1 d � i=1 I(i) ≤ 1 + tA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us now consider the remaining elements in the sum.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we take advantage that a velocity flip is an order t event.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider for the moment only the i-th component of the velocity vector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The probability that this is flipped (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' wi = −vi) satisfies λi(x + vt, Fiw) − λi(x + vt, v)e−(λi(x+vt,v)+λi(x+vt,Fiv))t λi(x + vt, v) + λi(x + vt, Fiv) ≤ λi(x + vt, v)(1 − e−t(|∂iψ(x+tv)|+2γi(x+tv))) |∂iψ(x + tv)| + 2γi(x + tv) ≤ tλi(x + vt, v) ≤ t max i=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=',d, (x,v)∈C×{±1}d, t∈(0,1) λi(x + vt, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Here we used that 1 − exp(−z) ≤ z for z ≥ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' All other probabilities can be bounded by 1 and hence � w̸=v K1 d � i=1 I(i) ≤ t max i=1,.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=',d, (x,v)∈C×{±1}d, t∈(0,1) λi(x + vt, v) � w̸=v V (x + vt + wt, w) V (x, v) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since V is continuous under our assumptions we proved that for every compact set C ×{±1}d there exists a constant B > 0 such that for all (x, v) ∈ C × {±1}d it holds P tV (x, v) ≤ (1 + tB)V (x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (50) Therefore we have (45) holds for all x ∈ Rd, v ∈ {±1}d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ SPLITTING SCHEMES FOR PDMPS 53 C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Equation (24).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us prove (24).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Fix a probability measure µ on Rd × {±1}d, and let (x, v) be a point in the support of µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we can construct the set D(x, v) corresponding to (x, v) and given by (21).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 there is a unique invariant measure πx,v δ for the Markov process with kernel P 2 δ , and by (20) for any probability measures ν, ν′ on D(x, v) ∥νP 2n δ − ν′P 2n δ ∥V ≤ C α ˜κnδ ∥ν − ν′∥V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Setting ν = δx,v, ν′ = πx,v δ and using that πx,v δ is an invariant measure for the kernel P 2 δ we have ∥δx,vP 2n δ − πx,v δ ∥V ≤ C α ˜κnδ ∥ν − ν′∥V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then integrating with respect to the probability measure µ we obtain ∥µP 2n δ − µπx,v δ ∥V = sup |g|≤V ���� � [P 2n δ g(x, v) − πx,v δ (g)]µ(dx, dv) ���� ≤ � sup |g|≤V ��P 2n δ g(x, v) − πx,v δ (g) �� µ(dx, dv) ≤ � ��δx,vP 2n δ − πx,v δ �� V µ(dx, dv) ≤ C α ˜κnδ � ∥δ(x,v) − πx,v δ ∥V µ(dx, dv).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Other splitting schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this Section we consider splitting schemes DRBRD and RDBDR of ZZS and prove geometric ergodicity in Theorem C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The minorisation and drift conditions are proved in Sections C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall work under the following assumption.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' There exists γ0 ∈ (0, ∞) such that the following conditions for the refreshment rate hold: (a) there exists R > 0 for which for any |x| ≥ R it holds that γ(x) d � j=1 (1 + |∂jψ(x)|) ≤ γ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) For |x| > R for some R > 0 it holds that sup t∈(0,1), y1,y2∈B(x,t √ d),v,w∈{−1,1}d γ(x + vt)et|∇ψ(x)|+ t2 2 (v+w)T ∇2ψ(y1)(v+w)+|∇ψ(y2)| d � i=1 (1 + 2|∂iψ(x)|) ≤ γ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Theorem C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider splitting schemes DRBRD and RDBDR of ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose Assumption 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a), (c) holds for switching rates λi(x, v) = (vi∂iψ(x))+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose moreover that the refreshment rate γ satisfies Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a) for scheme RDBDR and Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(b) for scheme DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then statements (1) and (2), as well as Equation (24), hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular (2) holds with δ0 < 2(1 + 2γ0 + γ2 0)−1 for RDBDR and with δ0 < 2(1 + 2γ0)−1 for DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Minorisation condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The chain obtained by DRBRD has the same periodic behaviour of DBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence this case can be treated in the same way and a minorisation condition follows by the same reasoning used in Section C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 for splitting DBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 54 SPLITTING SCHEMES FOR PDMPS Splitting RDBDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case we give a sketch of the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The chain obtained by RDBDR breaks the grid behaviour exhibited by DBD because of the two refreshment steps at the beginning and end of each step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed, consider the one-dimensional case and recall the definition of the grid D(x, v) as in Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since v = ±1, there are two disjoint grids: D(x, +1) and D(x, −1), with the idea being that after even steps of DBD the process lives on the same grid it started from.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' However, the process RDBDR can swap between one grid and the other by having a velocity refreshment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed, starting the process at (x, +1) and having a velocity flip due to a refreshment at the end of the first step and having no other jumps, we find the state of the process is (X2, V2) = (x, −1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore after even steps this process lives on the grid D(x) = {y : y = x + mδ, m ∈ Z}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If the initial and final condition are on the same grid D(x, v), then no refreshment is required and one can simply use the proof of the scheme DBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' On the other hand, if the two states are on different grids, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' one is on D(x, +1) and the other on D(x, −1), then a refreshment is required to choose the right grid.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order to maintain the right dependence on the step size δ it is required to give the process additional ⌈ M δ ⌉ iterations, for a constant M > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Indeed with this modification the probability of having a refreshment in the first ⌈ M δ ⌉ is constant has a lower bound which is independent of δ, assuming δ ≤ δ0 for some δ0 > 0 (see for instance the second case in the proof of Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' After the first ⌈ M δ ⌉ iterations the process is on the right grid and Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 can be applied with the further constraint that no (more) refreshments take place.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that this event again has a lower bounded probability independent of δ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since in the first ⌈ M δ ⌉ iterations the process can go out of the initial compact set C = [−R, R], it follows that the Lemma should be applied with set ˜C = [−R−M, R+M].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The extension to the multidimensional case follows by applying this same intuition to every com- ponent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Drift condition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us start with an auxiliary result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose the refreshment rate γ satisfies Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then P R t V ≤ (1+γ0t)V + Mt, where γ0 is as in Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and M independent of t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let V be as in Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Applying the transition kernel P R t to V we find P R t V (x, v) = V (x, v)e−tγ(x) + 1 2d (1 − e−tγ(x)) � w̸=v V (x, w) = V (x, v)e−tγ(x) + (1 − e−tγ(x))V (x, v) 1 2d � w̸=v V (x, w) V (x, v) = V (x, v) � e−tγ(x) + (1 − e−tγ(x)) 1 2d � w̸=v � j:vj̸=wj (1 + |∂jψ(x)|) � ≤ V (x, v) � �e−tγ(x) + tγ(x) d � j=1 (1 + |∂jψ(x)|) � � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Clearly for x inside of a compact set this implies P R t V (x, v) ≤ (1 + Bt)V (x, v) by taking maximum over x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Outside of a compact set we use Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 to obtain P R t V (x, v) ≤ V (x, v)(1 + tγ0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting scheme RDBDR of ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose Assumptions 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(c) and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(a) hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then there exist a function V and constants ρ ∈ (0, 1), C > 0 such that for any t ∈ (0, t0) with t0 < (1 + 2γ0 + γ2 0)−1 it holds that P R t P D t P B 2tP D t P R t V (x, v) ≤ (1 − ρt)V (x, v) + Ct.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 55 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let V be as in Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In the current context the result of the Lemma is that for all t ∈ (0, t0) with t0 < 1 it holds that P D t P B 2tP D t V (x, v) ≤ (1−ρt)V (x, v)+Bt where ρ = (1−e−Rt0)t−1 0 −1 for R sufficiently large such that ρ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Applying Lemmas C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7 we obtain P R t P D t P B 2tP D t P R t V (x, v) ≤ (1 + tγ0)P R t P D t P B 2tP D t V (x, v) + Mt ≤ (1 + tγ0)(1 − ρt)P R t V (x, v) + t(M + (1 + γ0)B) ≤ (1 + tγ0)2(1 − ρt)V (x, v) + t(M(2 + γ0) + (1 + γ0)B).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is left to ensure that (1 + tγ0)2(1 − ρt) ≤ (1 − ˜ρt) for ˜ρ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We have (1 + tγ0)2(1 − ρt) ≤ (1 − t(ρ − 2γ0 − γ2 0)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence it is needed that (1 − e−Rt0) t0 − 1 > 2γ0 + γ2 0 and thus that e−Rt0 < 1 − t0(1 + 2γ0 + γ2 0), which is valid as R can be taken as large as needed and t0 < (1 + 2γ0 + γ2 0)−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the splitting scheme DRBRD of ZZS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Suppose Assumptions 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(c) and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then there exist a function V and constants ρ ∈ (0, 1), C > 0 such that for any t ∈ (0, t0) with t0 < (1 + 2γ0)−1 it holds that P D t P R t P B 2tP R t P D t V (x, v) ≤ (1 − ρt)V (x, v) + Ct.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let V be as in Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that by Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7 we have that P R t P D t V (x, v) = P R t V (x + vt, v) ≤ (1 + γ0t)V (x + vt, v) + Mt and thus P R t P D t V (x, v) ≤ (1 + γ0t)P D t V (x, v) + Mt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then P D t P R t P B 2tP R t P D t V (x, v) ≤ (1 + γ0t)P D t P R t P B 2tP D t V (x, v) + Mt and P D t P R t P B 2tP D t V (x, v) = e−tγ(x+vt)P D t P B 2tP D t V (x, v) + V (x, v)(1 − e−tγ(x+vt)) � w∈{±1}d 1 2d V (x + vt + wt, w) V (x, v) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The first term corresponds to the case of no refreshments, while in the second term a refreshment takes place.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the first term we can directly apply Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4, which in the current context shows that for t < t0 < 1 it holds P D t P B 2tP D t V (x, v) ≤ (1 − ρt)V (x, v) + Mt for ρ = (1 − e−β1Mt0)t−1 0 − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The second term can be rewritten as in (47), that is for x1 = x1(x, v, w, t) ∈ B(x, t √ d) V (x + vt + wt, w) V (x, v) = exp � β(ψ(x + vt + wt) − ψ(x)) + d � i=1 (φ(wi∂iψ(x + vt + wt)) − φ(vi∂iψ(x))) � = et|∇ψ(x)|+ t2 2 (v+w)T ∇2ψ(x1)(v+w)+|∇ψ(x2)| d � i=1 (1 + 2|∂iψ(x)|) Using Assumption C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5(b) we find (1 − e−tγ(x+vt)) � w∈{±1}d 1 2d V (x + vt + wt, w) V (x, v) ≤ tγ(x + vt) � w∈{±1}d 1 2d V (x + vt + wt, w) V (x, v) ≤ tγ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 56 SPLITTING SCHEMES FOR PDMPS Therefore we have shown P D t P R t P B 2tP R t P D t V ≤ (1 + γ0t)(1 − ρt + tγ0)V + ˜ Mt ≤ (1 − t(ρ − 2γ0))V + ˜ Mt.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence it is sufficient to ensure that ρ > 2γ0, which can be done similarly to the proof of Lemma C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Appendix D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 and related results In this section we collect statements and proofs that are not included in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this section we prove Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We start by focusing on the left hand side of (26), i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ BPS(µf2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We find since µ is rotationally invariant in v L∗ BPS(µf2)(x, v) = µ(x, v) � ⟨v, ∇ψ(x)⟩f2(x, v) − ⟨v, ∇xf2(x, v)⟩ + (−⟨v, ∇ψ(x)⟩)+f2(x, R(x)v) − ⟨v, ∇ψ(x)⟩+f2(x, v) + λr � f2(x, y)ν(dy) − λrf2(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We shall consider the case of v = ±1, hence ν = (1/2)δ+1 +(1/2)δ−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular this choice satisfies Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Introduce the notation f+ 2 (x) = f2(x, 1), f− 2 (x) = f2(x, −1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We have in the 1-dimensional setting L∗ BPS(µf2)(x, +1) = −µ(x, +1) � (f+ 2 )′(x) + ((−ψ′(x))+ + λr/2)f+ 2 (x) − (λr/2 + (−ψ′(x))+)f− 2 (x) � , L∗ BPS(µf2)(x, −1) = +µ(x, −1) � (f− 2 )′(x) + ((+ψ′(x))+ + λr/2)f+ 2 (x) − (λr/2 + (+ψ′(x))+)f− 2 (x) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Define function h such that hµ = L∗ 2µ, and also h+(x) = h(x, +1) and h−(x) = h(x, −1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore we wish to solve the following system of ODEs � (f+ 2 )′(x) = −(λr/2 + (−ψ′(x))+)f+ 2 (x) + (λr/2 + (−ψ′(x))+)f− 2 (x) − h+(x), (f− 2 )′(x) = −(λr/2 + (+ψ′(x))+)f+ 2 (x) + (λr/2 + (+ψ′(x))+)f− 2 (x) + h−(x), (51) with compatibility condition (27), which in this case can be written as � ∞ −∞ (f+ 2 (x) + f− 2 (x))π(x)dx = 0 (52) with π(x) = µ(x, 1) + µ(x, −1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us find a solution to (51) for a generic (continuous and locally lipschitz) function h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Start by subtracting the first line to the second line in (51): (f− 2 )′(x) − (f+ 2 )′(x) = ((ψ′(x))+ − (−ψ′(x))+)(f− 2 (x) − f+ 2 (x)) + hs(x), (53) where hs(x) = h+(x) + h−(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Define g = f− 2 − f+ 2 and notice that (ψ′(x))+ − (−ψ′(x))+ = ψ′(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then we can rewrite (53) as g′(x) = ψ′(x)g(x) + hs(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Solving this ODE using an integrating factor we find g(x) = exp (ψ(x)) lim y→−∞ [exp (−ψ(y)) g(y)] + exp (ψ(x)) � x −∞ hs(y) exp(−ψ(y))dy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recall that g = f− − f+ and f+, f− satisfy (52).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In order for f2 to define a proper density we require � ∞ −∞ g(x)π(x)dx < ∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For this to hold it must be that limy→−∞ exp (−ψ(y)) g(y) = 0 and thus g(x) = exp (ψ(x)) � x −∞ hs(y) exp(−ψ(y))dy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (54) SPLITTING SCHEMES FOR PDMPS 57 Since f− 2 (x) = f+ 2 (x) + g(x) and plugging this in the first equation of (51) we obtain the ODE (f+ 2 )′(x) = (λr/2 + (−ψ′(x))+)g(x) − h+(x) which can be integrated as f+ 2 (x) = f+ 2 (0) + � x 0 � (λr/2 + (−ψ′(y))+)g(y) − h+(y) � dy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (55) It follows that f− 2 (x) = f+ 2 (0) + � x 0 � (λr/2 + (−ψ′(y))+)g(y) − h+(y) � dy + exp (ψ(x)) � x −∞ hs(y) exp(−ψ(y))dy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (56) Finally we compute f+ 2 (0) enforcing the compatibility condition (52).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Plugging (55) and (56) in (52) we find the condition f+ 2 (0) = − � ∞ −∞ � g(x)/2 + � x 0 � (λr/2 + (−ψ′(y))+)g(y) − h+(y) � dy � π(x)dx.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (57) D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Fix x ∈ R, δ > 0 and let G(x, δ) := {(z, v) ∈ R×{±1} : (z −x)/δ ∈ Z} be the state space of the chain with initial position x.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For now, let µδ be any probability measure on G(x, δ) such that µδ(y, w) = µδ(y, −w) for all (y, w) ∈ G(x, δ), and let us give a sufficient and necessary condition for it to be invariant by the chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since such a µδ is invariant by the refreshment step, it is invariant for the scheme RDBDR if and only if it is invariant for the scheme R’DBD, where R’ is a deterministic flip of the velocity (which, as R, preserves µδ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Besides, from a state (y, w) ∈ G(x, δ), one transition of R’DBD can only lead to (y, w) or (y + δw, −w), from which it can only stay or come back to the initial (y, w).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In other words the pair {(y, w), (y+δw, −w)} is irreducible for this chain, and thus µδ is invariant for R’DBD if and only if its restrictions on all these sets for (y, w) ∈ G(x, δ) are invariant by this scheme, which by definition reads ∀(y, w) ∈ G(x, δ) , µδ(y, w)e−δλ(y+wδ/2,w) = µδ(y + δw, −w)e−δλ(y+wδ/2,−w).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It turns out that this is exactly the skew detailed balance condition (6) for the scheme DBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Writing that µδ(y, w) ∝ exp(−ψδ(y)) for some ψδ and recalling that λ(y, w) − λ(y, −w) = ψ′(y) for all y, w, this is equivalent to ∀(y, w) ∈ G(x, δ), ψδ(y + δw) − ψδ(y) = δψ′ (y + δw/2) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Up to an additive constant, the only function ψδ which satisfies this is the one given in the statement of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' As a conclusion, we have proven that a probability measure on G(x, δ) which is independent from the velocity is invariant for the scheme RDBDR if and only if it is the one given in the proposition, which concludes the proof of the first statement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now we focus on the convergence of empirical means, assuming that the conditions of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 are met.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The reference position x ∈ R is still fixed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The long-time convergence established in Theo- rem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6 (for P 2 δ where Pδ is one step of the scheme) is well-known to imply an ergodic Law of Large Numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, for all initial conditions in G(x, δ) and all bounded f, distinguishing odd and even indexes, we see that 1 N �N k=1 f(Ztk) (where (Ztk)k∈N is a trajectory of the scheme) converges almost surely as N → ∞ to ˜µδ(f) := (µ′ δ(f) + µ′′ δ(f))/2, where µ′ δ and µ′′ δ are the unique invariant measures of P 2 δ on each periodic component of the state space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular, ˜µδ is an invariant measure for Pδ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In dimension 1, the scheme DBD is such that for all y, for all times, the number of visits of the points (y, 1) and (y, −1) differ at most by 1, which implies by ergodicity that ˜µδ(y, w) = µδ(y, −w) for all (y, w) ∈ G(x, δ), and we conclude thanks to the first part of the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 58 SPLITTING SCHEMES FOR PDMPS (a) Refreshment rate λr = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (b) Refreshment rate λr = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Figure 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Plots of the theoretical invariant measure up to second order for a standard Gaussian target, as given by Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The step size is δ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Application of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 to three one-dimensional targets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this section we give the function f2 corresponding to the three cases considered in Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Gaussian target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us start with a one-dimensional Gaussian target with mean zero and variance σ2 > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let ψ(x) = x2/(2σ2) for σ2 > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then: For the splitting scheme DBRBD it holds that f2(x, +1) = f2(x, −1) = λr 24 � 2 √ 2 σ√π − x3 σ4 sign(x) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme BDRDB it holds that f2(x, +1) = 1 8σ2 − 1 4σ4 x21x<0, f2(x, −1) = 1 8σ2 − 1 4σ4 x21x>0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme RDBDR it holds that f2(x, +1) = f2(x, −1) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme DRBRD it holds that f2(x, +1) = f2(x, −1) = λr 12 � 2 √ 2 σ√π − |x|3 σ4 � + λ2 r 16 � 1 − x2 σ2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recalling that v ∈ {−1, +1}, for all splitting schemes we can compute the functions h = (L∗ 2µ)/µ: hDBRBD(x, v) = λr 8σ4 (x2 + 2vx(−vx)+), hBDRDB(x, v) = 1 8σ6 � −λrσ2(x2 + 2vx(−vx)+) + 2(−vx)+(x2 − 2σ2) � , hRDBDR(x, v) = 0, SPLITTING SCHEMES FOR PDMPS 59 hDRBRD(x, v) = 1 8σ4 λr � x2 + vx(3(−vx)+ + (vx)+) + λrvxσ2� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DBRBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that hs(x) = λr 4σ4 (x2 + x((−x)+ − (x)+)) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence by (54) it holds g(x) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then by (55) f+ 2,DBRBD(x) = f+ 2 (0) − � x 0 λr 8σ4 (y2 + 2y(−y)+)dy = f+ 2 (0) − λr 24σ4 x3(1 − 21x<0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since g = 0 we have f+ 2,DBRBD = f− 2,DBRBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To find f+ 2 (0) we enforce (52): � +∞ −∞ f+ 2 (x)π(x)dx = f+ 2 (0) − λr 24σ4 � +∞ −∞ x3(1 − 21x<0)π(x)dx = f+ 2 (0) − λr 12σ4 � +∞ 0 x3π(x)dx = f+ 2 (0) − λr 12σ4 σ3 � 2 π = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Clearly this is satisfied for f+ 2 (0) = λr/(6σ √ 2π).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting RDBDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Clearly in this case hs(x) = 0, hence g(x) = 0 and f+ 2,RDBDR = f− 2,RDBDR = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting BDRDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We have hs(x) = 1 4σ6 |x|(x2 − 2σ2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now inserting this into the expression for g we get g(x) = 1 4σ6 exp(x2/(2σ2)) � x −∞ |y|(y2 − 2σ2) exp(−x2/(2σ2))dy = 1 4σ6 exp(x2/(2σ2))(−σ2sign(x) exp(−x2/(2σ2))x2) = − 1 4σ4 x2sign(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We compute f+ 2 by applying (55).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First observe that � x 0 h+(y)dy = − λr 8σ4 � x 0 y2sign(y)dy + 1 4σ6 � x 0 (−y)+(y2 − 2σ2)dy and � x 0 (λr/2 + (−y/σ2)+)g(y)dy = − λr 8σ4 � x 0 y2sign(y)dy − 1 4σ6 � x 0 (−y)+y2sign(y)dy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore we obtain f+ 2,BDRDB(x) = f+ 2 (0) + 1 2σ4 � x 0 (−y)+dy = f+ 2 (0) + 1 4σ4 x21x<0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Enforcing the compatibility condition (57) we obtain f+ 2 (0) = 1/(8σ2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 60 SPLITTING SCHEMES FOR PDMPS Splitting DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Similarly to the case of DBRBD observe that hs = 0 and thus g(x) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that h+ DRBRD(x) = λr 8σ4 (2x2sign(x) + λrxσ2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then by (55) f+ 2,DRBRD(x) = f+ 2 (0) − λr 8σ4 � x 0 (2y2sign(y) + σ2λry)dy = f+ 2 (0) − λr 12σ4 x3sign(x) − λ2 r 16σ2 x2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To find f+ 2 (0) we enforce (57): f+ 2 (0) = λr 6σ � 2 π + λ2 r 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Non-Lipschitz potential.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now we focus on a target distribution with non-Lipschitz potential.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let ψ(x) = x4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then: For the splitting scheme DBRBD it holds that f2(x, +1) = f2(x, −1) = λr 7 � 1 2Γ(5/4) − 2x7sign(x) � + 1 2 �Γ(3/4) Γ(1/4) − x2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme BDRDB it holds that f2(x, +1) = 5Γ(3/4) 2Γ(1/4) − x2 − 4x61x<0 , f2(x, −1) = 5Γ(3/4) 2Γ(1/4) − x2 − 4x61x≥0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme RDBDR it holds that f2(x, +1) = f2(x, −1) = Γ(3/4) 2Γ(1/4) − 1 2x2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme DRBRD it holds that f2(x, +1) = f2(x, −1) = λr 7 � 1 Γ(5/4) − 4x7sign(x) � + 1 2 �Γ(3/4) Γ(1/4) − x2 � + λ2 r 8 �1 4 − x4 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we obtain hDBRBD(x, v) = +2λr(x6 + 2vx3(−vx3)+) + vx, hBDRDB(x, v) = −2λr(x6 + 2vx3(−vx3)+) + 8(−vx3)+(−3x2 + 2x6) − 2vx, hRDBDR(x, v) = vx, hDRBRD(x, v) = +2λr(x6 + vx3(3(−vx3)+ + (vx3)+)) + vx + (λ2 rvx3)/2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denote the normalisation constant of the target π(x) by Z = 2Γ(5/4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DBRBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since hs(x) = 0 we have f+ 2,DBRBD(x) = f− 2,DBRBD(x) = f+ 2 (0) − 2 7λrx7sign(x) − 1 2x2, with f+ 2 (0) = 4λr 7 � ∞ 0 x7π(x)dx + � ∞ 0 x2π(x)dx = λr 14Γ(5/4) + Γ(3/4) 2Γ(1/4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 61 Splitting BDRDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case hs(x) = 8x2|x3|(2x4 − 3) and thus we find g(x) = −4x6sign(x).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It follows that f+ 2,BDRDB(x) = f+ 2 (0) − 4x61x<0 − x2, f− 2,BDRDB(x) = f+ 2 (0) − 4x61x≥0 − x2, where f+ 2 (0) = Γ(7/4) 2Γ(5/4) + Γ(3/4) Γ(1/4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting RDBDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since hs(x) = 0 we have f+ 2,RDBDR(x) = f− 2,RDBDR(x) = f+ 2 (0) − x2/2 with f+ 2 (0) = Γ(3/4) 2Γ(1/4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since hs(x) = 0 we have f+ 2,DRBRD(x) = f− 2,DRBRD(x) = f+ 2 (0) − 4 7λrx7sign(x) − 1 2x2 − 1 8λ2 rx4, with f+ 2 (0) = λr 7Γ(5/4) + Γ(3/4) 2Γ(1/4) + 1 32λ2 r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Heavy tailed target.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally we consider a Cauchy distribution π(x) = γ/(π(γ2 +x2)) for γ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let ψ(x) = ln(γ2 + x2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then: For the splitting scheme DBRBD it holds that f2(x, +1) = f2(x, −1) = λr 4γ �π 4 − |arctan(x/γ)| + γ|x| γ2 + x2 − 1 π � + 1 12 � 1 4γ2 + x2 − γ2 (γ2 + x2)2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme BDRDB it holds that f2(x, v) = � (x2 − 3γ2)2 48γ2(x2 + γ2)2 � 1xv<0 + �x4 − 54x2γ2 + 9γ4 48γ2(x2 + γ2)2 � 1xv≥0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme RDBDR it holds that f2(x, +1) = f2(x, −1) = 1 12 � 1 4γ2 + x2 − γ2 (γ2 + x2)2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For the splitting scheme DRBRD it holds that f2(x, +1) = f2(x, −1) = λr 2γ �π 4 − |arctan(x/γ)| + γ|x| γ2 + x2 − 1 π � + 1 12 � 1 4γ2 + x2 − γ2 (γ2 + x2)2 � + λ2 r 8 � ln 4 − ln � 1 + x2 γ2 �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 we obtain hDBRBD(x, v) = λr 2(γ2 + x2)2 (x2 + 2vx(−vx)+) + 1 24vψ(3)(x), hBDRDB(x, v) = − λr 2(γ2 + x2)2 (x2 + 2vx(−vx)+) + 2 (γ2 + x2)3 (−vx)+(−γ2 + 2x2) − 1 12vψ(3)(x), hRDBDR(x, v) = 1 24vψ(3)(x), 62 SPLITTING SCHEMES FOR PDMPS hDRBRD(x, v) = λr 2(γ2 + x2)2 � x2 + vx(3(−vx)+ + (vx)+) � + 1 24vψ(3)(x) + λ2 r xv 4(γ2 + x2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Denote the normalisation constant of the target π(x) by Z = π/γ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DBRBD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since hs(x) = 0 we have f+ 2,DBRBD(x) = f− 2,DBRBD(x) = f+ 2 (0) − λr 4 sign(x) �arctan(x/γ) γ − x γ2 + x2 � − 1 12 � γ2 − x2 (γ2 + x2)2 − 1 γ2 � , with f+ 2 (0) = λr 4γ � π 4 − 1 π � − 1 16γ2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting BDRDB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this case hs(x) = 2(2x2 − γ2)|x|/(γ2 + x2)3 and thus we find g(x) = −x2sign(x)/(γ2 + x2)2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It follows that f+ 2 (0) = 3 16γ2 and f+ 2,BDRDB(x) = � (x2 − 3γ2)2 48γ2(x2 + γ2)2 � 1x<0 + �x4 − 54x2γ2 + 9γ4 48γ2(x2 + γ2)2 � 1x≥0, f− 2,BDRDB(x) = � (x2 − 3γ2)2 48γ2(x2 + γ2)2 � 1x>0 + �x4 − 54x2γ2 + 9γ4 48γ2(x2 + γ2)2 � 1x<0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting RDBDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since hs(x) = 0 we have f+ 2,RR(x) = f− 2,RDBDR(x) = 1 12γ2 �1 4 − 1 � − 1 12 � γ2 − x2 (γ2 + x2)2 − 1 γ2 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Splitting DRBRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since hs(x) = 0 we have f+ 2,DRBRD(x) = f− 2,DRBRD(x) = f+ 2 (0) − λr 2 sign(x) �arctan(x/γ) γ − x γ2 + x2 � − 1 12 � γ2 − x2 (γ2 + x2)2 − 1 γ2 � − λ2 r 8 ln � 1 + x2 γ2 � , with f+ 2 (0) = λr 2γ � π 4 − 1 π � + λ2 r 8 ln 4 − 1 16γ2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Appendix E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3 In Section E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 we obtain the first and second order commutators of BPS, while in Section E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 we use the BCH formula and the obtained results to prove Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Computing the commutators of BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In this section we compute the first and second order commutators for the various components of the adjoint of the BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 we write down the commutator of the BPS and its decomposition in the three terms that represent the free transport, reflection mechanism, and velocity refreshments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In Section E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 we start with first order commutators, which are essential to compute second order commutators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The latter are computed in Section E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Now let us write the following identities, which form a lemma for convenience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' These will be used countless times in the computation of the commutators below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 63 Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For λ(x, v) = ⟨v, ∇ψ(x)⟩+ it holds that λ1(x, R(x)v) − λ1(x, v) = −⟨v, ∇ψ(x)⟩, (58) λ1(x, R(x)v) + λ1(x, v) = +|⟨v, ∇ψ(x)⟩|.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (59) The proof is trivial.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The adjoint of BPS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the generator Lf(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)=⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇xf(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)⟩ + λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)[f(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) − f(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)] + λ2 � � f(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' w) − f(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � ν(dw) Then one obtains that the adjoint is given by L∗ BPSg(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = −⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇xg(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)⟩ + ((gλ1)(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) − (gλ1)(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)) + λr � ν(v) � g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' y)dy − g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � = (L∗ D + L∗ B + L∗ R)g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' where L∗ Dg(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = −⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇xg(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)⟩,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ Bg(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v)λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) − g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ Rg(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = λr � ν(v) � g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' y)dy − g(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � Here the letters D,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' B,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' and R stand for drift,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' bounce,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' refreshment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' If we take g to be the invariant measure of BPS, µ, then L∗ Dµ(x, v) = ⟨v, ∇ψ(x)⟩µ(x, v), L∗ Bµ(x, v) = −⟨v, ∇ψ(x)⟩µ(x, v), L∗ Rµ(x, v) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' To obtain L∗ Dµ(x, v) we used the trivial, but useful, identity ∇xµ(x, v) = −∇ψ(x)µ(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' First order commutators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us start computing the three first order commutators [L∗ B, L∗ D], [L∗ R, L∗ D], and [L∗ R, L∗ B], which are essential to compute higher order commutators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This is done below respectively in Lemmas E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let g be a suitable function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ B, L∗ D]g(x, v) = −⟨R(x)v, (∇xg)(x, R(x)v)⟩λ1(x, R(x)v) + ⟨v, ∇xg(x, v)⟩λ1(x, v) + ⟨v, ∇x � g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v) � ⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular if g = µ [L∗ B, L∗ D]µ(x, v) = µ(x, v) � ⟨v, ∇ψ(x)⟩ � ⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩| � − ⟨v, ∇2ψ(x))v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Alternative ways to write [L∗ B, L∗ D]µ(x, v) can be found using the identities in Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We find [L∗ B, L∗ D]µ(x, v) = µ(x, v) � λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩ � = µ(x, v) � λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' 64 SPLITTING SCHEMES FOR PDMPS Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We have [L∗ B, L∗ D]g(x, v) = L∗ B(−⟨v, ∇xg(x, v)⟩) − L∗ D(g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v)) = −⟨R(x)v, (∇xg)(x, R(x)v)⟩λ1(x, R(x)v) + ⟨v, ∇xg(x, v)⟩λ1(x, v) + ⟨v, ∇x � g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v) � ⟩ and hence [L∗ B, L∗ D]µ(x, v) = −µ(x, v)⟨v, ∇ψ(x)⟩(λ1(x, R(x)v) + λ1(x, v)) + ⟨v, ∇x � µ(x, v)(λ1(x, R(x)v) − λ1(x, v)) � ⟩ = µ(x, v)(λ2 1(x, R(x)v) − λ2 1(x, v)) − ⟨v, ∇x � µ(x, v)⟨v, ∇ψ(x)⟩ � ⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then note that ⟨v, ∇x � µ(x, v)⟨v, ∇ψ(x)⟩ � ⟩ = ⟨v, ∇2ψ(x)v − ∇ψ(x)⟨v, ∇ψ(x)⟩ ⟩ µ(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' and hence [L∗ B, L∗ D]µ(x, v) = µ(x, v) � ⟨v, ∇ψ(x) � ⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩| � ⟩ − ⟨v, ∇2ψ(x))v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let g be a suitable function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ R, L∗ D]g(x, v) = λrν(v) � ⟨v, � ∇xg(x, y)dy⟩ − � ⟨y, ∇xg(x, y)⟩dy � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular if g = µ [L∗ R, L∗ D]µ(x, v) = −λr⟨v, ∇ψ(x)⟩µ(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We find [L∗ R, L∗ D]g(x, v) = −L∗ R(⟨v, ∇xg(x, v)⟩) − L∗ D � λr � ν(v) � g(x, y)dy − g(x, v) �� = −λr � ν(v) � ⟨y, ∇xg(x, y)⟩dy − ⟨v, ∇xg(x, v)⟩ � + λr⟨v, ∇x(ν(v) � g(x, y)dy − g(x, v))⟩ = λrν(v) � ⟨v, � ∇xg(x, y)dy⟩ − � ⟨y, ∇xg(x, y)⟩dy � and thus [L∗ R, L∗ D]µ(x, v) = L∗ R(⟨v, ∇ψ(x)⟩µ(x, v)) = −λr⟨v, ∇ψ(x)⟩µ(x, v) □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let g be a suitable function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ R, L∗ B]g(x, v) = λr � ν(v) � (g(x, R(x)y)λ1(x, R(x)y) − g(x, y)λ1(x, y))dy SPLITTING SCHEMES FOR PDMPS 65 + � ν(v) � g(x, y)dy � ⟨v, ∇ψ(x)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular if g = µ [L∗ R, L∗ B]µ(x, v) = λr⟨v, ∇ψ(x)⟩µ(x, v).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Compute [L∗ R, L∗ B]g(x, v) = L∗ R(g(x, R(x)v)λ1(x, R(x)v) − g(x, v)λ1(x, v)) − L∗ B � λr � ν(v) � g(x, y)dy − g(x, v) �� = λr � ν(v) � (g(x, R(x)y)λ1(x, R(x)y) − g(x, y)λ1(x, y))dy − g(x, R(x)v)λ1(x, R(x)v) � �� � A + g(x, v)λ1(x, v) � �� � B � − λr � � �ν(R(x)v) � g(x, y)dy − g(x, R(x)v) � �� � A � � λ1(x, R(x)v) − � �ν(v) � g(x, y)dy − g(x, v) � �� � B � � λ1(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It is now sufficient to cancel out the terms denoted by A and B that appear twice with opposite signs to obtain the final statement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' For g = µ [L∗ R, L∗ B]µ(x, v) = L∗ R(−⟨v, ∇ψ(x)µ(x, v)) = −λr � ν(v) � ⟨y, ∇ψ(x)⟩p(x, y)dy − ⟨v, ∇ψ(x)⟩µ(x, v) � which concludes by Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Note E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It follows from Lemmas E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 and E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='5 that [L∗ R, L∗ B + L∗ D]µ(x, v) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (60) E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Higher order commutators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us now compute higher order commutators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ B, [L∗ R, L∗ D]]µ(x, v) = λrµ(x, v) � ⟨v, ∇ψ(x)⟩ � λ1(x, R(x)v) + λ1(x, v) � + b tr � ∇ψ(x)(∇ψ(x))T − ∇2ψ(x) �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Applying Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 we obtain [L∗ B,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ D]]µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = −λrL∗ B � ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � + [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ D] � ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � = −λr � ⟨R(x)v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) − ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � + λrν(v) � ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇x � (⟨y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' y))dy⟩ − � ⟨y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇x(⟨y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' y))⟩dy � = λrµ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v)⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩ � λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) + λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � 66 SPLITTING SCHEMES FOR PDMPS − λrµ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � � (⟨y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇2ψ(x)y⟩ − ⟨y,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩2)ν(dy) � = λrµ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩ � λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) + λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � + b tr � ∇ψ(x)(∇ψ(x))T − ∇2ψ(x) �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Note that in the last line we used that ⟨a, b⟩2 = ⟨a, bbT a⟩ and that � ⟨y, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))y⟩ν(dy) = d � j=1 d � ℓ=1 (∇ψ(x)∇ψ(x)T − ∇2ψ(x))jℓ � (yjyℓ)ν(dy) = b tr � ∇ψ(x)∇ψ(x)T − ∇2ψ(x) � (61) which is a consequence of Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ R, [L∗ R, L∗ B]]µ(x, v) = −λ2 rµ(x, v)⟨v, ∇ψ(x)⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Next we compute [L∗ R, [L∗ R, L∗ B]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Since L∗ Rµ(x, v) = 0 we easily find [L∗ R, [L∗ R, L∗ B]]µ(x, v) = L∗ R(λr⟨v, ∇ψ(x)⟩µ(x, v)) = −λ2 rµ(x, v)⟨v, ∇ψ(x)⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ R, [L∗ R, L∗ D]]µ(x, v) = λ2 rµ(x, v)⟨v, ∇ψ(x)⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The result follows from Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='8 and (60).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ R, [L∗ B, L∗ D]]µ(x, v) = λrµ(x, v) � b tr � ∇ψ(x)∇ψ(x)T − ∇2ψ(x) � − � λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩ �� Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Taking advantage of Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 [L∗ R, [L∗ B, L∗ D]]µ(x, v) = L∗ R � µ(x, v) � λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩ �� = λrµ(x, v) � � � λ2 1(x, R(x)y) − λ2 1(x, y) + ⟨y, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))y⟩ � ν(dy) − � λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, (∇ψ(x)∇ψ(x)T − ∇2ψ(x))v⟩ �� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that for A = {y : ⟨y, ∇ψ(x)⟩ ≥ 0} we have � (λ2 1(x, R(x)y) − λ2 1(x, y))ν(dy) = � AC⟨y, ∇ψ(x)⟩2ν(y)dy − � A ⟨y, ∇ψ(x)⟩2ν(y)dy = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' This can be seen by the change of variables y′ = R(x)y in the first integral.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The result then follows by using (61).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ SPLITTING SCHEMES FOR PDMPS 67 Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ B, [L∗ R, L∗ B]]µ(x, v) = −λrµ(x, v)⟨v, ∇ψ(x)⟩(λ1(x, R(x)v) + λ1(x, v)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider now [L∗ B, [L∗ R, L∗ B]]µ(x, v) = L∗ B(λr⟨v, ∇ψ(x)⟩µ(x, v)) + [L∗ R, L∗ B](⟨v, ∇ψ(x)⟩µ(x, v)) = λrµ(x, v) � ⟨R(x)v, ∇ψ(x)⟩λ1(x, R(x)v) − ⟨v, ∇ψ(x)⟩λ1(x, v) + � � ⟨R(x)y, ∇ψ(x)⟩λ1(x, R(x)y⟩ − ⟨y, ∇ψ(x)⟩λ1(x, y) � ν(dy) + � (⟨y, ∇ψ(x)⟩ν(dy)⟨v, ∇ψ(x)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The last term equals zero as ν has mean zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then observe that by Identity (58) � � ⟨R(x)y, ∇ψ(x)⟩λ1(x, R(x)y) − ⟨y, ∇ψ(x)⟩λ1(x, y) � ν(dy) = = − � ⟨y, ∇ψ(x)⟩(λ1(x, R(x)y) + λ1(x, y))ν(dy) = − � � λ1(x, R(x)y)2ν(dy) − � λ1(x, y)2ν(dy) � = 0, (62) where the last equality follows by invariance under rotation of ν as required in Assumption 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence we have obtained the statement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ B, [L∗ B, L∗ D]]µ(x, v) = 2µ(x, v)λ1(x, R(x)v) � ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇ψ2(x)R(x)v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 we find [L∗ B, [L∗ B, L∗ D]]µ(x, v) = L∗ B � µ(x, v)(λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩) � (*) + [L∗ B, L∗ D] � ⟨v, ∇ψ(x)⟩µ(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (**) Let us treat the two terms separately, starting with (*).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' After applying L∗ B and using that R(x)(R(x)v) = v the first term becomes (*) = µ(x, v) �� (λ2 1(x, v) − λ2 1(x, R(x)v) + ⟨R(x)v, ∇ψ(x)⟩2 − ⟨R(x)v, ∇2ψ(x)R(x)v⟩ � λ1(x, R(x)v) − � λ2 1(x, R(x)v) − λ2 1(x, v) + ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩) � λ1(x, v) � = µ(x, v) � (λ2 1(x, v) − λ2 1(x, R(x)v))(λ1(x, R(x)v) + λ1(x, v)) + ⟨v, ∇ψ(x)⟩2(λ1(x, R(x)v) − λ1(x, v)) − ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩λ1(x, v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Using Identity (58) we obtain that ⟨v, ∇ψ(x)⟩2(λ1(x, R(x)v) − λ1(x, v)) = (λ2 1(x, v) − λ2 1(x, R(x)v))(λ1(x, R(x)v) + λ1(x, v)) 68 SPLITTING SCHEMES FOR PDMPS and thus cancelling out the corresponding terms in (*) it follows that (*) = µ(x, v) � ⟨v, ∇2ψ(x)v⟩λ1(x, v) − ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Focusing now on (**), we apply Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 to find (**) = −⟨R(x)v, ∇x � ⟨v, ∇ψ(x)⟩µ(x, v) � (x, R(x)v)⟩λ1(x, R(x)v) + ⟨v, ∇x � ⟨v, ∇ψ(x)⟩µ(x, v) � ⟩λ1(x, v) + ⟨v, ∇x � ⟨R(x)v, ∇ψ(x)⟩µ(x, v)λ1(x, R(x)v) − ⟨v, ∇ψ(x)⟩µ(x, v)λ1(x, v) � ⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Recalling that ∇x(⟨v, ∇ψ(x)⟩µ(x, v)) = µ(x, v)(∇2ψ(x)v − ∇ψ(x)⟨v, ∇ψ(x)⟩), we find (**) = µ(x, v) � � −⟨R(x)v, ∇2ψ(x)R(x)v⟩ + ⟨v, ∇ψ(x)⟩2� λ1(x, R(x)v) + � ⟨v, ∇2ψ(x)v⟩ − ⟨v, ∇ψ(x)⟩2� λ1(x, v) � − ⟨v, ∇x � ⟨v, ∇ψ(x)⟩µ(x, v)|⟨v, ∇ψ(x)⟩| � ⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' In particular we used Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 to write the last term more compactly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The derivative in the last term can be computed as follows − ⟨v, ∇x � ⟨v, ∇ψ(x)⟩µ(x, v)|⟨v, ∇ψ(x)⟩| � ⟩ = = −µ(x, v)⟨v, ∇2ψ(x)v|⟨v, ∇ψ(x)⟩| − ∇ψ(x)⟨v, ∇ψ(x)⟩|⟨v, ∇ψ(x)⟩| + ⟨v, ∇ψ(x)⟩sign(⟨v, ∇ψ(x)⟩)∇2ψ(x)v⟩ = −µ(x, v) � − ⟨v, ∇ψ(x)⟩2|⟨v, ∇ψ(x)⟩| + ⟨v, ∇2ψ(x)v⟩ (|⟨v, ∇ψ(x)⟩| + ⟨v, ∇ψ(x)⟩sign(⟨v, ∇ψ(x)⟩)) � = −µ(x, v) � − ⟨v, ∇ψ(x)⟩2|⟨v, ∇ψ(x)⟩| + ⟨v, ∇2ψ(x)v⟩2|⟨v, ∇ψ(x)⟩| � = µ(x, v)|⟨v, ∇ψ(x)⟩| � ⟨v, ∇ψ(x)⟩2 − 2⟨v, ∇2ψ(x)v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Hence re-applying Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 we find (**) = µ(x, v) � − ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v) + 2⟨v, ∇ψ(x)⟩2λ1(x, R(x)v) − (2λ1(x, R(x)v) + λ1(x, v))⟨v, ∇2ψ(x)v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The proof is now concluded by summing (*) and (**).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ D, [L∗ R, L∗ B]]µ(x, v) = λrµ(x, v) � ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider now [L∗ D, [L∗ R, L∗ B]]: [L∗ D, [L∗ R, L∗ B]]µ(x, v) = L∗ D(λr⟨v, ∇ψ(x)⟩µ(x, v)) − [L∗ R, L∗ B](⟨v, ∇ψ(x)⟩µ(x, v)) = −⟨v, ∇x(λr⟨v, ∇ψ(x)⟩µ(x, v))⟩ − λr � µ(x, v) � (−⟨y, ∇ψ(x)⟩)(λ1(x, R(x)y) + λ1(x, y))ν(dy) � = λrµ(x, v) � ⟨v, ∇ψ(x)⟩2 − ⟨v, ∇2ψ(x)v⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' SPLITTING SCHEMES FOR PDMPS 69 In particular we used that � (⟨y, ∇ψ(x)⟩)(λ1(x, R(x)y) + λ1(x, y))ν(dy) = � λ1(x, y)2ν(dy) − � λ1(x, R(x)y)2ν(dy) = 0 which was shown in (62).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ D, [L∗ R, L∗ D]]µ(x, v) = λrµ(x, v) � ⟨v, ∇2ψ(x)v⟩ − ⟨v, ∇ψ(x)⟩2 + b tr � ∇2ψ(x) − ∇ψ(x)∇ψ(x)T � � Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='4 [L∗ D, [L∗ R, L∗ D]]µ(x, v) = −λrL∗ D(⟨v, ∇ψ(x)⟩µ(x, v)) − [L∗ R, L∗ D](⟨v, ∇ψ(x)⟩µ(x, v)) = λrµ(x, v) � ⟨v, ∇2ψ(x)v⟩ − ⟨v, ∇ψ(x)⟩2 + � (⟨y, ∇2ψ(x)y⟩ − ⟨y, ∇ψ(x)⟩2)ν(dy) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The statement follows by Equation 61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' It holds that [L∗ D, [L∗ B, L∗ D]] = µ(x, v) � − 4⟨v, ∇ψ(x)⟩2λ1(x, R(x)v) + 7⟨v, ∇2ψ(x)v⟩λ1(x, R(x)v) + ⟨v, ∇x(⟨v, ∇2ψ(x))v⟩)⟩ + ⟨R(x)v, ∇2ψ(x)R(x)v⟩λ1(x, R(x)v) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' By Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2 together with Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1 [L∗ D, [L∗ B, L∗ D]]µ(x, v) = L∗ D � µ(x, v) � ⟨v, ∇ψ(x)⟩ � ⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩| � − ⟨v, ∇2ψ(x)v⟩ �� (†) − [L∗ B, L∗ D](⟨v, ∇ψ(x)⟩µ(x, v)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' (††) = (†) − (††).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider the two terms separately, starting from the first one: (†) = µ(x, v)⟨v, ∇ψ(x)⟩ � ⟨v, ∇ψ(x)⟩ � ⟨v, ∇ψ(x)⟩ − |⟨v, ∇ψ(x)⟩| � − ⟨v, ∇2ψ(x))v⟩ � − µ(x, v)⟨v, 2⟨v, ∇ψ(x)⟩∇2ψ(x)v − 2∇2ψ(x)v|⟨v, ∇ψ(x)⟩| − ∇x(⟨v, ∇2ψ(x))v⟩)⟩ = µ(x, v) � − 2⟨v, ∇ψ(x)⟩2λ1(x, R(x)v) + ⟨v, ∇2ψ(x)⟩(−3⟨v, ∇ψ(x)⟩ + 2|⟨v, ∇ψ(x)⟩|) + ⟨v, ∇x(⟨v, ∇2ψ(x))v⟩)⟩ � The second term (††) is the same as term (**) in the proof of Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' The statement follows taking the difference of the two terms (†) and (††) and using Lemma E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' □ E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Proof of Proposition 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider symmetric splitting schemes of the form eδLS = e δ 2 LAe δ 2 LBeδLCe δ 2 LBe δ 2 LA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' We have by the Baker-Campbell-Haussdorff formula L∗ S = L∗ + δ2 12 � [L∗ C, [L∗ C, L∗ A + L∗ B]] + [L∗ B, [L∗ B, L∗ A]] + [L∗ C, [L∗ B, L∗ A]] + [L∗ B, [L∗ C, L∗ A]] − 1 2[L∗ B, [L∗ B, L∗ C]] − 1 2[L∗ A, [L∗ A, L∗ C]] − 1 2[L∗ A, [L∗ A, L∗ B]] � + O(δ4) 70 SPLITTING SCHEMES FOR PDMPS = L∗ + δ2L∗ 2 + O(δ4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' where L∗ = L∗ A + L∗ B + L∗ C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Therefore it is sufficient to use the commutators of Section E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Observe that L∗ BPSµ(x, v) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Let us start with L∗ DBRBD: L∗ DBRBDµ(x, v) = δ2 12 � [L∗ R, [L∗ R, L∗ D + L∗ B]] + [L∗ B, [L∗ B, L∗ D]] + [L∗ R, [L∗ B, L∗ D]] + [L∗ B, [L∗ R, L∗ D]] − 1 2[L∗ B, [L∗ B, L∗ R]] − 1 2[L∗ D, [L∗ D, L∗ R]] − 1 2[L∗ D, [L∗ D, L∗ B]] � + O(δ4) = δ2 12µ(x, v) � 3 2λr � b tr � ∇ψ(x)∇ψ(x)T − ∇2ψ(x) � + 2⟨v, ∇ψ(x)⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩ � + 3 2λ1(x, R(x)v) � ⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩ � + 1 2⟨v, ∇x(⟨v, ∇2ψ(x)v⟩)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Then focus on L∗ BDRDB: L∗ BDRDBµ(x, v) = δ2 12 � [L∗ R, [L∗ R, L∗ D + L∗ B]] + [L∗ D, [L∗ D, L∗ B]] + [L∗ R, [L∗ D, L∗ B]] + [L∗ D, [L∗ R, L∗ B]] − 1 2[L∗ D, [L∗ D, L∗ R]] − 1 2[L∗ B, [L∗ B, L∗ R]] − 1 2[L∗ B, [L∗ B, L∗ D]] � + O(δ4) = δ2 12µ(x, v) � − 3 2λr � b tr � ∇ψ(x)∇ψ(x)T − ∇2ψ(x) � + 2⟨v, ∇ψ(x)⟩λ1(x, R(x)v) + ⟨v, ∇2ψ(x)v⟩ � + 3λ1(x, R(x)v) � − 2⟨v, ∇2ψ(x)v⟩ + ⟨v, ∇ψ(x)⟩2� − ⟨v, ∇(⟨v, ∇2ψ(x)v⟩)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Consider L∗ RDBDR: L∗ RDBDRµ(x, v) = δ2 12 � [L∗ B, [L∗ B, L∗ R + L∗ D]] + [L∗ D, [L∗ D, L∗ R]] + [L∗ B, [L∗ D, L∗ R]] + [L∗ D, [L∗ B, L∗ R]] − 1 2[L∗ D, [L∗ D, L∗ B]] − 1 2[L∗ R, [L∗ R, L∗ B]] − 1 2[L∗ R, [L∗ R, L∗ D]] � + O(δ4) = δ2 12µ(x, v) � 3 2λ1(x, R(x)v) � ⟨v, ∇2ψ(x)v⟩ − ⟨R(x)v, ∇2ψ(x)R(x)v⟩ � + 1 2⟨v, ∇x(⟨v, ∇2ψ(x)v⟩)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' Finally focus on L∗ DRBRD: L∗ DRBRDµ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) = δ2 12 � [L∗ B,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ B,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ D + L∗ R]] + [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ D]] + [L∗ B,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ D]] + [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ B,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ D]] − 1 2[L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ R,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ B]] − 1 2[L∗ D,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ D,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ B]] − 1 2[L∗ D,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' [L∗ D,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' L∗ R]] � + O(δ4) = δ2 12µ(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � 3 2λr � b tr � ∇ψ(x)∇ψ(x)T − ∇2ψ(x) � + ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩ � 3λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) + λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' v) � + ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇2ψ(x)v⟩ � + 3 2λ1(x,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' R(x)v) � ⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇2ψ(x)v⟩ − ⟨R(x)v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇2ψ(x)R(x)v⟩ � + 1 2⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇(⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇2ψ(x)v⟩)⟩ + 3 2λ2 r⟨v,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
+page_content=' ∇ψ(x)⟩ � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/hNE0T4oBgHgl3EQfpQGK/content/2301.02537v1.pdf'}
diff --git a/i9E2T4oBgHgl3EQfdAdf/content/2301.03901v1.pdf b/i9E2T4oBgHgl3EQfdAdf/content/2301.03901v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..775831140a7f4ad232a06e1ec1d96f6f43569914
--- /dev/null
+++ b/i9E2T4oBgHgl3EQfdAdf/content/2301.03901v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4981dc70b1ee779afb094a9b8a145e85647b85ef4c43a3c39f68597311e59e4
+size 803303
diff --git a/i9E2T4oBgHgl3EQfdAdf/vector_store/index.faiss b/i9E2T4oBgHgl3EQfdAdf/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..494b2f53a926832e356814eb2c37ccecb973c57a
--- /dev/null
+++ b/i9E2T4oBgHgl3EQfdAdf/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d28f8b2b04a47c35b41bd4b3796f24faa7d3f83d6e2a9639c5db554e5a7f5cac
+size 4194349
diff --git a/i9E2T4oBgHgl3EQfdAdf/vector_store/index.pkl b/i9E2T4oBgHgl3EQfdAdf/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..2e76affb699857998026aab860e069c727404021
--- /dev/null
+++ b/i9E2T4oBgHgl3EQfdAdf/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:633d47f321222c12e72895ba10a9801c4d8f4b1caa704f16bdd02cf424dd215f
+size 146433
diff --git a/idE4T4oBgHgl3EQfSgzP/vector_store/index.pkl b/idE4T4oBgHgl3EQfSgzP/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..b0db93334c6ce4b7fbacc6dcaa8546c773ed64bc
--- /dev/null
+++ b/idE4T4oBgHgl3EQfSgzP/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:891cb3f70ac365bb297ecb023ddaa29bdd4a8428e6948d4a3e79d81a56783a4f
+size 730075
diff --git a/j9E4T4oBgHgl3EQfTQyV/vector_store/index.faiss b/j9E4T4oBgHgl3EQfTQyV/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..7ef5f98fe1c330406656cbbdd92ca1030ff714fd
--- /dev/null
+++ b/j9E4T4oBgHgl3EQfTQyV/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e70b97a299267239106d77606f5fb678f2986d85171f757ea418f06b2282d20
+size 10223661
diff --git a/jNA0T4oBgHgl3EQfIv8h/vector_store/index.pkl b/jNA0T4oBgHgl3EQfIv8h/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..c6e77f7301ff5c69cb90502a13d696b91dd1dc6f
--- /dev/null
+++ b/jNA0T4oBgHgl3EQfIv8h/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58c1d87720f211a6bac487a49942dcb7c434c137e2858160f4851b6d3e99b5f5
+size 178434
diff --git a/jNAyT4oBgHgl3EQfkfir/content/tmp_files/2301.00435v1.pdf.txt b/jNAyT4oBgHgl3EQfkfir/content/tmp_files/2301.00435v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4c3546a23b3b54a89e41a6f24f5df0f5c7a330cd
--- /dev/null
+++ b/jNAyT4oBgHgl3EQfkfir/content/tmp_files/2301.00435v1.pdf.txt
@@ -0,0 +1,1851 @@
+Trojaning semi-supervised learning model via poisoning wild
+images on the web
+Le Feng
+Fudan University
+China
+Zhengxing Qian
+Fudan University
+China
+Sheng Li
+Fudan University
+China
+Xinpeng Zhang
+Fudan University
+China
+ABSTRACT
+Wild images on the web are vulnerable to backdoor (also called
+trojan) poisoning, causing machine learning models learned on
+these images to be injected with backdoors. Most previous attacks
+assumed that the wild images are labeled. In reality, however, most
+images on the web are unlabeled. Specifically, we study the effects
+of unlabeled backdoor images under semi-supervised learning (SSL)
+on widely studied deep neural networks. To be realistic, we assume
+that the adversary is zero-knowledge and that the semi-supervised
+learning model is trained from scratch. Firstly, we find the fact that
+backdoor poisoning always fails when poisoned unlabeled images
+come from different classes, which is different from poisoning the
+labeled images. The reason is that the SSL algorithms always strive
+to correct them during training. Therefore, for unlabeled images,
+we implement backdoor poisoning on images from the target class.
+Then, we propose a gradient matching strategy to craft poisoned
+images such that their gradients match the gradients of target im-
+ages on the SSL model, which can fit poisoned images to the target
+class and realize backdoor injection. To the best of our knowledge,
+this may be the first approach to backdoor poisoning on unlabeled
+images of trained-from-scratch SSL models. Experiments show that
+our poisoning achieves state-of-the-art attack success rates on most
+SSL algorithms while bypassing modern backdoor defenses.
+KEYWORDS
+Backdoor, Semi-supervised Learning, Trained-from-scratch, Neural
+Network
+1
+INTRODUCTION
+The excellent performance of deep neural networks [12, 31, 37, 49] is
+largely due to numerous training examples. To obtain enough train-
+ing examples, trainers usually grab them from the web. However,
+these examples from the wild may not be safe, they are vulnerable
+to backdoor poisoning. Previous backdoor poisonings [10, 21, 28,
+39, 45] mainly focus on the labeled examples, which rely on the
+guidance of the target label to inject backdoors into the models. Yan
+et al. [45, 46] initially propose two unlabeled backdoor poisoning
+schemes for pre-trained SSL models: DeNeB [45] and DeHiB [46].
+They assume that the SSL learner first trains the model on labeled
+examples. The obtained pre-trained model is then fine-tuned using
+the SSL algorithm in combination with unlabeled examples. Actu-
+ally, most advanced SSL algorithms are end-to-end, i.e., unlabeled
+examples along with labeled examples are fed into the model to
+train from scratch. There is no intermediate pre-trained model. Be-
+sides, the backdoor patterns proposed by Yan et al. are perceptible,
+which can be detected by DePuD [46].
+Adversary
+Clean Ship
+Imperceptible backdoor patterns
+Backdoor Ship
+Labeled part
+SSL training set
+⋯
+⋯
+Bird
+Car
+Plane
+Ship
+Unlabeled part
+Victim
+network Trained-from-
+scratch SSL
++
+=
+Poisoning
+Victim
+Ship
+Ship
+Ship
+Ship
+Feeding
+Predicting
+Unseen backdoor images
+Backdoor
+network
+Adversary
+Releasing
+Figure 1: Attack pipelines of unlabeled backdoor poisoning
+on the trained-from-scratch SSL model. Assuming that the
+target class is Ship.
+To be practical and stealthy, we propose a zero-knowledge and
+imperceptible backdoor poisoning on unlabeled examples of trained-
+from-scratch SSL models. Zero-knowledge means that the adversary
+does not require knowledge of the victim model, the SSL algorithm,
+the complete dataset, and the training process. Our attack pipeline
+is shown in Fig. 1. The adversary adds the imperceptible backdoor
+patterns to the clean Ship images. Then use the resulting backdoor
+images to poison the unlabeled part of the SSL training set. The
+victim uses the poisoned training set to train the network from
+scratch by the SSL algorithm, thus causing the backdoor to be
+injected inadvertently. Finally, in the inference stage, the images
+with backdoor patterns will be misclassified as the target class Ship.
+Specifically, first of all, we find that unlike poisoning labeled
+examples, since SSL algorithms strive to learn correctly the unla-
+beled examples, if the poisoned examples are from different classes,
+i.e., label-inconsistent backdoor poisoning, they will be re-learned
+into the correct class by the trained-from-scratch SSL model. As
+a result, the backdoor cannot be injected. On the contrary, if poi-
+soning only is implemented on the examples from the target class,
+arXiv:2301.00435v1 [cs.CY] 1 Jan 2023
+
+Conference’17, July 2017, Washington, DC, USA
+Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang
+i.e., label-consistent backdoor poisoning, they will end up being
+classified into the target class by the SSL model. And since various
+regularizations of SSL algorithms mitigate overfitting on the target
+class, backdoor patterns can be generalized to non-target classes.
+Thus, our poisoning is only implemented on the examples from the
+target class. Then, to achieve zero-knowledge poisoning, we resort
+to the transferability of neural networks. Only the target class and
+the distribution of the victim dataset are required. Specifically, We
+first prepare a surrogate dataset with a similar distribution to the
+victim dataset containing the examples from the target class. The
+surrogate network is then trained on the surrogate dataset. For the
+obtained surrogate network, we train a backdoor pattern generator
+that takes clean examples from different classes as input and outputs
+the corresponding imperceptible backdoor patterns, and then adds
+the backdoor patterns to the examples to get poisoned examples.
+Besides the imperceptibility of backdoor patterns, the other train-
+ing target of the generator is to achieve gradient matching which
+is proposed to make the gradients of poisoned examples match
+the gradients of the target examples on the surrogate model. We
+hope that in the trained-from-scratch SSL model, through gradient
+matching, the poisoned examples can be naturally learned into the
+target class as the target examples are learned into the target class,
+thereby injecting the backdoor. In summary, our contributions are
+as follows:
+1. To the best of our knowledge, we are the first to investigate
+the vulnerability of unlabeled examples of trained-from-scratch
+SSL models to backdoor poisoning.
+2. We find that for unlabeled examples of trained-from-scratch
+SSL models, label-consistent backdoor poisoning is more effective.
+3. We propose a zero-knowledge and imperceptible backdoor
+poisoning on unlabeled examples of trained-from-scratch SSL mod-
+els.
+4. We implement our poisoning on SSL algorithms of three types.
+Attack success rates are significantly higher than baseline poison-
+ings, and ours can successfully bypass various defenses including
+DePuD.
+2
+RELATED WORK
+2.1
+Backdoor poisoning
+If the poisoned examples are all from the target class, it is called
+label-consistent backdoor poisoning, otherwise, it is label-inconsistent
+backdoor poisoning.
+Label-inconsistent backdoor poisoning: BadNets [10] is the
+first backdoor poisoning for neural networks. The scheme is rudi-
+mentary so that numerous backdoor defenses [3, 8, 11, 22, 42] can de-
+tect or remove the backdoor. Many follow-up works propose more
+threatening poisonings. Invisible backdoor patterns [21, 23, 28]
+make it difficult for victims to visually detect the abnormality of
+the poisoned examples. Dynamic backdoor patterns [27, 33] can
+make it difficult for victims to capture the regular pattern of back-
+door patterns. There are also schemes [6, 21] that achieve invisible
+and dynamic backdoor patterns. DeHiB [45] and DeNeB [46] that
+poison unlabeled examples of pre-trained models are also label-
+inconsistent.
+Label-consistent backdoor poisoning: Due to not changing
+the correct labels of poisoned examples, it is more stealthy. How-
+ever, backdoor patterns may overfit on the target class and fail
+to generalize to non-target classes. CLB (Clean Label backdoor)
+[39] first proposes to mitigate this overfitting by adversarial per-
+turbation and interpolation. Later, [48] implements this backdoor
+poisoning on the videos. [20] implements this backdoor poisoning
+on the point clouds.
+2.2
+Backdoor defense
+For different phases of backdoor poisoning, backdoor defenses can
+be categorized into four types: pre-training defense, post-training
+defense, testing-time defense, and blind defense.
+Pre-training defense: The defender checks training examples
+to determine whether there are suspicious examples. For label-
+inconsistent poisoning in supervised learning, poisoned examples
+can be screened by the inconsistency between the content of the
+examples and their labels. Formally, activation clustering [3] can
+detect outlier examples by clustering training examples according
+to their labels. Recently, DePuD [46] is proposed to detect unlabeled
+poisoned examples in semi-supervised learning, which uses heavy
+regularization to distinguish suspicious unlabeled examples.
+Post-training defense: This defense is to detect anomalies in
+the learned model. A typical detection is Neural Cleanse [42]. Re-
+verse engineering is first used for all classes to get their triggers. If
+the trigger intensity of the class is abnormally smaller than those
+of other classes, this class is detected as a backdoor class. Later,
+many variants based on Neural Cleanse appear. For example, [4, 43]
+improve Neural Cleanse with better objective functions. [7, 11]
+propose the detection in black box scenarios.
+Testing-time defense: The defense is deployed during the model
+testing phase. The testing example is checked. A typical detection
+is STRIP [8]. The testing example is fused with a set of pre-prepared
+clean examples to obtain synthetic examples. Then, feed these syn-
+thetic examples to the model for prediction. If prediction results
+present a low-entropy distribution, then the testing example may
+be a backdoor example.
+Blind defense: Instead of detecting examples or models, unified
+operations against the examples or model are adopted. Data aug-
+mentation is a natural blind defense method. In the testing phase,
+processing such as JPEG compression on the examples may also
+destroy backdoor patterns. Fine-pruning [22] prunes and fine-tunes
+the model to try to destroy possible backdoors in the model.
+2.3
+Semi-Supervised Learning
+Existing SSL algorithms can be categorized into three types: consis-
+tency regularization [17, 25, 30, 38, 41, 44], pseudo-labeling [15, 19,
+29], and pseudo-labeling with consistency regularization [1, 2, 35].
+Consistency regularization: It assumes that randomness within
+the neural network or data augmentation transformations should
+not modify model predictions given the same input. For example,
+PI-Model [30] minimizes the difference between two passes through
+the network with stochastic transformations for the same point.
+MeanTeacher [38] minimizes the difference between the predic-
+tions of the student model and the teacher model for the same point.
+
+Trojaning semi-supervised learning model via poisoning wild images on the web
+Conference’17, July 2017, Washington, DC, USA
+Bird
+Car
+Plane
+Ship
+Ship
+Ship
+Ship
+Ship
+Adding backdoor
+patterns
+Changing their labels
+to the target label
+Clean labeled images
+Poisoned labeled images
+(a) Label-inconsistent poisoning in SL
+Ship
+Ship
+Ship
+Ship
+Adding backdoor
+patterns
+Keeping their labels
+unchanged
+Ship
+Ship
+Ship
+Ship
+Clean labeled images
+Poisoned labeled images
+(b) Label-consistent poisoning in SL
+Adding backdoor
+patterns
+Poisoned unlabeled images
+Clean unlabeled images
+(c) Label-inconsistent poisoning in SSL
+Adding backdoor
+patterns
+Poisoned unlabeled images
+Clean unlabeled images
+(d) Label-consistent poisoning in SSL
+Backdoor network
+Adding backdoor
+patterns
+Ship
+Ship
+Ship
+Ship
+Feeding
+Predicting
+Unseen clean images
+Unseen backdoor images
+Predicted classes
+Backdoor
+network
+(e) Poisoning target
+Figure 2: Backdoor poisoning in SL and SSL. In these poisonings, the target class is "ship". Their poisoning targets are the
+same, which all cause unseen backdoor images to be misclassified as "ship" by the backdoor network trained on the poisoned
+training set.
+VAT [25], ICT [41], and UDA [44] aim to develop more efficient
+augmentations to exploit unlabeled data.
+Pseudo-labeling: It assigns pseudo labels to unlabeled exam-
+ples based on the predictions of the current model and then trains
+unlabeled examples by supervised learning. For example, pseudo
+labeling [19] uses the pretrained network trained on the labeled
+examples to predict pseudo labels. MPL (Meta Pseudo Labeling) [29]
+maintains two models: a student model and a teacher model. The
+teacher model predicts unlabeled examples to give pseudo labels.
+Pseudo-labeling with consistency regularization: MixMatch
+[2] uses MixUp augmentation to create multiple augmentations for
+each unlabeled example, and then takes the maximum class of the
+average of the predictions of these augmentations as the pseudo
+label. ReMixMatch [1] improves MixMatch by introducing two new
+mechanisms: distribution alignment and augmentation anchoring.
+FixMatch [35] performs weak augmentation and strong augmenta-
+tion for each unlabeled example, and the predicted label of weak
+augmentation is used as the pseudo label of strong augmentation.
+3
+OUR NOVEL FINDING
+In the context of supervised learning, as shown in Fig. 2(a) and Fig.
+2(b), both label-consistent and label-inconsistent backdoor poison-
+ings rely on the guidance of the target label. The difference is that
+label-inconsistent backdoor poisoning changes their labels to target
+labels. This is not required for label-consistent backdoor poisoning.
+However, this also leads to the fact that since backdoor patterns are
+not added to the non-target class examples, the model may overfit
+backdoor patterns on the target class, so that backdoor patterns do
+not work on non-target class. Although CLB [39] proposes interpo-
+lation and adversarial perturbation to improve the generalization
+of backdoor patterns on non-target classes, attack success rates
+are lower than label-inconsistent backdoor poisoning. Thus, label-
+inconsistent backdoor poisoning is easier to be implemented than
+label-consistent backdoor poisoning.
+However, in the context of semi-supervised learning, on the one
+hand, backdoor poisoning on unlabeled examples will lose the guid-
+ance of the target label, as shown in Fig. 2(c) and 2(d). On the other
+hand, the difference in the mechanism of semi-supervised learning
+and supervised learning brings a novel finding:
+Since the semi-supervised learning algorithms strive to correctly learn
+unlabeled examples through various regularizations, for unlabeled
+examples, label-inconsistent backdoor poisoning is much more diffi-
+cult to implement than label-consistent backdoor poisoning.
+Next, we will experimentally verify our finding. We use exist-
+ing schemes BadNets [10], CLB [39], and DeNeB [46] to poison
+unlabeled examples of trained-from-scratch SSL models. Note that
+although some recent backdoor poisonings, e.g., invisible backdoor
+poisonings [21, 23, 28], are better at resisting backdoor defenses,
+BadNets is still excellent in terms of attack success rate. Likewise,
+in the context of a pretrained network, the attack success rate of
+DeNeB is much higher than that of DeHiB.
+Figure 3: Poisoning unlabeled examples of trained-from-
+scratch SSL model using existing schemes. The tested vic-
+tim dataset is CIFAR10 [16] and the network is CNN13 [38].
+In the three backdoor poisoning schemes, the backdoor pat-
+terns are all 8 × 8 pixel squares. The target class is 8.
+The attack success rates are shown in Fig. 3. Both BadNets and
+DeNeB fail to poison completely, and the attack success rates are
+close to the probability 10.00% of random classification. In contrast,
+CLB obtains certain attack success rates (56.88%, 42.03%, 23.45%).
+Specifically, as shown in Fig. 2(c), for label-inconsistent backdoor
+
+Attack success rate of existing schemes
+100
+MeanTeacher[41]
+80
+Pseudolabeling[2o]
+Attack success rate
+FixMatch [38]
+60
+56.88
+42.03
+40
+30.1130.36
+30.12
+23.45
+22.27
+25.01
+20
+18.9
+8.64 9.65 9.84
+8.45 8.49 9.26
+0
+BadNets
+CLB
+DeNeB
+BadNets-C
+DeNeB-CConference’17, July 2017, Washington, DC, USA
+Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang
+poisoning, i.e., BadNets and DeNeB, pseudo-labeling based SSL
+algorithms [1, 2, 15, 19, 29, 35] strive to assign correct labels to un-
+labeled examples, while the poisoned unlabeled examples coming
+from different classes expect themselves to be misclassified into the
+target class. This opposition makes backdoor patterns difficult to
+be learned. Likewise, when consistency regularization based SSL
+algorithms [1, 2, 17, 25, 30, 35, 38, 41, 44] are employed, the noises
+or augmentations the SSL algorithms add to the examples or models
+will make the models to unlearn backdoor patterns but to focus
+on the semantic information of the poisoned unlabeled examples.
+As shown in Fig. 4, as SSL proceeds, the poisoned unlabeled exam-
+ples are gradually classified into their respective correct classes.
+However, for label-consistent backdoor poisoning, i.e., CLB, since
+poisoning only is implemented on examples from the target class
+(Fig. 2(d)), SSL algorithms classify all of them into the target class.
+Such opposition does not exist. Moreover, various regularizations
+of SSL algorithms prevent the model from overfitting on the target
+class, so backdoor patterns can be slightly generalized to non-target
+classes.
+To further verify our finding, we generalize DeNeB and BadNets
+to label-consistent versions DeNeB-C and BadNets-C, where C
+indicates consistent. With all settings unchanged, as shown in Fig.
+3, the attack success rates have been significantly improved, e.g.,
+for DeNeB, the increase from 8.45%, 8.49%, 9.26% to 30.12%,
+25.01%, 18.9%. However, CLB, DeNeB-C, and BadNets-C have
+three significant shortcomings.
+(1) The attack success rate is not ideal, the highest is only 56.88%.
+(2) The backdoor patterns are perceptible and easily detected by
+the victim as suspicious, as shown in Fig. 6(b), 6(c), and 6(d).
+(3) DePuD [46], a detection solution for poisoned unlabeled
+examples, can detect the anomaly.
+To remedy these shortcomings, we propose a zero-knowledge
+and imperceptible backdoor poisoning on unlabeled examples of
+trained-from-scratch SSL models.
+(a)
+(b)
+(c)
+Figure 4: t-SNE [40] feature distribution of poisoned un-
+labeled examples in label-inconsistent backdoor poisoning
+DeNeB in the trained-from-sratch SSL. The SSL algorithm is
+FixMatch.
+4
+OUR METHOD
+4.1
+Threat model
+Assume that the victim who trains a neural network model has
+only limited labeled examples. To improve model performance, he
+intends to scrape more unlabeled examples from the web for semi-
+supervised learning. For example, a state-of-the-art image classifier
+[24] scrapes 1 billion images from Instagram. At this point, an
+adversary who can upload data to the network can control a portion
+of the unlabeled examples, thereby realizing backdoor poisoning.
+Since our attack is zero-knowledge, an adversary has very limited
+knowledge. Specifically, what an adversary cannot obtain are:
+(1) The architecture, weights, and outputs of the trained-from-
+scratch victim model.
+(2) The training process, hyperparameter settings, and the SSL
+algorithm employed.
+(3) The complete victim dataset and whether the examples are
+labeled.
+The only knowledge an adversary can obtain is:
+(1) The distribution Z of the victim dataset and the target class
+𝑦𝑡 of poisoning.
+Formally, for a victim model F parameterized by 𝜃, the train-
+ing set of semi-supervised learning consists of a labeled part X =
+{(𝑥𝑛,𝑦𝑛) : 𝑛 ∈ (1, ..., 𝑁)} and an unlabeled part U = {𝑢𝑛 : 𝑛 ∈
+(1, ..., 𝜇𝑠𝑁)}, where 𝜇𝑠 is a hyperparameter that determines the
+relative sizes of X and U. Let F (𝑥;𝜃) be the predicted class distri-
+bution produced by the model F for input 𝑥. For convenience, we
+always use Γ to represent the SSL algorithm, and the SSL process
+can be formalized as:
+arg min
+𝜃
+Γ(F (X ∪ U;𝜃)).
+(1)
+The threat model can be formalized as the bilevel problem listed
+in Eq. 2. The outer optimization is to achieve two targets, one is the
+fundamental target of adversary backdoor poisoning: to maximize
+the attack success rate of backdoor examples with backdoor patterns
+without degrading model accuracy on unseen examples X𝑣𝑎𝑙, and
+the other is to ensure that backdoor patterns are the least perceptible
+to avoid arousing the suspicion of the victim. The inner optimization
+is that the victim uses the SSL algorithm Γ to train the model on
+the poisoned training set.
+min
+P𝑡𝑟 (·) E(𝑥,𝑦) ∈X𝑣𝑎𝑙 (ℓ(𝑦, F (𝑥;𝜃∗)) + ℓ(𝑦𝑡, F (P𝑣𝑎𝑙 (𝑥);𝜃∗))
++∥P𝑣𝑎𝑙 (𝑥) − 𝑥∥2)
+s.t.
+𝜃∗ = arg min
+𝜃
+Γ(F (X ∪ (1 − 𝜇𝑏)U ∪ P𝑡𝑟 (𝜇𝑏U);𝜃))
+,
+(2)
+where P𝑡𝑟 (𝜇𝑏U) indicates that the unlabeled data 𝜇𝑏U in the train-
+ing set is poisoned, and 𝜇𝑏 is a hyperparameter that determines the
+proportion of backdoor poisoning. P𝑣𝑎𝑙 (𝑥) is to add the backdoor
+pattern to the example 𝑥 to get the backdoor example.
+4.2
+Achieving P𝑡𝑟 and P𝑣𝑎𝑙
+The first thing to note is that P𝑣𝑎𝑙 (𝑥) and P𝑡𝑟 (𝑥) in Eq. 2 are differ-
+ent in CLB and DeNeB. In CLB, for the model to remember backdoor
+patterns well, interpolation or adversarial perturbation is used to
+keep the selected images away from their correct classification
+when adding backdoor patterns. In DeNeB, when adding backdoor
+patterns, it also makes the features and classifications of selected
+images close to the target class through adversarial perturbation. In
+contrast, P𝑣𝑎𝑙 (𝑥) and P𝑡𝑟 (𝑥) in our poisoning are the same, that is,
+poisoned unlabeled images are obtained by only adding backdoor
+patterns to clean unlabeled images. Our poisoning focus on how
+to craft backdoor patterns so that the SSL models trained on poi-
+soned images can remember them. In general, our poisoning can
+be divided into the following three steps: preparing the surrogate
+network and dataset, crafting backdoor patterns, and poisoning
+unlabeled images.
+
+Feature distribution of poisoned
+unlabeledexamplesat4oothepoch
+1.0
+0.8
+0.6
+1
+2
+0.4-
+3
+4
+5
+0.2 -
+6
+8
+9
+0.0
+0.0
+0.2
+0.4
+0.6
+0.8
+1.0Feature distribution of poisoned
+unlabeledexamplesatothepoch
+1.0
+0
+1
+2
+0.8
+3
+4
+5
+0.6
+6
+7
+8
+9
+0.4
+0.2
+0.0
+0.0
+0.2
+0.4
+0.6
+0.8
+1.0Feature distribution of poisoned
+unlabeled examples at 2ooth epoch
+1.0
+0
+2
+0.8
+3
+5
+0.6
+8
+9
+0.4
+0.2
+0.0
+0.0
+0.2
+0.4
+0.6
+0.8
+1.0Trojaning semi-supervised learning model via poisoning wild images on the web
+Conference’17, July 2017, Washington, DC, USA
+⋮
+⋮
+ℒ𝑏
+ℒ𝑡
+∇𝜃𝑠ℒ𝑏
+∇𝜃𝑠ℒ𝑡
+ℒ𝑏𝑡
+ℒ𝑖𝑛𝑣
+Clean images
+Backdoor patterns
+Backdoor images
+Backward
+Backward
+Target class 𝑦𝑡
+Target image 𝑥𝑡
+Frozen
+surrogate
+network
+Backdoor pattern
+generator
+Clip
+𝜀Tanh
+⋱
+⋱
+⋱
+: ship
+Figure 5: Crafting backdoor patterns on the surrogate network and the surrogate dataset.
+Preparing the surrogate network and dataset: Since only
+the target class 𝑦𝑡 and the distribution Z of the victim dataset are
+grasped by the adversary. With the help of the transferability of
+neural networks, the adversary crafts backdoor patterns on the sur-
+rogate dataset X𝑠 and the surrogate network F 𝑠 parameterized by
+𝜃𝑠. X𝑠 should contain the images for the target class𝑦𝑡 and conform
+to the distribution Z. F 𝑠 should ensure considerable classification
+accuracy on X𝑠, thus mining backdoor patterns that are as imper-
+ceptible as possible. F 𝑠 is then trained on X𝑠. For simplicity, in
+the following, the learned parameters of F 𝑠 are still denoted by 𝜃𝑠.
+Note that X𝑠 is not required to be labeled, which can be labeled-less
+or unlabeled. This is because F 𝑠 can be trained by semi-supervised
+learning [29, 35] or unsupervised learning [5, 9], which is beyond
+our research scope.
+Crafting backdoor patterns: This step crafts backdoor pat-
+terns based on the surrogate dataset X𝑠 and learned surrogate
+network F 𝑠. As concluded in Section 3, to avoid the failure of back-
+door poisoning caused by the correct labeling of SSL algorithms,
+poisoning only is implemented on the images from the target class
+𝑦𝑡. However, as listed in Eq. 2, the target of backdoor poisoning
+requires that the images from different classes are all misclassified
+as the target class 𝑦𝑡 by the backdoor model after adding backdoor
+patterns. Thus, although only images from the target class are poi-
+soned, the images from other classes need to be taken into account
+when crafting the backdoor patterns. Specifically, this step includes
+the design of two aspects.
+One is to make backdoor patterns imperceptible. As shown in Fig.
+5, we use a backdoor pattern generator to generate the raw backdoor
+pattern. Let G parameterized by 𝜗 denote this generator. The input
+is a clean image 𝑥𝑠 ∈ X𝑠, and the output is a raw backdoor pattern
+G(𝑥𝑠;𝜗) corresponding to this image. It is then constrained to a
+reasonable range using the activation function Tanh and multiplied
+by the budget 𝜖 to make the generator search for imperceptible
+backdoor patterns within the given budget. Finally, the obtained
+backdoor pattern is added to the image, and the clip function is
+connected to make the backdoor image 𝑥𝑏 in the normal range. To
+further ensure that the backdoor pattern is imperceptible, we add a
+loss function L𝑖𝑛𝑠 listed in Eq. 3, which makes the backdoor image
+look more like the clean image.
+𝐿𝑖𝑛𝑠 = E𝑥𝑠∼X𝑠
+���𝑥𝑏 − 𝑥𝑠���
+2
+s.t.
+���𝑥𝑏 − 𝑥𝑠���∞ ≤ 𝜀
+(3)
+The other is to make the backdoor image 𝑥𝑏 be learned by
+the trained-from-scratch SSL model F into the target class 𝑦𝑡. To
+achieve this target, we propose a gradient matching strategy. First,
+let’s see the learning of the target image 𝑥𝑡 from the target class 𝑦𝑡
+by F . Regardless of whether 𝑥𝑡 is labeled, whether pseudo-labeling
+or consistency regularization is employed, the target of learning 𝑥𝑡
+is to make it classified into the target class, which can be formalized
+as:
+𝜃𝑘 = 𝜃𝑘−1 − 𝜂∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑡;𝜃𝑘−1))
+for 𝑘 ∈ [1,𝑚𝑎𝑥_𝑘]
+(4)
+where 𝜃𝑘 indicates the weights of the 𝑘th iteration, 𝜂 is the learning
+rate, 𝑚𝑎𝑥_𝑘 indicates the number of iterations. When crafting the
+backdoor image𝑥𝑏, as listed in Eq. 5, we hope that the SSL algorithm
+learns them just like fitting the target image 𝑥𝑡 to the target class𝑦𝑡,
+so that the SSL algorithm can be tricked into injecting the backdoor.
+This means that at each iteration, the gradients of the backdoor
+image 𝑥𝑏 on the model F should match the gradients of the target
+image 𝑥𝑡 on the model F .
+𝜃𝑘−1 − 𝜂∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑡;𝜃𝑘−1)) ≈ 𝜃𝑘−1 − 𝜂∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑏;𝜃𝑘−1))
+→ ∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑡;𝜃𝑘−1)) ≈ ∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑏;𝜃𝑘−1))
+for 𝑘 ∈ [1,𝑚𝑎𝑥_𝑘]
+,
+(5)
+However, since our poisoning is zero-knowledge, the gradient infor-
+mation during model F training cannot be obtained. To circumvent
+this problem, we think of mimicking such gradient information on
+the surrogate network F 𝑠. Furthermore, gradient information of
+all iterations is not required. Because on the one hand, the model F
+trained from scratch will generate numerous gradient information,
+and it is extremely costly and not practical to mimic all of this.
+On the other hand, gradient information in early training does not
+carry meaningful information. Thus, considering the computational
+cost, we only take the gradient information of the well-trained sur-
+rogate network. Experiments in Fig. 7 have demonstrated that our
+approach is effective.
+Specifically, first, from the target class, we select images that
+can be classified as the target class with high confidence by the
+surrogate network F 𝑠 as target images 𝑥𝑡, thereby ensuring that
+the gradients of the backdoor image 𝑥𝑏 can well match those of the
+images from the target class. Then feed the target image 𝑥𝑡 to the
+frozen well-trained F 𝑠 and get the loss:
+L𝑡 = ℓ(𝑦𝑡, F 𝑠 (𝑥𝑡;𝜃𝑠)),
+(6)
+and calculate the gradient ∇𝜃𝑠 L𝑡 to the parameters 𝜃𝑠. Likewise,
+as shown in Fig. 5, the backdoor images are also fed to the F 𝑠, and
+
+Conference’17, July 2017, Washington, DC, USA
+Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang
+Table 1: Poisoning performance. SL CA represents the model accuracy trained only on labeled examples by supervised learn-
+ing. SSL CA represents the model accuracy trained on the complete training set (including labeled examples and unlabeled
+examples) by semi-supervised learning. CA indicates the accuracy of the poisoned SSL model. In the column IMP, the data
+from top to bottom are PSNR, SSIM, and L-∞ norm, respectively.
+Dataset
+SSL algorithm
+SL CA
+SSL CA
+BadNets-C [10]
+DeNeB-C [46]
+CLB [39]
+Ours
+CA
+ASR
+IMP
+CA
+ASR
+IMP
+CA
+ASR
+IMP
+CA
+ASR
+IMP
+CIFAR10
+PseudoLabel [19]
+78.86
+89.18
+89.21
+30.36
+20.20
+0.8559
+214.70
+89.40
+25.01
+20.55
+0.8211
+208.93
+89.18
+42.03
+19.98
+0.8020
+214.70
+89.15
+91.17
+31.34
+0.9515
+24.51
+PI-Model [30]
+87.21
+86.88
+60.38
+87.26
+48.32
+87.26
+68.56
+87.05
+90.24
+MeanTeacher [38]
+90.54
+90.41
+30.11
+90.18
+30.12
+90.27
+56.88
+90.43
+88.70
+VAT [25]
+87.33
+87.29
+59.33
+87.31
+46.21
+87.12
+67.93
+86.90
+87.13
+ICT [41]
+93.26
+93.21
+49.87
+93.09
+43.21
+93.15
+57.84
+93.54
+94.54
+FixMatch [35]
+93.56
+93.39
+22.27
+93.48
+18.90
+93.65
+23.45
+93.83
+97.12
+SVHN
+PseudoLabel [19]
+86.54
+92.28
+92.09
+8.66
+20.89
+0.8361
+195.22
+92.06
+8.29
+20.59
+0.7825
+195.22
+92.21
+10.12
+20.53
+0.7741
+195.22
+92.16
+66.31
+40.15
+0.9832
+19.62
+PI-Model [30]
+92.19
+91.69
+10.26
+91.58
+9.36
+91.19
+8.24
+91.83
+75.78
+MeanTeacher [38]
+93.52
+93.19
+8.64
+93.15
+8.45
+93.18
+6.58
+93.40
+69.76
+VAT [25]
+94.16
+93.58
+7.96
+93.29
+9.58
+93.54
+5.63
+93.05
+28.71
+ICT [41]
+95.62
+95.26
+8.26
+95.34
+6.98
+95.41
+9.26
+95.21
+45.27
+FixMatch [35]
+97.10
+96.89
+67.21
+96.95
+54.63
+96.79
+75.47
+97.03
+79.59
+the loss L𝑏 is obtained.
+L𝑏 = ℓ(𝑦𝑡, F 𝑠 (𝑥𝑏;𝜃𝑠)),
+(7)
+and calculate the gradient ∇𝜃𝑠 L𝑏. Finally, optimize G so that ∇𝜃𝑠 L𝑏
+is close to ∇𝜃𝑠 L𝑡, as listed in Eq. 8.
+L𝑏𝑡 = ∥∇𝜃𝑠 L𝑡 − ∇𝜃𝑠 L𝑏 ∥2.
+(8)
+The whole loss function for crafting backdoor patterns can be
+expressed as:
+L𝑐𝑟𝑎𝑓 𝑡 = arg min
+𝜗
+E𝑥𝑠∼X𝑠 L𝑏𝑡 + 𝜆𝑖𝑛𝑠L𝑖𝑛𝑠
+s.t.
+���𝑥𝑏 − 𝑥𝑠��� ≤ 𝜀
+,
+(9)
+where 𝜆𝑖𝑛𝑠 is the hyperparameter that determines the imperceptibil-
+ity of backdoor patterns. We adopt a gradually increasing strategy
+for 𝜆𝑖𝑛𝑠, that is, multiply 𝜆𝑖𝑛𝑠 by 2 every 50 epochs, thus finding
+backdoor patterns that are as imperceptible as possible.
+Poisoning unlabeled images: According to the poisoning ratio
+𝜇𝑏, the images to be poisoned are selected from the unlabeled images
+from the target class. Then, feed them into the learned backdoor
+pattern generator G to get the corresponding backdoor patterns,
+and add them to the images to get the poisoned images. Finally,
+the poisoned images are posted on the Internet for the victim to
+scratch or secretly re-injected into the victim dataset U.
+5
+EXPERIMENT EVALUATION
+5.1
+Experiment setup
+5.1.1
+Victim network and dataset: We implement our poisoning
+on CIFAR10 [16] and SVHN [26], which are widely used in semi-
+supervised learning. CIFAR10 contains 50,000 training images and
+10,000 testing images from 10 classes. SVHN consists of 73257
+training images and 26032 testing images of house digits from
+10 classes. Moreover, CIFAR10 is trained on CNN13 [38] and SVHN
+is trained on WideResNet-28-2 [47]. To implement semi-supervised
+learning, in CIFAR10, only 4000 training images are labeled, i.e.,
+𝑁 = 4000. In SVHN, only 1000 images are labeled, i.e., 𝑁 = 1000.
+5.1.2
+SSL algorithms: We select some representative SSL algo-
+rithms from three types introduced in Section 2.3: consistency
+regularization is PI-Model [30], MeanTeacher [38], VAT [25], and
+ICT [41], pseudo-labeling is PseudoLabel [19], pseudo-labeling with
+consistency regularization is FixMatch [35]. Some of these algo-
+rithms are old and do not perform well, while others are recently
+proposed and have outstanding performance, which can fully verify
+the generality of our poisoning. The implementations of these SSL
+algorithms on SVHN and CIFAR10 come from the public Pytorch
+open source codes, [36] and [14], respectively.
+5.1.3
+Baseline poisonings: We compare our poisoning with BadNets-
+C [10], CLB [39], and DeNeB-C [46] , which have been described in
+detail in Section 3. Their backdoor patterns are all the same 8 × 8
+pixel block located at the position (20, 20) of the image, as shown
+in Fig. 6(b) 6(c), and 6(d). The target class is all 8.
+5.1.4
+Poisoning setup: In our experiments, except that the sur-
+rogate dataset for victim dataset CIFAR10 is CIFAR10, and the
+surrogate dataset for victim dataset SVHN is SVHN, the other ex-
+perimental settings are the same. The surrogate network and the
+backdoor pattern generator are WideResNet-28-2 [47] and UNet
+[32], respectively. The target class is 8, the number of poisons is 500,
+𝜆𝑖𝑛𝑠 is 0.05, 𝜖 is 27 which indicates pixel perturbation maximum.
+5.1.5
+Evaluation metrics: Evaluation metrics include three: the
+accuracy of clean examples (CA), the attack success rate (ASR) of
+the backdoor, and the imperceptibility (IMP) of backdoor patterns.
+CA: Backdoor poisoning should not degrade the accuracy of the
+SSL model, that is, the CA on the poisoned model should be close
+to the CA on the unpoisoned model.
+ASR: In the inference stage, the probability that the examples
+with backdoor patterns added are misclassified by the poisoned
+model into the target class, higher ASR means better poisoning
+performance.
+IMP: The more imperceptible backdoor patterns are, the better
+they can evade the detection of the victim. We quantify imper-
+ceptibility by computing the distance between clean images and
+poisoned images by PSNR [13], SSIM [13], and L-∞ norm. The
+larger the PSNR, the closer the SSIM is to 1, and the smaller the
+L-∞ norm, the better the imperceptibility.
+5.2
+Poisoning performance
+Let’s first verify that the poisoned unlabeled examples and gradient
+matching work in the trained-from-scratch SSL model. In the SSL
+
+Trojaning semi-supervised learning model via poisoning wild images on the web
+Conference’17, July 2017, Washington, DC, USA
+(a) Clean images
+(b) Poisoned images of BadNets
+(c) Poisoned images of CLB
+(d) Poisoned images of DeNeB
+(e) Our poisoned images
+Figure 6: Clean images and poisoned images from CIFAR10. These images from SVHN are posted on the supplementary.
+model of the 𝑘th epoch, the loss of poisoned examples taking the
+target label as the label is:
+D = ℓ(𝑦𝑡, F (P𝑡𝑟 (𝜇𝑏U);𝜃𝑘))
+(10)
+To more accurately reflect that predictions of backdoor examples
+are far away from the clean classes and close to the target class, we
+adopt relative distance to define the degree C of gradient matching.
+C = E(𝑥,𝑦) ∈X𝑣𝑎𝑙
+��∇𝜃𝑘 ℓ(𝑦𝑡, F (P𝑣𝑎𝑙 (𝑥);𝜃𝑘)) − ∇𝜃𝑘 ℓ(𝑦𝑡, F (𝑥𝑡;𝜃𝑘))
+��2
+��∇𝜃𝑘 ℓ(𝑦, F (P𝑣𝑎𝑙 (𝑥);𝜃𝑘)) − ∇𝜃𝑘 ℓ(𝑦𝑡, F (𝑥𝑡;𝜃𝑘))
+��2
+(11)
+where the upper term calculates the gradient distance between
+the backdoor example P𝑣𝑎𝑙 (𝑥) taking the target label as the label
+and the target example 𝑥𝑡 taking the target label as the label. The
+lower term calculates the gradient distance between the backdoor
+example taking the correct label as the label and the target example
+taking the target label as the label.
+Since poisoned unlabeled examples are all from the target class,
+the SSL model will correctly classify them as the target class 𝑦𝑡,
+so D will gradually decrease, as shown in the bottom of Fig. 7. As
+a result, the gradient distance C between backdoor examples and
+target examples also gradually decreases, as shown in the middle
+of Fig. 7. This brings about a gradual increase in ASR, as shown in
+the top of Fig. 7. In Table 1, we present the final poisoning results,
+from which we can draw three conclusions.
+First, SSL CAs are significantly higher than SL CAs, which means
+that the learning of SSL algorithms on unlabeled examples improves
+the model accuracy. Moreover, when poisonings are implemented,
+the CAs of poisoned models do not show significant degradations
+compared to CAs of clean models, which means that unlabeled
+backdoor poisoning can be achieved without degenerating model
+accuracy.
+Second, on these SSL algorithms, the ASRs of our poisoning are
+much higher than those of several baseline poisonings. As listed
+in Table 1, on CIFAR10, although these poisonings achieve certain
+ASRs, the highest is only 68.56%, while our lowest is 87.13%. On
+SVHN, baseline poisonings fail on all SSL algorithms except Fix-
+Match which has the ASR of 75.47%. In contrast, our poisoning
+can be applied to these SSL algorithms. Although ASRs of our poi-
+soning are lower on SVHN than on CIFAR10, the imperceptibility
+of backdoor patterns is better. If the adversary is willing to sacrifice
+imperceptibility, it will bring an increase in ASRs. In addition, differ-
+ent SSL algorithms have different vulnerabilities to our poisoning.
+On FixMatch, our poisoning performs the best, while on VAT, the
+ASR is the lowest. This may be because VAT considers adversarial
+perturbations as image augmentations and implements adversarial
+(a) PseudoLabel
+(b) PI-Model
+(c) MeanTeacher
+(d) VAT
+(e) ICT
+(f) FixMatch
+Figure 7: The evolutions of C, D, and ASR with increasing
+epochs on trained-from-scratch SSL model CNN13.
+training to ensure that unlabeled examples are resistant to these
+adversarial perturbations. The backdoor patterns we craft are simi-
+lar to adversarial perturbations, so adversarial training improves
+the model’s ability to resist poisoned examples.
+Finally, thanks to the imperceptibility design of our poisoning,
+poisoned images look very similar to clean images, as shown in
+Fig. 6. In contrast, the poisoned images of baseline poisonings have
+obvious backdoor squares, which are easily detected by victims as
+suspicious. Quantitatively, as listed in Table 1, our PNSR, SSIM and
+L-∞ norm all significantly outperform those of these schemes. To
+sum up, our poisoning well achieves the poisoning target in Eq. 2.
+5.3
+Ablation study
+To focus on the impact of varying hyperparameters or situations
+on poisoning performance, the evaluations in this section are per-
+formed on CIFAR10 trained with FixMatch. Poisoning on other
+target classes are posted on the supplementary.
+5.3.1
+Evaluation across network architectures. We evaluate the im-
+pact of different architectures of generators and surrogate networks
+on the ASR and imperceptibility. Alternative generators include
+SimNet and UNet. Alternative surrogate networks are LeNet [18],
+
+1.0
+0.8
+R 0.6
+0.4
+0.2
+8:8
+2.5
+2.0
+1.5
+1.0
+0.5
+0.0
+0.8
+0.6
+D
+0.4
+0.2
+0.0
+0
+50
+100
+150
+200
+250
+300
+350
+400
+SSL epochs1.0
+0.8
+S
+0.4
+0.2
+0.0
+1.5
+1.0
+0.5
+0.0
+0.8
+0.6
+D
+0.4
+0.2
+0.0
+0
+50
+100
+150
+200
+250
+300
+350
+400
+SSL epochs1.0
+0.8
+ASF
+0.4
+0.2
+0.0
+1.50
+1.25
+1.00
+0.75
+0.50
+0.25
+0.00
+0.30
+0.25
+0.20
+D
+0.15
+0.10
+0.05
+MW
+0.00
+0
+50
+100
+150
+200
+250
+300
+350
+400
+SSL epochs1.0
+0.8
+R
+0.6
+0.4
+0.2
+0.0
+2.0
+1.5
+C
+1.0
+0.5
+0.0
+1.25
+1.00
+0.75
+0.50
+0.25
+0.00
+0
+50
+100
+150
+200
+250
+300
+350
+400
+SSL epochs1.0
+0.8
+R 0.6
+0.4
+0.2
+2.9
+2.0
+1.5
+1.0
+0.5
+0.0
+0.5
+0.4
+0.3
+0.2
+0.1
+0
+50
+100
+150
+200
+250
+300
+350
+400
+SSL epochs1.0
+0.8
+R 0.6
+0.4
+0.2
+0.0
+1.5
+1.0
+0.5
+0.0
+0.125
+0.100
+D
+0.075
+0.050
+0.025
+0.000
+0
+50
+100
+150
+200
+250
+300
+350
+400
+SSL epochsConference’17, July 2017, Washington, DC, USA
+Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang
+Table 2: Evaluation across network architectures. SimNet is
+a simple network we build that only consists of one convo-
+lutional layer (3 × 64) for down-sampling and one deconvo-
+lutional layer (64 × 3) for up-sampling.
+Generator
+Surrogate Nets
+CA
+ASR
+IMP
+PSNR
+SSIM
+L-∞
+SimNet
+LeNet
+93.72
+99.33
+21.72
+0.6950
+27.00
+CNN13
+93.73
+90.40
+26.54
+0.8703
+26.89
+WideResNet-28-2
+93.65
+94.39
+26.63
+0.8966
+26.87
+UNet
+LeNet
+93.77
+94.37
+24.65
+0.8057
+27.00
+CNN13
+93.86
+95.24
+31.43
+0.9423
+24.88
+WideResNet-28-2
+93.83
+97.12
+31.34
+0.9515
+24.51
+CNN13, and WideResNet-28-2. The poisoning results are listed in
+Table 2. Comparing SimNet and UNet, it can be seen that because
+the network is too simple, it is more difficult for SimNet to ex-
+plore imperceptible backdoor patterns. Although a higher ASR is
+obtained on LeNet, imperceptibility is greatly sacrificed. On CNN13
+and WideResNet-28-2, UNet achieves higher ASR and better imper-
+ceptibility. Comparing these surrogate networks, WideResNet-28-2
+is larger in scale, better at exploring the least perceptible backdoor
+patterns, and obtaining higher ASR.
+5.3.2
+Evaluation across perturbation budgets 𝜖. Although a higher
+perturbation budget 𝜖 leads to more significant changes in indi-
+vidual pixels, it allows us more space to search for imperceptible
+backdoor patterns and leads to higher ASRs. As shown in Table 3,
+at 𝜖 = 7, although the maximum value of pixel perturbation is only
+7, the imperceptibility of backdoor patterns does not bring more
+significant improvement than at 𝜖 = 27, while the ASR significantly
+drops, from 97.12% to 64.40%. On the other hand, when 𝜖 = 54,
+the maximum value of pixel perturbation is improved to 32.32 com-
+pared to at 𝜖 = 27, while the ASR is only improved by 1.51%.
+Considering ASR and imperceptibility, 𝜖 = 27 is more suitable.
+(a) Poisoned images of CLB
+(b) Our poisoned images
+Figure 8: Grad-CAM [34] visualiztion.
+5.3.3
+Evaluation across other situations. In this section, we consider
+three situations of our poisoning.
+S1: Assume that the adversary does not know that the victim
+dataset is CIFAR10, but only knows the target class and dataset
+distribution. He then employs CIFAR100, which has a similar dis-
+tribution to CIFAR10, as a surrogate dataset, and replace examples
+of a certain class with examples from the target class.
+S2: Assume that the adversary does not know the labeling sit-
+uation of the examples in the victim training set. The poisoned
+unlabeled examples are then labeled proportionally.
+S3: Assume that the poisoned unlabeled examples come from
+different classes.
+Table 3: Evaluation across perturbation budgets 𝜖
+𝜖
+SSL CA
+CA
+ASR
+IMP
+PSNR
+SSIM
+L-∞
+7
+93.56
+93.62
+64.40
+33.12
+0.9502
+7.00
+27
+93.83
+97.12
+31.34
+0.9515
+24.51
+54
+93.78
+98.63
+31.57
+0.9497
+32.32
+Table 4: Evaluation across other situations
+Situation
+SSL CA
+CA
+ASR
+IMP
+PSNR
+SSIM
+L-∞
+S1
+93.56
+93.69
+94.25
+30.25
+0.9389
+24.59
+S2
+93.75
+96.89
+31.34
+0.9515
+24.51
+S3
+93.86
+8.79
+31.34
+0.9515
+24.51
+The poisoning results are listed in Table 4. In the S1 situation,
+since the distributions of CIFAR100 and CIFAR10 are similar, back-
+door patterns crafted on CIFAR100 can be migrated to CIFAR10,
+obtaining an ASR of 94.25%. In the S2 situation, since we only
+poison examples from the target class, even if these examples are
+correctly labeled, it will not impact the ASR. The poisoning failure
+(the ASR is only 8.79%) in the S3 situation further validates the
+finding in Section 3: label-inconsistent backdoor poisoning is much
+more difficult to use for unlabeled examples.
+5.4
+Defense evaluation
+The section evaluates that our poisoning can bypass five represen-
+tative defenses from four types mentioned in Section 2.2, includ-
+ing Activation Cluster [3], Neural Cleanse [42], Fine-pruning [22],
+STRIP [8], and DePuD [46]. The DePuD is posted below, and the
+other four are posted on the supplementary.
+5.4.1
+DePuD. DePuD is proposed to detect poisoned unlabeled
+examples. First, all the training examples are divided into two cate-
+gories according to whether they are labeled. The labeled ones are
+assigned the label 0, and the unlabeled ones are assigned the label
+1. These examples are then classified using a heavy regularization
+model. If poisoned examples have significant backdoor patterns,
+the predictions will be extremely close to 1, and the separation from
+clean unlabeled examples will appear, so that it is detected as abnor-
+mal. DePuD works for CLB, as shown in Fig. 8(a), backdoor patterns
+can be detected prominently in the lower right corner. However,
+the backdoor patterns of our poisoning are imperceptible, it is diffi-
+cult to be captured by the heavy regularization model, as shown in
+Fig. 8(b). Moreover, as shown in Fig. 9, clean unlabeled examples
+almost overlap with poisoned unlabeled examples, which means
+that DePuD cannot separate the poisoned unlabeled examples.
+(a)
+(b)
+Figure 9: DePuD detection.
+
+DePuD detection on poisoned SVHN
+1.0
+Labeledtrainingimages
+Cleanunlabeledtrainingimages
+0.8
+Poisoned unlabeledtraining images
+Distribution
+0.6
+0.4
+0.2
+0.0
+0.0
+0.2
+0.4
+0.6
+0.8
+1.0
+PredictionDePuD detectiononpoisonedCiFARlo
+1.0
+Labeledtrainingimages
+Cleanunlabeledtrainingimages
+0.8
+Poisoned unlabeledtraining images
+Distribution
+0.6
+0.4
+0.2
+0.0
+0.0
+0.2
+0.4
+0.6
+0.8
+1.0
+PredictionTrojaning semi-supervised learning model via poisoning wild images on the web
+Conference’17, July 2017, Washington, DC, USA
+6
+CONCLUSION
+This paper is the first to investigate the vulnerability of unlabeled
+examples of trained-from-scratch SSL models to backdoor poison-
+ing, revealing the flaws in the security design of SSL algorithms.
+We first find that label-inconsistent backdoor poisoning cannot be
+used for unlabeled examples due to the opposition to the SSL algo-
+rithms that strive to correctly learn unlabeled examples. Thus, for
+unlabeled examples, poisoning only is implemented on examples
+from the target class. Based on this, we propose a zero-knowledge
+and imperceptible backdoor poisoning. Experiments show that our
+poisoning achieves state-of-the-art attack success rates when by-
+passing various defenses.
+REFERENCES
+[1] Berthelot, D.; Carlini, N.; Cubuk, E. D.; Kurakin, A.; Sohn, K.; Zhang, H.; and Raf-
+fel, C. 2019. ReMixMatch: Semi-Supervised Learning with Distribution Matching
+and Augmentation Anchoring. In International Conference on Learning Represen-
+tations.
+[2] Berthelot, D.; Carlini, N.; Goodfellow, I.; Papernot, N.; Oliver, A.; and Raffel, C. A.
+2019. Mixmatch: A holistic approach to semi-supervised learning. Advances in
+Neural Information Processing Systems, 32.
+[3] Chen, B.; Carvalho, W.; Baracaldo, N.; Ludwig, H.; Edwards, B.; Lee, T.; Molloy, I.;
+and Srivastava, B. 2018. Detecting backdoor attacks on deep neural networks by
+activation clustering. arXiv preprint arXiv:1811.03728.
+[4] Chen, H.; Fu, C.; Zhao, J.; and Koushanfar, F. 2019. DeepInspect: A Black-box
+Trojan Detection and Mitigation Framework for Deep Neural Networks. In IJCAI,
+volume 2, 8.
+[5] Chen, T.; Kornblith, S.; Norouzi, M.; and Hinton, G. 2020. A simple framework
+for contrastive learning of visual representations. In International conference on
+machine learning, 1597–1607. PMLR.
+[6] Doan, K.; Lao, Y.; Zhao, W.; and Li, P. 2021. LIRA: Learnable, Imperceptible and
+Robust Backdoor Attacks. In Proceedings of the IEEE/CVF International Conference
+on Computer Vision, 11966–11976.
+[7] Dong, Y.; Yang, X.; Deng, Z.; Pang, T.; Xiao, Z.; Su, H.; and Zhu, J. 2021. Black-box
+detection of backdoor attacks with limited information and data. In Proceedings
+of the IEEE/CVF International Conference on Computer Vision, 16482–16491.
+[8] Gao, Y.; Xu, C.; Wang, D.; Chen, S.; Ranasinghe, D. C.; and Nepal, S. 2019. Strip:
+A defence against trojan attacks on deep neural networks. In Proceedings of the
+35th Annual Computer Security Applications Conference, 113–125.
+[9] Gidaris, S.; Singh, P.; and Komodakis, N. 2018. Unsupervised Representation
+Learning by Predicting Image Rotations. In International Conference on Learning
+Representations.
+[10] Gu, T.; Dolan-Gavitt, B.; and Garg, S. 2017. Badnets: Identifying vulnerabilities
+in the machine learning model supply chain. arXiv preprint arXiv:1708.06733.
+[11] Guo, J.; Li, A.; and Liu, C. 2021. AEVA: Black-box Backdoor Detection Using
+Adversarial Extreme Value Analysis. arXiv preprint arXiv:2110.14880.
+[12] He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep residual learning for image
+recognition. In Proceedings of the IEEE conference on computer vision and pattern
+recognition, 770–778.
+[13] Hore, A.; and Ziou, D. 2010. Image quality metrics: PSNR vs. SSIM. In 2010 20th
+international conference on pattern recognition, 2366–2369. IEEE.
+[14] iBelieveCJM. 2020. Tricks of Semi-supervised Deep Leanring–Pytorch. https:
+//github.com/iBelieveCJM/Tricks-of-Semi-supervisedDeepLeanring-Pytorch.
+[15] Iscen, A.; Tolias, G.; Avrithis, Y.; and Chum, O. 2019. Label propagation for deep
+semi-supervised learning. In Proceedings of the IEEE/CVF Conference on Computer
+Vision and Pattern Recognition, 5070–5079.
+[16] Krizhevsky, A.; Hinton, G.; et al. 2009. Learning multiple layers of features from
+tiny images.
+[17] Laine, S.; and Aila, T. 2016. Temporal ensembling for semi-supervised learning.
+arXiv preprint arXiv:1610.02242.
+[18] LeCun, Y.; Bottou, L.; Bengio, Y.; and Haffner, P. 1998. Gradient-based learning
+applied to document recognition. Proceedings of the IEEE, 86(11): 2278–2324.
+[19] Lee, D.-H.; et al. 2013. Pseudo-label: The simple and efficient semi-supervised
+learning method for deep neural networks. In Workshop on challenges in repre-
+sentation learning, ICML, volume 3, 896.
+[20] Li, X.; Chen, Z.; Zhao, Y.; Tong, Z.; Zhao, Y.; Lim, A.; and Zhou, J. T. 2021. PointBA:
+Towards Backdoor Attacks in 3D Point Cloud. In Proceedings of the IEEE/CVF
+International Conference on Computer Vision, 16492–16501.
+[21] Li, Y.; Li, Y.; Wu, B.; Li, L.; He, R.; and Lyu, S. 2021. Invisible backdoor attack with
+sample-specific triggers. In Proceedings of the IEEE/CVF International Conference
+on Computer Vision, 16463–16472.
+[22] Liu, K.; Dolan-Gavitt, B.; and Garg, S. 2018. Fine-pruning: Defending against
+backdooring attacks on deep neural networks. In International Symposium on
+Research in Attacks, Intrusions, and Defenses, 273–294. Springer.
+[23] Liu, Y.; Ma, X.; Bailey, J.; and Lu, F. 2020. Reflection backdoor: A natural backdoor
+attack on deep neural networks. In European Conference on Computer Vision,
+182–199. Springer.
+[24] Mahajan, D.; Girshick, R.; Ramanathan, V.; He, K.; Paluri, M.; Li, Y.; Bharambe,
+A.; and Van Der Maaten, L. 2018. Exploring the limits of weakly supervised
+pretraining. In Proceedings of the European conference on computer vision (ECCV),
+181–196.
+[25] Miyato, T.; Maeda, S.-i.; Koyama, M.; and Ishii, S. 2018. Virtual adversarial
+training: a regularization method for supervised and semi-supervised learning.
+IEEE transactions on pattern analysis and machine intelligence, 41(8): 1979–1993.
+[26] Netzer, Y.; Wang, T.; Coates, A.; Bissacco, A.; Wu, B.; and Ng, A. Y. 2011. Reading
+digits in natural images with unsupervised feature learning.
+[27] Nguyen, T. A.; and Tran, A. 2020. Input-aware dynamic backdoor attack. Advances
+in Neural Information Processing Systems, 33: 3454–3464.
+[28] Nguyen, T. A.; and Tran, A. T. 2020. WaNet-Imperceptible Warping-based Back-
+door Attack. In International Conference on Learning Representations.
+[29] Pham, H.; Dai, Z.; Xie, Q.; and Le, Q. V. 2021. Meta pseudo labels. In Proceedings
+of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11557–
+11568.
+[30] Rasmus, A.; Berglund, M.; Honkala, M.; Valpola, H.; and Raiko, T. 2015. Semi-
+supervised learning with ladder networks. Advances in neural information pro-
+cessing systems, 28.
+[31] Redmon, J.; Divvala, S.; Girshick, R.; and Farhadi, A. 2016. You only look once:
+Unified, real-time object detection. In Proceedings of the IEEE conference on
+computer vision and pattern recognition, 779–788.
+[32] Ronneberger, O.; Fischer, P.; and Brox, T. 2015. U-net: Convolutional networks
+for biomedical image segmentation. In International Conference on Medical image
+computing and computer-assisted intervention, 234–241. Springer.
+[33] Salem, A.; Wen, R.; Backes, M.; Ma, S.; and Zhang, Y. 2020. Dynamic backdoor
+attacks against machine learning models. arXiv preprint arXiv:2003.03675.
+[34] Selvaraju, R. R.; Cogswell, M.; Das, A.; Vedantam, R.; Parikh, D.; and Batra, D.
+2017. Grad-cam: Visual explanations from deep networks via gradient-based
+localization. In Proceedings of the IEEE international conference on computer vision,
+618–626.
+[35] Sohn, K.; Berthelot, D.; Carlini, N.; Zhang, Z.; Zhang, H.; Raffel, C. A.; Cubuk, E. D.;
+Kurakin, A.; and Li, C.-L. 2020. Fixmatch: Simplifying semi-supervised learning
+with consistency and confidence. Advances in Neural Information Processing
+Systems, 33: 596–608.
+[36] Suzuki, T. 2020. Consistency Regularization for Semi-supervised Learning with
+PyTorch. https://github.com/perrying/pytorch-consistency-regularization.
+[37] Szegedy, C.; Ioffe, S.; Vanhoucke, V.; and Alemi, A. A. 2017.
+Inception-v4,
+inception-resnet and the impact of residual connections on learning. In Thirty-
+first AAAI conference on artificial intelligence.
+[38] Tarvainen, A.; and Valpola, H. 2017. Mean teachers are better role models:
+Weight-averaged consistency targets improve semi-supervised deep learning
+results. Advances in neural information processing systems, 30.
+[39] Turner, A.; Tsipras, D.; and Madry, A. 2019. Label-consistent backdoor attacks.
+arXiv preprint arXiv:1912.02771.
+[40] Van der Maaten, L.; and Hinton, G. 2008. Visualizing data using t-SNE. Journal
+of machine learning research, 9(11).
+[41] Verma, V.; Kawaguchi, K.; Lamb, A.; Kannala, J.; Bengio, Y.; and Lopez-Paz, D.
+2019. Interpolation consistency training for semi-supervised learning. arXiv
+preprint arXiv:1903.03825.
+[42] Wang, B.; Yao, Y.; Shan, S.; Li, H.; Viswanath, B.; Zheng, H.; and Zhao, B. Y. 2019.
+Neural cleanse: Identifying and mitigating backdoor attacks in neural networks.
+In 2019 IEEE Symposium on Security and Privacy (SP), 707–723. IEEE.
+[43] Wang, R.; Zhang, G.; Liu, S.; Chen, P.-Y.; Xiong, J.; and Wang, M. 2020. Practical
+detection of trojan neural networks: Data-limited and data-free cases. In European
+Conference on Computer Vision, 222–238. Springer.
+[44] Xie, Q.; Dai, Z.; Hovy, E.; Luong, T.; and Le, Q. 2020. Unsupervised data aug-
+mentation for consistency training. Advances in Neural Information Processing
+Systems, 33: 6256–6268.
+[45] Yan, Z.; Li, G.; TIan, Y.; Wu, J.; Li, S.; Chen, M.; and Poor, H. V. 2021. DeHiB: Deep
+hidden backdoor attack on semi-supervised learning via adversarial perturbation.
+In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 10585–
+10593.
+[46] Yan, Z.; Wu, J.; Li, G.; Li, S.; and Guizani, M. 2021. Deep Neural Backdoor in
+Semi-Supervised Learning: Threats and Countermeasures. IEEE Transactions on
+Information Forensics and Security, 16: 4827–4842.
+[47] Zagoruyko, S.; and Komodakis, N. 2016. Wide residual networks. arXiv preprint
+arXiv:1605.07146.
+[48] Zhao, S.; Ma, X.; Zheng, X.; Bailey, J.; Chen, J.; and Jiang, Y.-G. 2020. Clean-label
+backdoor attacks on video recognition models. In Proceedings of the IEEE/CVF
+Conference on Computer Vision and Pattern Recognition, 14443–14452.
+
+Conference’17, July 2017, Washington, DC, USA
+Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang
+[49] Zhao, Z.-Q.; Zheng, P.; Xu, S.-t.; and Wu, X. 2019. Object detection with deep
+learning: A review. IEEE transactions on neural networks and learning systems,
+30(11): 3212–3232.
+APPENDIX
+A
+SVHN IMAGES
+In Fig. 12, we show clean images and poisoned images on SVHN. It
+can be seen that our poisoned images are visually very similar to
+the clean images, and the victim is difficult to detect.
+B
+EVALUATION ACROSS TARGET CLASSES
+We select other classes to act as target classes. The poisoning results
+are listed in Table 5. First, likewise, poisoning on other target classes
+does not degrade the model accuracy. Secondly, it can be seen that
+the poisoning difficulty of different target classes is different. For
+example, on the target class Bird, the ASR can reach 99.52%, while
+on Truck, the ASR is lower, 81.88%. Of course, we also can sacrifice
+a little backdoor pattern imperceptibility to improve the ASR.
+C
+DEFENSE EVALUATION
+We evaluate our poisoning on Activation Cluster, Neural Cleanse,
+Fine-pruning, and STRIP.
+C.1
+Activation Cluster
+The process of Activation Cluster is to input all training examples
+into the already trained victim model, thereby obtaining the activa-
+tion of these examples in the last hidden layer. These activations
+are then divided into different clusters based on their labels. Finally,
+it is determined whether there are poisoned examples by detecting
+the abnormality of these clusters. However, our backdoor poison-
+ing does not rely on labels and poisons only unlabeled examples.
+Thus, poisoned unlabeled examples cannot be divided into different
+clusters based on labels. Thus, our poisoning can naturally bypass
+Activation Cluster detection.
+C.2
+Neural Cleanse
+On potentially poisoned models, reverse-engineer the minimum-
+intensity backdoor triggers for all classes. They then determine
+whether a certain class is the target class based on the prior knowl-
+edge that the target class injected into the backdoor has a trigger
+with abnormally small intensity, i.e., 𝑎𝑛𝑜𝑚𝑎𝑙𝑦 𝑖𝑛𝑑𝑒𝑥 > 2. In semi-
+supervised learning, the defender may not have enough labeled
+examples for more accurate reverse engineering, but we assume
+the most stringent condition that the defender has enough labeled
+examples. However, even so, as shown in the Fig. 10(a), the trigger
+intensity of the target class injected into the backdoor is not signifi-
+cant outliers, and 𝑎𝑛𝑜𝑚𝑎𝑙𝑦 𝑖𝑛𝑑𝑒𝑥 is all less than 2 (Fig. 10(b)). Thus,
+our poisoning can successfully bypass Neural Cleanse detection.
+We think this is because reverse engineering in Nerucal Cleanse
+detection relies on classification layers, whereas our poisoning is
+gradient matching that controls the entire network.
+C.3
+Fine-pruning
+This is a blind defense strategy, instead of detecting whether the
+model or example is poisoned, it uses pruning and fine-tuning for
+(a)
+(b)
+Figure 10: Neural Cleanse detection.
+Table 5: Evaluation across target classes
+Target class
+SSL CA
+CA
+ASR
+IMP
+PSNR
+SSIM
+L-∞
+Airplane
+93.56
+93.49
+99.15
+33.29
+0.9518
+24.26
+Automobile
+93.44
+89.22
+30.21
+0.9490
+24.88
+Bird
+93.76
+99.52
+33.08
+0.9492
+23.99
+Cat
+93.95
+99.95
+33.33
+0.9655
+20.06
+Deer
+93.78
+98.18
+32.51
+0.9560
+22.18
+Dog
+93.81
+91.08
+32.32
+0.9634
+23.81
+Frog
+93.89
+95.64
+31.47
+0.9519
+23.22
+Horse
+93.84
+94.11
+32.36
+0.9653
+22.91
+Ship
+93.83
+97.12
+31.34
+0.9515
+24.51
+Truck
+93.93
+81.88
+32.65
+0.9765
+23.29
+any model to try to eliminate possible backdoors. Specifically, clean
+examples are first fed into the model, and then 𝛼% (i.e., Pruning
+rate) of neurons with minimal activation are dormant by pruning,
+thereby attempting to remove possible backdoors. Fine-tuning is
+then used to compensate for the degradation of clean example
+accuracy caused by pruning. As shown in Fig. 11, even with a
+pruning rate of 90%, there is no significant drop in the ASR. When
+the pruning rate is 99%, on CIFAR10, the CA drops to 80%, and the
+ASR is still 62%. An interesting phenomenon is that on SVHN, the
+ASR increases significantly, which may be because the excessive
+pruning makes the model’s ability to distinguish clean examples
+weakened, which makes it easier to be misclassified as the target
+class once backdoor patterns are added.
+(a)
+(b)
+Figure 11: Fine-pruning.
+C.4
+STRIP
+STRIP is deployed in the model inference stage. Before the test-
+ing example is fed to the model, it is synthesized with a set of
+pre-prepared clean examples. Then these obtained synthesized ex-
+amples are fed into the model for prediction. If the entropy of their
+prediction results is abnormally small, it is determined that the
+testing example is a backdoor example, and the model is poisoned.
+
+Trigger intensity of all class
+100
+90
+80
+70
+60
+Target class
+50
+Clean
+Poisoned
+Clean
+Poisoned
+SVHN
+SVHN
+CIFAR10
+CIFAR10Neural cleanse detection result
+3.0
+2.5
+Anomaly index
+2.0
+1.5
+1.37
+1.29
+1.16
+1.0
+0.5
+0.19
+0.0
+Clean
+Poisoned
+Clean
+Poisoned
+SVHN
+SVHN
+CIFAR10
+CIFAR10Fine-pruning on SVHN
+1.0
+1.0
+由
+0.8
+0.8
+rate
+success
+0.6
+0.6
+0.4
+0.2
+0.2
+ASR
+中
+BMA
+0.0
+0.0
+0.00
+0.10
+0.20
+0.30
+0.50
+0.60
+0.70
+0.80
+0.90
+0.99
+Pruning rateFine-pruning on CIFAR10
+1.0
+1.0
+由
+由
+0.8
+0.8
+0.6
+0.6
+ASR
+0.4
+0.4
+0.2
+0.2
+ASR
+中
+CA
+0.0
+0.0
+0.00
+0.10
+0.20
+0.30
+0.400.50
+0.60
+0.70
+0.80
+0.90
+0.99
+Pruning rateTrojaning semi-supervised learning model via poisoning wild images on the web
+Conference’17, July 2017, Washington, DC, USA
+(a) Clean images
+(b) Poisoned images of BadNets
+(c) Poisoned images of CLB
+(d) Poisoned images of DeNeB
+(e) Our poisoned images
+Figure 12: Clean images and poisoned images on SVHN.
+As shown in Fig. 13, we show the entropy distribution for 500
+testing examples and 500 backdoor examples, and it can be seen
+that the distributions almost coincide. The entropy of the backdoor
+examples does not exhibit abnormally small property. Thus, our
+backdoor poisoning can bypass STRIP detection.
+(a)
+(b)
+Figure 13: STRIP detection.
+
+EntropydistributionofSVHN
+clean
+0.12
+backdoor
+0.10
+0.08
+0.06
+0.04
+0.02
+0.00
+0.0
+0.1
+0.2
+0.3
+0.4
+0.5
+0.6EntropydistributionofClFAR1o
+0.12
+clean
+backdoor
+0.10
+0.08
+0.06
+0.04
+0.02
+0.00
+0.0
+0.2
+0.4
+0.6
+0.8
+1.0
+1.2
+1.4
\ No newline at end of file
diff --git a/jNAyT4oBgHgl3EQfkfir/content/tmp_files/load_file.txt b/jNAyT4oBgHgl3EQfkfir/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..5bda4bd643ae26d83b260f4a46c0704af98a0d95
--- /dev/null
+++ b/jNAyT4oBgHgl3EQfkfir/content/tmp_files/load_file.txt
@@ -0,0 +1,1617 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf,len=1616
+page_content='Trojaning semi-supervised learning model via poisoning wild images on the web Le Feng Fudan University China Zhengxing Qian Fudan University China Sheng Li Fudan University China Xinpeng Zhang Fudan University China ABSTRACT Wild images on the web are vulnerable to backdoor (also called trojan) poisoning, causing machine learning models learned on these images to be injected with backdoors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Most previous attacks assumed that the wild images are labeled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In reality, however, most images on the web are unlabeled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, we study the effects of unlabeled backdoor images under semi-supervised learning (SSL) on widely studied deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To be realistic, we assume that the adversary is zero-knowledge and that the semi-supervised learning model is trained from scratch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Firstly, we find the fact that backdoor poisoning always fails when poisoned unlabeled images come from different classes, which is different from poisoning the labeled images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The reason is that the SSL algorithms always strive to correct them during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Therefore, for unlabeled images, we implement backdoor poisoning on images from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then, we propose a gradient matching strategy to craft poisoned images such that their gradients match the gradients of target im- ages on the SSL model, which can fit poisoned images to the target class and realize backdoor injection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To the best of our knowledge, this may be the first approach to backdoor poisoning on unlabeled images of trained-from-scratch SSL models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Experiments show that our poisoning achieves state-of-the-art attack success rates on most SSL algorithms while bypassing modern backdoor defenses.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' KEYWORDS Backdoor, Semi-supervised Learning, Trained-from-scratch, Neural Network 1 INTRODUCTION The excellent performance of deep neural networks [12, 31, 37, 49] is largely due to numerous training examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To obtain enough train- ing examples, trainers usually grab them from the web.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, these examples from the wild may not be safe, they are vulnerable to backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Previous backdoor poisonings [10, 21, 28, 39, 45] mainly focus on the labeled examples, which rely on the guidance of the target label to inject backdoors into the models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Yan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [45, 46] initially propose two unlabeled backdoor poisoning schemes for pre-trained SSL models: DeNeB [45] and DeHiB [46].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' They assume that the SSL learner first trains the model on labeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The obtained pre-trained model is then fine-tuned using the SSL algorithm in combination with unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Actu- ally, most advanced SSL algorithms are end-to-end, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', unlabeled examples along with labeled examples are fed into the model to train from scratch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' There is no intermediate pre-trained model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Be- sides, the backdoor patterns proposed by Yan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' are perceptible, which can be detected by DePuD [46].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Adversary Clean Ship Imperceptible backdoor patterns Backdoor Ship Labeled part SSL training set ⋯ ⋯ Bird Car Plane Ship Unlabeled part Victim network Trained-from- scratch SSL + = Poisoning Victim Ship Ship Ship Ship Feeding Predicting Unseen backdoor images Backdoor network Adversary Releasing Figure 1: Attack pipelines of unlabeled backdoor poisoning on the trained-from-scratch SSL model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Assuming that the target class is Ship.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To be practical and stealthy, we propose a zero-knowledge and imperceptible backdoor poisoning on unlabeled examples of trained- from-scratch SSL models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zero-knowledge means that the adversary does not require knowledge of the victim model, the SSL algorithm, the complete dataset, and the training process.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Our attack pipeline is shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The adversary adds the imperceptible backdoor patterns to the clean Ship images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then use the resulting backdoor images to poison the unlabeled part of the SSL training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The victim uses the poisoned training set to train the network from scratch by the SSL algorithm, thus causing the backdoor to be injected inadvertently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Finally, in the inference stage, the images with backdoor patterns will be misclassified as the target class Ship.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, first of all, we find that unlike poisoning labeled examples, since SSL algorithms strive to learn correctly the unla- beled examples, if the poisoned examples are from different classes, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', label-inconsistent backdoor poisoning, they will be re-learned into the correct class by the trained-from-scratch SSL model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As a result, the backdoor cannot be injected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On the contrary, if poi- soning only is implemented on the examples from the target class, arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00435v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='CY] 1 Jan 2023 Conference’17, July 2017, Washington, DC, USA Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', label-consistent backdoor poisoning, they will end up being classified into the target class by the SSL model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' And since various regularizations of SSL algorithms mitigate overfitting on the target class, backdoor patterns can be generalized to non-target classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, our poisoning is only implemented on the examples from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then, to achieve zero-knowledge poisoning, we resort to the transferability of neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Only the target class and the distribution of the victim dataset are required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, We first prepare a surrogate dataset with a similar distribution to the victim dataset containing the examples from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The surrogate network is then trained on the surrogate dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For the obtained surrogate network, we train a backdoor pattern generator that takes clean examples from different classes as input and outputs the corresponding imperceptible backdoor patterns, and then adds the backdoor patterns to the examples to get poisoned examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Besides the imperceptibility of backdoor patterns, the other train- ing target of the generator is to achieve gradient matching which is proposed to make the gradients of poisoned examples match the gradients of the target examples on the surrogate model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We hope that in the trained-from-scratch SSL model, through gradient matching, the poisoned examples can be naturally learned into the target class as the target examples are learned into the target class, thereby injecting the backdoor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In summary, our contributions are as follows: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To the best of our knowledge, we are the first to investigate the vulnerability of unlabeled examples of trained-from-scratch SSL models to backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We find that for unlabeled examples of trained-from-scratch SSL models, label-consistent backdoor poisoning is more effective.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We propose a zero-knowledge and imperceptible backdoor poisoning on unlabeled examples of trained-from-scratch SSL mod- els.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We implement our poisoning on SSL algorithms of three types.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Attack success rates are significantly higher than baseline poison- ings, and ours can successfully bypass various defenses including DePuD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2 RELATED WORK 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 Backdoor poisoning If the poisoned examples are all from the target class, it is called label-consistent backdoor poisoning, otherwise, it is label-inconsistent backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Label-inconsistent backdoor poisoning: BadNets [10] is the first backdoor poisoning for neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The scheme is rudi- mentary so that numerous backdoor defenses [3, 8, 11, 22, 42] can de- tect or remove the backdoor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Many follow-up works propose more threatening poisonings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Invisible backdoor patterns [21, 23, 28] make it difficult for victims to visually detect the abnormality of the poisoned examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dynamic backdoor patterns [27, 33] can make it difficult for victims to capture the regular pattern of back- door patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' There are also schemes [6, 21] that achieve invisible and dynamic backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DeHiB [45] and DeNeB [46] that poison unlabeled examples of pre-trained models are also label- inconsistent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Label-consistent backdoor poisoning: Due to not changing the correct labels of poisoned examples, it is more stealthy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' How- ever, backdoor patterns may overfit on the target class and fail to generalize to non-target classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' CLB (Clean Label backdoor) [39] first proposes to mitigate this overfitting by adversarial per- turbation and interpolation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Later, [48] implements this backdoor poisoning on the videos.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [20] implements this backdoor poisoning on the point clouds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 Backdoor defense For different phases of backdoor poisoning, backdoor defenses can be categorized into four types: pre-training defense, post-training defense, testing-time defense, and blind defense.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Pre-training defense: The defender checks training examples to determine whether there are suspicious examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For label- inconsistent poisoning in supervised learning, poisoned examples can be screened by the inconsistency between the content of the examples and their labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Formally, activation clustering [3] can detect outlier examples by clustering training examples according to their labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Recently, DePuD [46] is proposed to detect unlabeled poisoned examples in semi-supervised learning, which uses heavy regularization to distinguish suspicious unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Post-training defense: This defense is to detect anomalies in the learned model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A typical detection is Neural Cleanse [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Re- verse engineering is first used for all classes to get their triggers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' If the trigger intensity of the class is abnormally smaller than those of other classes, this class is detected as a backdoor class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Later, many variants based on Neural Cleanse appear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For example, [4, 43] improve Neural Cleanse with better objective functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [7, 11] propose the detection in black box scenarios.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Testing-time defense: The defense is deployed during the model testing phase.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The testing example is checked.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A typical detection is STRIP [8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The testing example is fused with a set of pre-prepared clean examples to obtain synthetic examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then, feed these syn- thetic examples to the model for prediction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' If prediction results present a low-entropy distribution, then the testing example may be a backdoor example.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Blind defense: Instead of detecting examples or models, unified operations against the examples or model are adopted.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Data aug- mentation is a natural blind defense method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In the testing phase, processing such as JPEG compression on the examples may also destroy backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Fine-pruning [22] prunes and fine-tunes the model to try to destroy possible backdoors in the model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 Semi-Supervised Learning Existing SSL algorithms can be categorized into three types: consis- tency regularization [17, 25, 30, 38, 41, 44], pseudo-labeling [15, 19, 29], and pseudo-labeling with consistency regularization [1, 2, 35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Consistency regularization: It assumes that randomness within the neural network or data augmentation transformations should not modify model predictions given the same input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For example, PI-Model [30] minimizes the difference between two passes through the network with stochastic transformations for the same point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' MeanTeacher [38] minimizes the difference between the predic- tions of the student model and the teacher model for the same point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Trojaning semi-supervised learning model via poisoning wild images on the web Conference’17,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' July 2017,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Washington,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DC,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' USA ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Bird ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Car ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Plane ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Adding backdoor ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='patterns ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Changing their labels ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='to the target label ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Clean labeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Poisoned labeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='(a) Label-inconsistent poisoning in SL ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Adding backdoor ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='patterns ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Keeping their labels ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='unchanged ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Clean labeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Poisoned labeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='(b) Label-consistent poisoning in SL ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Adding backdoor ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='patterns ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Poisoned unlabeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Clean unlabeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='(c) Label-inconsistent poisoning in SSL ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Adding backdoor ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='patterns ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Poisoned unlabeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Clean unlabeled images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='(d) Label-consistent poisoning in SSL ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Backdoor network ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Adding backdoor ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='patterns ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Ship ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Feeding ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Predicting ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Unseen clean images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Unseen backdoor images ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Predicted classes ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Backdoor ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='network ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='(e) Poisoning target ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='Figure 2: Backdoor poisoning in SL and SSL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In these poisonings, the target class is "ship".' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Their poisoning targets are the same, which all cause unseen backdoor images to be misclassified as "ship" by the backdoor network trained on the poisoned training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' VAT [25], ICT [41], and UDA [44] aim to develop more efficient augmentations to exploit unlabeled data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Pseudo-labeling: It assigns pseudo labels to unlabeled exam- ples based on the predictions of the current model and then trains unlabeled examples by supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For example, pseudo labeling [19] uses the pretrained network trained on the labeled examples to predict pseudo labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' MPL (Meta Pseudo Labeling) [29] maintains two models: a student model and a teacher model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The teacher model predicts unlabeled examples to give pseudo labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Pseudo-labeling with consistency regularization: MixMatch [2] uses MixUp augmentation to create multiple augmentations for each unlabeled example, and then takes the maximum class of the average of the predictions of these augmentations as the pseudo label.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ReMixMatch [1] improves MixMatch by introducing two new mechanisms: distribution alignment and augmentation anchoring.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' FixMatch [35] performs weak augmentation and strong augmenta- tion for each unlabeled example, and the predicted label of weak augmentation is used as the pseudo label of strong augmentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 3 OUR NOVEL FINDING In the context of supervised learning, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2(a) and Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2(b), both label-consistent and label-inconsistent backdoor poison- ings rely on the guidance of the target label.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The difference is that label-inconsistent backdoor poisoning changes their labels to target labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' This is not required for label-consistent backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, this also leads to the fact that since backdoor patterns are not added to the non-target class examples, the model may overfit backdoor patterns on the target class, so that backdoor patterns do not work on non-target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Although CLB [39] proposes interpo- lation and adversarial perturbation to improve the generalization of backdoor patterns on non-target classes, attack success rates are lower than label-inconsistent backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, label- inconsistent backdoor poisoning is easier to be implemented than label-consistent backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, in the context of semi-supervised learning, on the one hand, backdoor poisoning on unlabeled examples will lose the guid- ance of the target label, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2(c) and 2(d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On the other hand, the difference in the mechanism of semi-supervised learning and supervised learning brings a novel finding: Since the semi-supervised learning algorithms strive to correctly learn unlabeled examples through various regularizations, for unlabeled examples, label-inconsistent backdoor poisoning is much more diffi- cult to implement than label-consistent backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Next, we will experimentally verify our finding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We use exist- ing schemes BadNets [10], CLB [39], and DeNeB [46] to poison unlabeled examples of trained-from-scratch SSL models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Note that although some recent backdoor poisonings, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', invisible backdoor poisonings [21, 23, 28], are better at resisting backdoor defenses, BadNets is still excellent in terms of attack success rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Likewise, in the context of a pretrained network, the attack success rate of DeNeB is much higher than that of DeHiB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Figure 3: Poisoning unlabeled examples of trained-from- scratch SSL model using existing schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The tested vic- tim dataset is CIFAR10 [16] and the network is CNN13 [38].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In the three backdoor poisoning schemes, the backdoor pat- terns are all 8 × 8 pixel squares.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The target class is 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The attack success rates are shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Both BadNets and DeNeB fail to poison completely, and the attack success rates are close to the probability 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00% of random classification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In contrast, CLB obtains certain attack success rates (56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88%, 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03%, 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='45%).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2(c), for label-inconsistent backdoor Attack success rate of existing schemes 100 MeanTeacher[41] 80 Pseudolabeling[2o] Attack success rate FixMatch [38] 60 56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03 40 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1130.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='36 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='45 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='27 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='01 20 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='64 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='65 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='84 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='45 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='49 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 0 BadNets CLB DeNeB BadNets-C DeNeB-CConference’17, July 2017, Washington, DC, USA Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang poisoning, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', BadNets and DeNeB, pseudo-labeling based SSL algorithms [1, 2, 15, 19, 29, 35] strive to assign correct labels to un- labeled examples, while the poisoned unlabeled examples coming from different classes expect themselves to be misclassified into the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' This opposition makes backdoor patterns difficult to be learned.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Likewise, when consistency regularization based SSL algorithms [1, 2, 17, 25, 30, 35, 38, 41, 44] are employed, the noises or augmentations the SSL algorithms add to the examples or models will make the models to unlearn backdoor patterns but to focus on the semantic information of the poisoned unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 4, as SSL proceeds, the poisoned unlabeled exam- ples are gradually classified into their respective correct classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, for label-consistent backdoor poisoning, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', CLB, since poisoning only is implemented on examples from the target class (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2(d)), SSL algorithms classify all of them into the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Such opposition does not exist.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Moreover, various regularizations of SSL algorithms prevent the model from overfitting on the target class, so backdoor patterns can be slightly generalized to non-target classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To further verify our finding, we generalize DeNeB and BadNets to label-consistent versions DeNeB-C and BadNets-C, where C indicates consistent.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' With all settings unchanged, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 3, the attack success rates have been significantly improved, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', for DeNeB, the increase from 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='45%, 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='49%, 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26% to 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12%, 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='01%, 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, CLB, DeNeB-C, and BadNets-C have three significant shortcomings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (1) The attack success rate is not ideal, the highest is only 56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (2) The backdoor patterns are perceptible and easily detected by the victim as suspicious, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 6(b), 6(c), and 6(d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (3) DePuD [46], a detection solution for poisoned unlabeled examples, can detect the anomaly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To remedy these shortcomings, we propose a zero-knowledge and imperceptible backdoor poisoning on unlabeled examples of trained-from-scratch SSL models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (a) (b) (c) Figure 4: t-SNE [40] feature distribution of poisoned un- labeled examples in label-inconsistent backdoor poisoning DeNeB in the trained-from-sratch SSL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The SSL algorithm is FixMatch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 4 OUR METHOD 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 Threat model Assume that the victim who trains a neural network model has only limited labeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To improve model performance, he intends to scrape more unlabeled examples from the web for semi- supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For example, a state-of-the-art image classifier [24] scrapes 1 billion images from Instagram.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' At this point, an adversary who can upload data to the network can control a portion of the unlabeled examples, thereby realizing backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Since our attack is zero-knowledge, an adversary has very limited knowledge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, what an adversary cannot obtain are: (1) The architecture, weights, and outputs of the trained-from- scratch victim model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (2) The training process, hyperparameter settings, and the SSL algorithm employed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (3) The complete victim dataset and whether the examples are labeled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The only knowledge an adversary can obtain is: (1) The distribution Z of the victim dataset and the target class 𝑦𝑡 of poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Formally, for a victim model F parameterized by 𝜃, the train- ing set of semi-supervised learning consists of a labeled part X = {(𝑥𝑛,𝑦𝑛) : 𝑛 ∈ (1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', 𝑁)} and an unlabeled part U = {𝑢𝑛 : 𝑛 ∈ (1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', 𝜇𝑠𝑁)}, where 𝜇𝑠 is a hyperparameter that determines the relative sizes of X and U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Let F (𝑥;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃) be the predicted class distri- bution produced by the model F for input 𝑥.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For convenience, we always use Γ to represent the SSL algorithm, and the SSL process can be formalized as: arg min 𝜃 Γ(F (X ∪ U;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (1) The threat model can be formalized as the bilevel problem listed in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The outer optimization is to achieve two targets, one is the fundamental target of adversary backdoor poisoning: to maximize the attack success rate of backdoor examples with backdoor patterns without degrading model accuracy on unseen examples X𝑣𝑎𝑙, and the other is to ensure that backdoor patterns are the least perceptible to avoid arousing the suspicion of the victim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The inner optimization is that the victim uses the SSL algorithm Γ to train the model on the poisoned training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' min P𝑡𝑟 (·) E(𝑥,𝑦) ∈X𝑣𝑎𝑙 (ℓ(𝑦, F (𝑥;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃∗)) + ℓ(𝑦𝑡, F (P𝑣𝑎𝑙 (𝑥);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃∗)) +∥P𝑣𝑎𝑙 (𝑥) − 𝑥∥2) s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 𝜃∗ = arg min 𝜃 Γ(F (X ∪ (1 − 𝜇𝑏)U ∪ P𝑡𝑟 (𝜇𝑏U);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃)) , (2) where P𝑡𝑟 (𝜇𝑏U) indicates that the unlabeled data 𝜇𝑏U in the train- ing set is poisoned, and 𝜇𝑏 is a hyperparameter that determines the proportion of backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' P𝑣𝑎𝑙 (𝑥) is to add the backdoor pattern to the example 𝑥 to get the backdoor example.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 Achieving P𝑡𝑟 and P𝑣𝑎𝑙 The first thing to note is that P𝑣𝑎𝑙 (𝑥) and P𝑡𝑟 (𝑥) in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2 are differ- ent in CLB and DeNeB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In CLB, for the model to remember backdoor patterns well, interpolation or adversarial perturbation is used to keep the selected images away from their correct classification when adding backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In DeNeB, when adding backdoor patterns, it also makes the features and classifications of selected images close to the target class through adversarial perturbation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In contrast, P𝑣𝑎𝑙 (𝑥) and P𝑡𝑟 (𝑥) in our poisoning are the same, that is, poisoned unlabeled images are obtained by only adding backdoor patterns to clean unlabeled images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Our poisoning focus on how to craft backdoor patterns so that the SSL models trained on poi- soned images can remember them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In general, our poisoning can be divided into the following three steps: preparing the surrogate network and dataset, crafting backdoor patterns, and poisoning unlabeled images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Feature distribution of poisoned unlabeledexamplesat4oothepoch 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 1 2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4- 3 4 5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 - 6 8 9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0Feature distribution of poisoned unlabeledexamplesatothepoch 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0 1 2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 3 4 5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 6 7 8 9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0Feature distribution of poisoned unlabeled examples at 2ooth epoch 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0 2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 3 5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 8 9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0Trojaning semi-supervised learning model via poisoning wild images on the web Conference’17, July 2017, Washington, DC, USA ⋮ ⋮ ℒ𝑏 ℒ𝑡 ∇𝜃𝑠ℒ𝑏 ∇𝜃𝑠ℒ𝑡 ℒ𝑏𝑡 ℒ𝑖𝑛𝑣 Clean images Backdoor patterns Backdoor images Backward Backward Target class 𝑦𝑡 Target image 𝑥𝑡 Frozen surrogate network Backdoor pattern generator Clip 𝜀Tanh ⋱ ⋱ ⋱ : ship Figure 5: Crafting backdoor patterns on the surrogate network and the surrogate dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Preparing the surrogate network and dataset: Since only the target class 𝑦𝑡 and the distribution Z of the victim dataset are grasped by the adversary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' With the help of the transferability of neural networks, the adversary crafts backdoor patterns on the sur- rogate dataset X𝑠 and the surrogate network F 𝑠 parameterized by 𝜃𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' X𝑠 should contain the images for the target class𝑦𝑡 and conform to the distribution Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' F 𝑠 should ensure considerable classification accuracy on X𝑠, thus mining backdoor patterns that are as imper- ceptible as possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' F 𝑠 is then trained on X𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For simplicity, in the following, the learned parameters of F 𝑠 are still denoted by 𝜃𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Note that X𝑠 is not required to be labeled, which can be labeled-less or unlabeled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' This is because F 𝑠 can be trained by semi-supervised learning [29, 35] or unsupervised learning [5, 9], which is beyond our research scope.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Crafting backdoor patterns: This step crafts backdoor pat- terns based on the surrogate dataset X𝑠 and learned surrogate network F 𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As concluded in Section 3, to avoid the failure of back- door poisoning caused by the correct labeling of SSL algorithms, poisoning only is implemented on the images from the target class 𝑦𝑡.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, as listed in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2, the target of backdoor poisoning requires that the images from different classes are all misclassified as the target class 𝑦𝑡 by the backdoor model after adding backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, although only images from the target class are poi- soned, the images from other classes need to be taken into account when crafting the backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, this step includes the design of two aspects.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' One is to make backdoor patterns imperceptible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5, we use a backdoor pattern generator to generate the raw backdoor pattern.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Let G parameterized by 𝜗 denote this generator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The input is a clean image 𝑥𝑠 ∈ X𝑠, and the output is a raw backdoor pattern G(𝑥𝑠;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜗) corresponding to this image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' It is then constrained to a reasonable range using the activation function Tanh and multiplied by the budget 𝜖 to make the generator search for imperceptible backdoor patterns within the given budget.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Finally, the obtained backdoor pattern is added to the image, and the clip function is connected to make the backdoor image 𝑥𝑏 in the normal range.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To further ensure that the backdoor pattern is imperceptible, we add a loss function L𝑖𝑛𝑠 listed in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 3, which makes the backdoor image look more like the clean image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 𝐿𝑖𝑛𝑠 = E𝑥𝑠∼X𝑠 ���𝑥𝑏 − 𝑥𝑠��� 2 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ���𝑥𝑏 − 𝑥𝑠���∞ ≤ 𝜀 (3) The other is to make the backdoor image 𝑥𝑏 be learned by the trained-from-scratch SSL model F into the target class 𝑦𝑡.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To achieve this target, we propose a gradient matching strategy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' First, let’s see the learning of the target image 𝑥𝑡 from the target class 𝑦𝑡 by F .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Regardless of whether 𝑥𝑡 is labeled, whether pseudo-labeling or consistency regularization is employed, the target of learning 𝑥𝑡 is to make it classified into the target class, which can be formalized as: 𝜃𝑘 = 𝜃𝑘−1 − 𝜂∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑡;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘−1)) for 𝑘 ∈ [1,𝑚𝑎𝑥_𝑘] (4) where 𝜃𝑘 indicates the weights of the 𝑘th iteration, 𝜂 is the learning rate, 𝑚𝑎𝑥_𝑘 indicates the number of iterations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' When crafting the backdoor image𝑥𝑏, as listed in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5, we hope that the SSL algorithm learns them just like fitting the target image 𝑥𝑡 to the target class𝑦𝑡, so that the SSL algorithm can be tricked into injecting the backdoor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' This means that at each iteration, the gradients of the backdoor image 𝑥𝑏 on the model F should match the gradients of the target image 𝑥𝑡 on the model F .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 𝜃𝑘−1 − 𝜂∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑡;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘−1)) ≈ 𝜃𝑘−1 − 𝜂∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑏;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘−1)) → ∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑡;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘−1)) ≈ ∇𝜃𝑘−1ℓ(𝑦𝑡, F (𝑥𝑏;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘−1)) for 𝑘 ∈ [1,𝑚𝑎𝑥_𝑘] , (5) However, since our poisoning is zero-knowledge, the gradient infor- mation during model F training cannot be obtained.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To circumvent this problem, we think of mimicking such gradient information on the surrogate network F 𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Furthermore, gradient information of all iterations is not required.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Because on the one hand, the model F trained from scratch will generate numerous gradient information, and it is extremely costly and not practical to mimic all of this.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On the other hand, gradient information in early training does not carry meaningful information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, considering the computational cost, we only take the gradient information of the well-trained sur- rogate network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Experiments in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 7 have demonstrated that our approach is effective.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, first, from the target class, we select images that can be classified as the target class with high confidence by the surrogate network F 𝑠 as target images 𝑥𝑡, thereby ensuring that the gradients of the backdoor image 𝑥𝑏 can well match those of the images from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then feed the target image 𝑥𝑡 to the frozen well-trained F 𝑠 and get the loss: L𝑡 = ℓ(𝑦𝑡, F 𝑠 (𝑥𝑡;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑠)), (6) and calculate the gradient ∇𝜃𝑠 L𝑡 to the parameters 𝜃𝑠.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Likewise, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5, the backdoor images are also fed to the F 𝑠, and Conference’17, July 2017, Washington, DC, USA Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang Table 1: Poisoning performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' SL CA represents the model accuracy trained only on labeled examples by supervised learn- ing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' SSL CA represents the model accuracy trained on the complete training set (including labeled examples and unlabeled examples) by semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' CA indicates the accuracy of the poisoned SSL model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In the column IMP, the data from top to bottom are PSNR, SSIM, and L-∞ norm, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dataset SSL algorithm SL CA SSL CA BadNets-C [10] DeNeB-C [46] CLB [39] Ours CA ASR IMP CA ASR IMP CA ASR IMP CA ASR IMP CIFAR10 PseudoLabel [19] 78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='86 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='18 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='36 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8559 214.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='70 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='40 25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='01 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='55 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8211 208.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='93 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='18 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03 19.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='98 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8020 214.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='70 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='15 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='17 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9515 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 PI-Model [30] 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='38 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 48.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='32 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='56 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='05 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='24 MeanTeacher [38] 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='54 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='41 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='11 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='18 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='27 56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='43 88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='70 VAT [25] 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='33 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='29 59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='33 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='31 46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='93 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='90 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='13 ICT [41] 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 49.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='87 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='09 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='15 57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='84 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='54 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='54 FixMatch [35] 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='56 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='39 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='27 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='48 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='90 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='65 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='45 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='83 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 SVHN PseudoLabel [19] 86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='54 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='28 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='09 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='66 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='89 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8361 195.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='22 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='06 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='29 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='59 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='7825 195.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='22 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='53 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='7741 195.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='22 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='16 66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='31 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9832 19.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='62 PI-Model [30] 92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='19 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='69 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='58 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='36 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='19 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='24 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='83 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='78 MeanTeacher [38] 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='52 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='19 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='64 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='15 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='45 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='18 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='58 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='40 69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='76 VAT [25] 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='16 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='58 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='96 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='29 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='58 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='54 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='63 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='05 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='71 ICT [41] 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='62 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='98 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='41 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='27 FixMatch [35] 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='10 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='89 67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='95 54.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='63 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='79 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='47 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03 79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='59 the loss L𝑏 is obtained.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' L𝑏 = ℓ(𝑦𝑡, F 𝑠 (𝑥𝑏;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑠)), (7) and calculate the gradient ∇𝜃𝑠 L𝑏.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Finally, optimize G so that ∇𝜃𝑠 L𝑏 is close to ∇𝜃𝑠 L𝑡, as listed in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' L𝑏𝑡 = ∥∇𝜃𝑠 L𝑡 − ∇𝜃𝑠 L𝑏 ∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (8) The whole loss function for crafting backdoor patterns can be expressed as: L𝑐𝑟𝑎𝑓 𝑡 = arg min 𝜗 E𝑥𝑠∼X𝑠 L𝑏𝑡 + 𝜆𝑖𝑛𝑠L𝑖𝑛𝑠 s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ���𝑥𝑏 − 𝑥𝑠��� ≤ 𝜀 , (9) where 𝜆𝑖𝑛𝑠 is the hyperparameter that determines the imperceptibil- ity of backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We adopt a gradually increasing strategy for 𝜆𝑖𝑛𝑠, that is, multiply 𝜆𝑖𝑛𝑠 by 2 every 50 epochs, thus finding backdoor patterns that are as imperceptible as possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Poisoning unlabeled images: According to the poisoning ratio 𝜇𝑏, the images to be poisoned are selected from the unlabeled images from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then, feed them into the learned backdoor pattern generator G to get the corresponding backdoor patterns, and add them to the images to get the poisoned images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Finally, the poisoned images are posted on the Internet for the victim to scratch or secretly re-injected into the victim dataset U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5 EXPERIMENT EVALUATION 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 Experiment setup 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 Victim network and dataset: We implement our poisoning on CIFAR10 [16] and SVHN [26], which are widely used in semi- supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' CIFAR10 contains 50,000 training images and 10,000 testing images from 10 classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' SVHN consists of 73257 training images and 26032 testing images of house digits from 10 classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Moreover, CIFAR10 is trained on CNN13 [38] and SVHN is trained on WideResNet-28-2 [47].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To implement semi-supervised learning, in CIFAR10, only 4000 training images are labeled, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', 𝑁 = 4000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In SVHN, only 1000 images are labeled, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', 𝑁 = 1000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 SSL algorithms: We select some representative SSL algo- rithms from three types introduced in Section 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3: consistency regularization is PI-Model [30], MeanTeacher [38], VAT [25], and ICT [41], pseudo-labeling is PseudoLabel [19], pseudo-labeling with consistency regularization is FixMatch [35].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Some of these algo- rithms are old and do not perform well, while others are recently proposed and have outstanding performance, which can fully verify the generality of our poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The implementations of these SSL algorithms on SVHN and CIFAR10 come from the public Pytorch open source codes, [36] and [14], respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 Baseline poisonings: We compare our poisoning with BadNets- C [10], CLB [39], and DeNeB-C [46] , which have been described in detail in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Their backdoor patterns are all the same 8 × 8 pixel block located at the position (20, 20) of the image, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 6(b) 6(c), and 6(d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The target class is all 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 Poisoning setup: In our experiments, except that the sur- rogate dataset for victim dataset CIFAR10 is CIFAR10, and the surrogate dataset for victim dataset SVHN is SVHN, the other ex- perimental settings are the same.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The surrogate network and the backdoor pattern generator are WideResNet-28-2 [47] and UNet [32], respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The target class is 8, the number of poisons is 500, 𝜆𝑖𝑛𝑠 is 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='05, 𝜖 is 27 which indicates pixel perturbation maximum.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 Evaluation metrics: Evaluation metrics include three: the accuracy of clean examples (CA), the attack success rate (ASR) of the backdoor, and the imperceptibility (IMP) of backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' CA: Backdoor poisoning should not degrade the accuracy of the SSL model, that is, the CA on the poisoned model should be close to the CA on the unpoisoned model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ASR: In the inference stage, the probability that the examples with backdoor patterns added are misclassified by the poisoned model into the target class, higher ASR means better poisoning performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' IMP: The more imperceptible backdoor patterns are, the better they can evade the detection of the victim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We quantify imper- ceptibility by computing the distance between clean images and poisoned images by PSNR [13], SSIM [13], and L-∞ norm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The larger the PSNR, the closer the SSIM is to 1, and the smaller the L-∞ norm, the better the imperceptibility.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 Poisoning performance Let’s first verify that the poisoned unlabeled examples and gradient matching work in the trained-from-scratch SSL model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In the SSL Trojaning semi-supervised learning model via poisoning wild images on the web Conference’17, July 2017, Washington, DC, USA (a) Clean images (b) Poisoned images of BadNets (c) Poisoned images of CLB (d) Poisoned images of DeNeB (e) Our poisoned images Figure 6: Clean images and poisoned images from CIFAR10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' These images from SVHN are posted on the supplementary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' model of the 𝑘th epoch, the loss of poisoned examples taking the target label as the label is: D = ℓ(𝑦𝑡, F (P𝑡𝑟 (𝜇𝑏U);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘)) (10) To more accurately reflect that predictions of backdoor examples are far away from the clean classes and close to the target class, we adopt relative distance to define the degree C of gradient matching.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C = E(𝑥,𝑦) ∈X𝑣𝑎𝑙 ��∇𝜃𝑘 ℓ(𝑦𝑡, F (P𝑣𝑎𝑙 (𝑥);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘)) − ∇𝜃𝑘 ℓ(𝑦𝑡, F (𝑥𝑡;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘)) ��2 ��∇𝜃𝑘 ℓ(𝑦, F (P𝑣𝑎𝑙 (𝑥);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘)) − ∇𝜃𝑘 ℓ(𝑦𝑡, F (𝑥𝑡;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='𝜃𝑘)) ��2 (11) where the upper term calculates the gradient distance between the backdoor example P𝑣𝑎𝑙 (𝑥) taking the target label as the label and the target example 𝑥𝑡 taking the target label as the label.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The lower term calculates the gradient distance between the backdoor example taking the correct label as the label and the target example taking the target label as the label.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Since poisoned unlabeled examples are all from the target class, the SSL model will correctly classify them as the target class 𝑦𝑡, so D will gradually decrease, as shown in the bottom of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As a result, the gradient distance C between backdoor examples and target examples also gradually decreases, as shown in the middle of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' This brings about a gradual increase in ASR, as shown in the top of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Table 1, we present the final poisoning results, from which we can draw three conclusions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' First, SSL CAs are significantly higher than SL CAs, which means that the learning of SSL algorithms on unlabeled examples improves the model accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Moreover, when poisonings are implemented, the CAs of poisoned models do not show significant degradations compared to CAs of clean models, which means that unlabeled backdoor poisoning can be achieved without degenerating model accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Second, on these SSL algorithms, the ASRs of our poisoning are much higher than those of several baseline poisonings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As listed in Table 1, on CIFAR10, although these poisonings achieve certain ASRs, the highest is only 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='56%, while our lowest is 87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='13%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On SVHN, baseline poisonings fail on all SSL algorithms except Fix- Match which has the ASR of 75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='47%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In contrast, our poisoning can be applied to these SSL algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Although ASRs of our poi- soning are lower on SVHN than on CIFAR10, the imperceptibility of backdoor patterns is better.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' If the adversary is willing to sacrifice imperceptibility, it will bring an increase in ASRs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In addition, differ- ent SSL algorithms have different vulnerabilities to our poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On FixMatch, our poisoning performs the best, while on VAT, the ASR is the lowest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' This may be because VAT considers adversarial perturbations as image augmentations and implements adversarial (a) PseudoLabel (b) PI-Model (c) MeanTeacher (d) VAT (e) ICT (f) FixMatch Figure 7: The evolutions of C, D, and ASR with increasing epochs on trained-from-scratch SSL model CNN13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' training to ensure that unlabeled examples are resistant to these adversarial perturbations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The backdoor patterns we craft are simi- lar to adversarial perturbations, so adversarial training improves the model’s ability to resist poisoned examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Finally, thanks to the imperceptibility design of our poisoning, poisoned images look very similar to clean images, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In contrast, the poisoned images of baseline poisonings have obvious backdoor squares, which are easily detected by victims as suspicious.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Quantitatively, as listed in Table 1, our PNSR, SSIM and L-∞ norm all significantly outperform those of these schemes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' To sum up, our poisoning well achieves the poisoning target in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 Ablation study To focus on the impact of varying hyperparameters or situations on poisoning performance, the evaluations in this section are per- formed on CIFAR10 trained with FixMatch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Poisoning on other target classes are posted on the supplementary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 Evaluation across network architectures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We evaluate the im- pact of different architectures of generators and surrogate networks on the ASR and imperceptibility.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Alternative generators include SimNet and UNet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Alternative surrogate networks are LeNet [18], 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 R 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 8:8 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 D 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0 50 100 150 200 250 300 350 400 SSL epochs1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 S 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 D 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0 50 100 150 200 250 300 350 400 SSL epochs1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 ASF 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='50 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='20 D 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='05 MW 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0 50 100 150 200 250 300 350 400 SSL epochs1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 R 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 C 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='75 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0 50 100 150 200 250 300 350 400 SSL epochs1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 R 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 0 50 100 150 200 250 300 350 400 SSL epochs1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 R 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='125 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='100 D 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='075 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='050 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='025 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='000 0 50 100 150 200 250 300 350 400 SSL epochsConference’17, July 2017, Washington, DC, USA Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang Table 2: Evaluation across network architectures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' SimNet is a simple network we build that only consists of one convo- lutional layer (3 × 64) for down-sampling and one deconvo- lutional layer (64 × 3) for up-sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Generator Surrogate Nets CA ASR IMP PSNR SSIM L-∞ SimNet LeNet 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='72 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='33 21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='72 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6950 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 CNN13 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='73 90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='40 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='54 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8703 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='89 WideResNet-28-2 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='65 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='39 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='63 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8966 26.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='87 UNet LeNet 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='77 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='37 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8057 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 CNN13 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='86 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='24 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='43 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9423 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88 WideResNet-28-2 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='83 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9515 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 CNN13, and WideResNet-28-2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The poisoning results are listed in Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Comparing SimNet and UNet, it can be seen that because the network is too simple, it is more difficult for SimNet to ex- plore imperceptible backdoor patterns.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Although a higher ASR is obtained on LeNet, imperceptibility is greatly sacrificed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On CNN13 and WideResNet-28-2, UNet achieves higher ASR and better imper- ceptibility.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Comparing these surrogate networks, WideResNet-28-2 is larger in scale, better at exploring the least perceptible backdoor patterns, and obtaining higher ASR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 Evaluation across perturbation budgets 𝜖.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Although a higher perturbation budget 𝜖 leads to more significant changes in indi- vidual pixels, it allows us more space to search for imperceptible backdoor patterns and leads to higher ASRs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As shown in Table 3, at 𝜖 = 7, although the maximum value of pixel perturbation is only 7, the imperceptibility of backdoor patterns does not bring more significant improvement than at 𝜖 = 27, while the ASR significantly drops, from 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12% to 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='40%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' On the other hand, when 𝜖 = 54, the maximum value of pixel perturbation is improved to 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='32 com- pared to at 𝜖 = 27, while the ASR is only improved by 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Considering ASR and imperceptibility, 𝜖 = 27 is more suitable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (a) Poisoned images of CLB (b) Our poisoned images Figure 8: Grad-CAM [34] visualiztion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 Evaluation across other situations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In this section, we consider three situations of our poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' S1: Assume that the adversary does not know that the victim dataset is CIFAR10, but only knows the target class and dataset distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' He then employs CIFAR100, which has a similar dis- tribution to CIFAR10, as a surrogate dataset, and replace examples of a certain class with examples from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' S2: Assume that the adversary does not know the labeling sit- uation of the examples in the victim training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The poisoned unlabeled examples are then labeled proportionally.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' S3: Assume that the poisoned unlabeled examples come from different classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Table 3: Evaluation across perturbation budgets 𝜖 𝜖 SSL CA CA ASR IMP PSNR SSIM L-∞ 7 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='56 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='62 64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='40 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9502 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 27 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='83 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9515 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 54 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='78 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='63 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='57 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9497 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='32 Table 4: Evaluation across other situations Situation SSL CA CA ASR IMP PSNR SSIM L-∞ S1 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='56 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='69 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9389 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='59 S2 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='75 96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='89 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9515 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 S3 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='86 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='79 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9515 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 The poisoning results are listed in Table 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In the S1 situation, since the distributions of CIFAR100 and CIFAR10 are similar, back- door patterns crafted on CIFAR100 can be migrated to CIFAR10, obtaining an ASR of 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='25%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In the S2 situation, since we only poison examples from the target class, even if these examples are correctly labeled, it will not impact the ASR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The poisoning failure (the ASR is only 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='79%) in the S3 situation further validates the finding in Section 3: label-inconsistent backdoor poisoning is much more difficult to use for unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 Defense evaluation The section evaluates that our poisoning can bypass five represen- tative defenses from four types mentioned in Section 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2, includ- ing Activation Cluster [3], Neural Cleanse [42], Fine-pruning [22], STRIP [8], and DePuD [46].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The DePuD is posted below, and the other four are posted on the supplementary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 DePuD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DePuD is proposed to detect poisoned unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' First, all the training examples are divided into two cate- gories according to whether they are labeled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The labeled ones are assigned the label 0, and the unlabeled ones are assigned the label 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' These examples are then classified using a heavy regularization model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' If poisoned examples have significant backdoor patterns, the predictions will be extremely close to 1, and the separation from clean unlabeled examples will appear, so that it is detected as abnor- mal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DePuD works for CLB, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 8(a), backdoor patterns can be detected prominently in the lower right corner.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, the backdoor patterns of our poisoning are imperceptible, it is diffi- cult to be captured by the heavy regularization model, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 8(b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Moreover, as shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 9, clean unlabeled examples almost overlap with poisoned unlabeled examples, which means that DePuD cannot separate the poisoned unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (a) (b) Figure 9: DePuD detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DePuD detection on poisoned SVHN 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 Labeledtrainingimages Cleanunlabeledtrainingimages 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 Poisoned unlabeledtraining images Distribution 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 PredictionDePuD detectiononpoisonedCiFARlo 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 Labeledtrainingimages Cleanunlabeledtrainingimages 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 Poisoned unlabeledtraining images Distribution 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 PredictionTrojaning semi-supervised learning model via poisoning wild images on the web Conference’17, July 2017, Washington, DC, USA 6 CONCLUSION This paper is the first to investigate the vulnerability of unlabeled examples of trained-from-scratch SSL models to backdoor poison- ing, revealing the flaws in the security design of SSL algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We first find that label-inconsistent backdoor poisoning cannot be used for unlabeled examples due to the opposition to the SSL algo- rithms that strive to correctly learn unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, for unlabeled examples, poisoning only is implemented on examples from the target class.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Based on this, we propose a zero-knowledge and imperceptible backdoor poisoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Experiments show that our poisoning achieves state-of-the-art attack success rates when by- passing various defenses.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' REFERENCES [1] Berthelot, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Carlini, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Cubuk, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Kurakin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Sohn, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Raf- fel, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ReMixMatch: Semi-Supervised Learning with Distribution Matching and Augmentation Anchoring.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In International Conference on Learning Represen- tations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [2] Berthelot, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Carlini, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Goodfellow, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Papernot, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Oliver, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Raffel, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Mixmatch: A holistic approach to semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [3] Chen, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Carvalho, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Baracaldo, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ludwig, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Edwards, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Lee, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Molloy, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Srivastava, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Detecting backdoor attacks on deep neural networks by activation clustering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:1811.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03728.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [4] Chen, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Fu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Koushanfar, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DeepInspect: A Black-box Trojan Detection and Mitigation Framework for Deep Neural Networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In IJCAI, volume 2, 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [5] Chen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Kornblith, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Norouzi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Hinton, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A simple framework for contrastive learning of visual representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In International conference on machine learning, 1597–1607.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' PMLR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [6] Doan, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Lao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhao, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Li, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' LIRA: Learnable, Imperceptible and Robust Backdoor Attacks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF International Conference on Computer Vision, 11966–11976.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [7] Dong, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Yang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Deng, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Pang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Xiao, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Su, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Zhu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Black-box detection of backdoor attacks with limited information and data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF International Conference on Computer Vision, 16482–16491.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [8] Gao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Xu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ranasinghe, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Nepal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Strip: A defence against trojan attacks on deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the 35th Annual Computer Security Applications Conference, 113–125.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [9] Gidaris, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Singh, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Komodakis, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Unsupervised Representation Learning by Predicting Image Rotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [10] Gu, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dolan-Gavitt, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Garg, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Badnets: Identifying vulnerabilities in the machine learning model supply chain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:1708.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='06733.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [11] Guo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Liu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' AEVA: Black-box Backdoor Detection Using Adversarial Extreme Value Analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:2110.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='14880.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [12] He, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ren, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Sun, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Deep residual learning for image recognition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE conference on computer vision and pattern recognition, 770–778.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [13] Hore, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Ziou, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Image quality metrics: PSNR vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' SSIM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In 2010 20th international conference on pattern recognition, 2366–2369.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' IEEE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [14] iBelieveCJM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Tricks of Semi-supervised Deep Leanring–Pytorch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' https: //github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='com/iBelieveCJM/Tricks-of-Semi-supervisedDeepLeanring-Pytorch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [15] Iscen, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Tolias, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Avrithis, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Chum, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Label propagation for deep semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 5070–5079.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [16] Krizhevsky, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Hinton, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Learning multiple layers of features from tiny images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [17] Laine, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Aila, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Temporal ensembling for semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:1610.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='02242.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [18] LeCun, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bottou, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bengio, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Haffner, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Gradient-based learning applied to document recognition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Proceedings of the IEEE, 86(11): 2278–2324.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [19] Lee, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Workshop on challenges in repre- sentation learning, ICML, volume 3, 896.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [20] Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Tong, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Lim, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Zhou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' PointBA: Towards Backdoor Attacks in 3D Point Cloud.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF International Conference on Computer Vision, 16492–16501.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [21] Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wu, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' He, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Lyu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Invisible backdoor attack with sample-specific triggers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF International Conference on Computer Vision, 16463–16472.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [22] Liu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dolan-Gavitt, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Garg, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Fine-pruning: Defending against backdooring attacks on deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In International Symposium on Research in Attacks, Intrusions, and Defenses, 273–294.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [23] Liu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ma, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bailey, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Lu, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Reflection backdoor: A natural backdoor attack on deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In European Conference on Computer Vision, 182–199.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [24] Mahajan, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Girshick, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ramanathan, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' He, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Paluri, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bharambe, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Van Der Maaten, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Exploring the limits of weakly supervised pretraining.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the European conference on computer vision (ECCV), 181–196.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [25] Miyato, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Maeda, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Koyama, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Ishii, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Virtual adversarial training: a regularization method for supervised and semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' IEEE transactions on pattern analysis and machine intelligence, 41(8): 1979–1993.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [26] Netzer, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Coates, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bissacco, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wu, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Ng, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Reading digits in natural images with unsupervised feature learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [27] Nguyen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Tran, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Input-aware dynamic backdoor attack.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 33: 3454–3464.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [28] Nguyen, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Tran, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' WaNet-Imperceptible Warping-based Back- door Attack.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In International Conference on Learning Representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [29] Pham, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dai, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Xie, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Le, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Meta pseudo labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11557– 11568.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [30] Rasmus, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Berglund, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Honkala, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Valpola, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Raiko, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Semi- supervised learning with ladder networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Advances in neural information pro- cessing systems, 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [31] Redmon, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Divvala, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Girshick, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Farhadi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' You only look once: Unified, real-time object detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE conference on computer vision and pattern recognition, 779–788.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [32] Ronneberger, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Fischer, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Brox, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' U-net: Convolutional networks for biomedical image segmentation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In International Conference on Medical image computing and computer-assisted intervention, 234–241.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [33] Salem, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wen, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Backes, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ma, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Zhang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dynamic backdoor attacks against machine learning models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03675.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [34] Selvaraju, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Cogswell, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Das, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Vedantam, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Parikh, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Batra, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Grad-cam: Visual explanations from deep networks via gradient-based localization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE international conference on computer vision, 618–626.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [35] Sohn, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Berthelot, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Carlini, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Raffel, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Cubuk, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Kurakin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Fixmatch: Simplifying semi-supervised learning with consistency and confidence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 33: 596–608.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [36] Suzuki, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Consistency Regularization for Semi-supervised Learning with PyTorch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' https://github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='com/perrying/pytorch-consistency-regularization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [37] Szegedy, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ioffe, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Vanhoucke, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Alemi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Inception-v4, inception-resnet and the impact of residual connections on learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Thirty- first AAAI conference on artificial intelligence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [38] Tarvainen, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Valpola, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Advances in neural information processing systems, 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [39] Turner, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Tsipras, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Madry, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Label-consistent backdoor attacks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='02771.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [40] Van der Maaten, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Hinton, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Visualizing data using t-SNE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Journal of machine learning research, 9(11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [41] Verma, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Kawaguchi, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Lamb, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Kannala, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bengio, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Lopez-Paz, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Interpolation consistency training for semi-supervised learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:1903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='03825.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [42] Wang, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Yao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Shan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Viswanath, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zheng, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Zhao, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Neural cleanse: Identifying and mitigating backdoor attacks in neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In 2019 IEEE Symposium on Security and Privacy (SP), 707–723.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' IEEE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [43] Wang, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zhang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Chen, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Xiong, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Wang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Practical detection of trojan neural networks: Data-limited and data-free cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In European Conference on Computer Vision, 222–238.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Springer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [44] Xie, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Dai, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Hovy, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Luong, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Le, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Unsupervised data aug- mentation for consistency training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Advances in Neural Information Processing Systems, 33: 6256–6268.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [45] Yan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' TIan, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Poor, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' DeHiB: Deep hidden backdoor attack on semi-supervised learning via adversarial perturbation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, 10585– 10593.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [46] Yan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Li, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Guizani, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Deep Neural Backdoor in Semi-Supervised Learning: Threats and Countermeasures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' IEEE Transactions on Information Forensics and Security, 16: 4827–4842.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [47] Zagoruyko, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Komodakis, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Wide residual networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' arXiv preprint arXiv:1605.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='07146.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' [48] Zhao, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Ma, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zheng, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Bailey, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Chen, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Jiang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Clean-label backdoor attacks on video recognition models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 14443–14452.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Conference’17, July 2017, Washington, DC, USA Le Feng, Zhengxing Qian, Sheng Li, and Xinpeng Zhang [49] Zhao, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Zheng, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Xu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='-t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=';' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' and Wu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Object detection with deep learning: A review.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' IEEE transactions on neural networks and learning systems, 30(11): 3212–3232.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' APPENDIX A SVHN IMAGES In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 12, we show clean images and poisoned images on SVHN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' It can be seen that our poisoned images are visually very similar to the clean images, and the victim is difficult to detect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' B EVALUATION ACROSS TARGET CLASSES We select other classes to act as target classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The poisoning results are listed in Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' First, likewise, poisoning on other target classes does not degrade the model accuracy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Secondly, it can be seen that the poisoning difficulty of different target classes is different.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' For example, on the target class Bird, the ASR can reach 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='52%, while on Truck, the ASR is lower, 81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Of course, we also can sacrifice a little backdoor pattern imperceptibility to improve the ASR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C DEFENSE EVALUATION We evaluate our poisoning on Activation Cluster, Neural Cleanse, Fine-pruning, and STRIP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 Activation Cluster The process of Activation Cluster is to input all training examples into the already trained victim model, thereby obtaining the activa- tion of these examples in the last hidden layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' These activations are then divided into different clusters based on their labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Finally, it is determined whether there are poisoned examples by detecting the abnormality of these clusters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, our backdoor poison- ing does not rely on labels and poisons only unlabeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, poisoned unlabeled examples cannot be divided into different clusters based on labels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, our poisoning can naturally bypass Activation Cluster detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 Neural Cleanse On potentially poisoned models, reverse-engineer the minimum- intensity backdoor triggers for all classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' They then determine whether a certain class is the target class based on the prior knowl- edge that the target class injected into the backdoor has a trigger with abnormally small intensity, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', 𝑎𝑛𝑜𝑚𝑎𝑙𝑦 𝑖𝑛𝑑𝑒𝑥 > 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' In semi- supervised learning, the defender may not have enough labeled examples for more accurate reverse engineering, but we assume the most stringent condition that the defender has enough labeled examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' However, even so, as shown in the Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 10(a), the trigger intensity of the target class injected into the backdoor is not signifi- cant outliers, and 𝑎𝑛𝑜𝑚𝑎𝑙𝑦 𝑖𝑛𝑑𝑒𝑥 is all less than 2 (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 10(b)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, our poisoning can successfully bypass Neural Cleanse detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' We think this is because reverse engineering in Nerucal Cleanse detection relies on classification layers, whereas our poisoning is gradient matching that controls the entire network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 Fine-pruning This is a blind defense strategy, instead of detecting whether the model or example is poisoned, it uses pruning and fine-tuning for (a) (b) Figure 10: Neural Cleanse detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Table 5: Evaluation across target classes Target class SSL CA CA ASR IMP PSNR SSIM L-∞ Airplane 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='56 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='49 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='15 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='29 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9518 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='26 Automobile 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='44 89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='22 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='21 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9490 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88 Bird 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='76 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='52 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9492 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='99 Cat 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='95 99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='95 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='33 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9655 20.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='06 Deer 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='78 98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='18 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9560 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='18 Dog 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='81 91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='08 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='32 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9634 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='81 Frog 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='89 95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='64 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9519 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='22 Horse 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='84 94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='11 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='36 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9653 22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='91 Ship 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='83 97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='34 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9515 24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='51 Truck 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='93 81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='88 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='9765 23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='29 any model to try to eliminate possible backdoors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Specifically, clean examples are first fed into the model, and then 𝛼% (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=', Pruning rate) of neurons with minimal activation are dormant by pruning, thereby attempting to remove possible backdoors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Fine-tuning is then used to compensate for the degradation of clean example accuracy caused by pruning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 11, even with a pruning rate of 90%, there is no significant drop in the ASR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' When the pruning rate is 99%, on CIFAR10, the CA drops to 80%, and the ASR is still 62%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' An interesting phenomenon is that on SVHN, the ASR increases significantly, which may be because the excessive pruning makes the model’s ability to distinguish clean examples weakened, which makes it easier to be misclassified as the target class once backdoor patterns are added.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (a) (b) Figure 11: Fine-pruning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 STRIP STRIP is deployed in the model inference stage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Before the test- ing example is fed to the model, it is synthesized with a set of pre-prepared clean examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Then these obtained synthesized ex- amples are fed into the model for prediction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' If the entropy of their prediction results is abnormally small, it is determined that the testing example is a backdoor example, and the model is poisoned.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Trigger intensity of all class 100 90 80 70 60 Target class 50 Clean Poisoned Clean Poisoned SVHN SVHN CIFAR10 CIFAR10Neural cleanse detection result 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 Anomaly index 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='37 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='29 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='16 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='19 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 Clean Poisoned Clean Poisoned SVHN SVHN CIFAR10 CIFAR10Fine-pruning on SVHN 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 由 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 rate success 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 ASR 中 BMA 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='60 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='80 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='90 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='99 Pruning rateFine-pruning on CIFAR10 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 由 由 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 ASR 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 ASR 中 CA 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='30 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='400.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='60 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='70 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='80 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='90 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='99 Pruning rateTrojaning semi-supervised learning model via poisoning wild images on the web Conference’17, July 2017, Washington, DC, USA (a) Clean images (b) Poisoned images of BadNets (c) Poisoned images of CLB (d) Poisoned images of DeNeB (e) Our poisoned images Figure 12: Clean images and poisoned images on SVHN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' As shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' 13, we show the entropy distribution for 500 testing examples and 500 backdoor examples, and it can be seen that the distributions almost coincide.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' The entropy of the backdoor examples does not exhibit abnormally small property.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' Thus, our backdoor poisoning can bypass STRIP detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' (a) (b) Figure 13: STRIP detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content=' EntropydistributionofSVHN clean 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 backdoor 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='5 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6EntropydistributionofClFAR1o 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='12 clean backdoor 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='08 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='00 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='8 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='0 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='2 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
+page_content='4' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNAyT4oBgHgl3EQfkfir/content/2301.00435v1.pdf'}
diff --git a/jNFRT4oBgHgl3EQfWDfG/content/tmp_files/2301.13542v1.pdf.txt b/jNFRT4oBgHgl3EQfWDfG/content/tmp_files/2301.13542v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4594173c8e3c1fbbc47f18cb298d90e066a3ed4c
--- /dev/null
+++ b/jNFRT4oBgHgl3EQfWDfG/content/tmp_files/2301.13542v1.pdf.txt
@@ -0,0 +1,596 @@
+arXiv:2301.13542v1 [math.OC] 31 Jan 2023
+Springer Nature 2021 LATEX template
+Theoretical aspects in penalty
+hyperparameters optimization
+Flavia Esposito1,2*†, Laura Selicato1,2† and Caterina
+Sportelli1,3†
+1*Department of Mathematics, Universit`a degli Studi di Bari
+Aldo Moro, via Orabona, 4, Bari, 70125, Italy.
+2Member of INDAM research group, GNCS.
+3Member of INDAM research group, GNAMPA.
+*Corresponding author(s). E-mail(s): flavia.esposito@uniba.it;
+Contributing authors: laura.selicato@uniba.it;
+caterina.sportelli@uniba.it;
+†These authors contributed equally to this work.
+Abstract
+Learning processes are useful methodologies able to improve knowl-
+edge of real phenomena. These are often dependent on hyperparameters,
+variables set before the training process and regulating the learning
+procedure. Hyperparameters optimization problem is an open issue in
+learning approaches since it can strongly affect any real data analysis.
+They are usually selected using Grid-Search or Cross Validation tech-
+niques. No automatic tuning procedure exists especially if we focus on
+an unsupervised learning scenario.
+This study aims to assess some theoretical considerations for tun-
+ing penalty hyperparameters in optimization problems. It considers
+a bi-level formulation tuning problem in an unsupervised context,
+by using Gradient-based methods. Suitable conditions for the exis-
+tence of a minimizer in an infinite-dimensional Hilbert space are
+outlined, together with some theoretical results, applicable in all
+those situations when it is unnecessary or not possible obtaining
+an exact minimizer. An iterative algorithmic strategy is considered,
+equipped with a stopping criterion via Ekeland’s variational principle.
+Keywords: Hyperparameters optimization, learning approaches, existence
+results.
+1
+
+Springer Nature 2021 LATEX template
+2
+HPO: theoretical results
+MSC Classification: 68Q32 , 46N10 , 90C46 , 49J27 , 90C48
+Acknowledgment
+The authors would like to thank Prof. A. M. Candela and Prof. N. Del Buono
+from Universit`a degli Studi di Bari Aldo Moro for the deep discussions on the
+preliminary version of this manuscript. This work was supported by INDAM-
+GNCS.
+1 Introduction
+Training a Machine Learning (ML) algorithm is quite important to produce
+data-driven models, which can be successfully applied in real life applica-
+tions. These processes often require to specify several variables by the users,
+namely hyperparameters, which must be set before the learning procedure
+starts. Hyperparameters govern the whole learning process and play a cru-
+cial role in guaranteeing good model performances. They are often manually
+specified, and the lack of an automatic tuning procedure makes the field of
+Hyperparameter Optimization (HPO) an ever-evolving topic. The literature
+offers various solutions for hyperparameters tuning, from Gradient-based to
+Black-Box or Bayesian’s approaches, beside some naive but daily used meth-
+ods such as Grid and Random search. A brief overview on existing methods
+can be found in [1]. Hyperparameters can be of different types (discrete, con-
+tinuous, categorical), and in most cases, the number of their configurations to
+explore is infinite. This paves the way for a mathematical formalization of the
+HPO in ML context with abstract spaces, such as Hilbert spaces.
+A supervised learning algorithm may be represented as a mapping that
+takes a configuration of hyperparameter and a dataset D and returns an
+hypothesis [2]:
+A : Λ × D → H;
+A(λ, D) = h,
+with
+D =
+�
+N∈N
+(X × Y )N
+(1)
+where D is the space of finite dimensional dataset, representing a task, X and
+Y are the input and output spaces, Λ is an hyperparameter space, and H is an
+hypothesis space. A quite standard claim for the hypotheses set is to be a linear
+function space, endowed with a suitable norm (more binding arising from an
+inner product): two requirements satisfied when H is a Hilbert space of func-
+tions over X1. Assuming an Hilbert space structure on the hypothesis space
+has some advantages: (i) practical computations reduced to ordinary linear
+algebra operations and (ii) self duality; that is for any x ∈ X a representative
+1If X is an infinite dimensional space the boundedness is needed, too.
+
+Springer Nature 2021 LATEX template
+HPO: theoretical results
+3
+of x can be found, i.e., kx ∈ H exists such that
+h(x) = ⟨kx, h⟩
+for all h ∈ H,
+(2)
+where kx is a suitable positive definite “kernel”. This construction gives the
+chance to connect the abstract structure of H and what its elements actually
+are, flipping the construction of the hypotheses set from the kernel. Providing
+a suitable positive function k on X, H can be set as the minimal complete
+space of functions involving all {kx}x∈X equipped with the scalar product in
+(2). Thus, H is outlined in a unique way, and it is named the Representing
+Kernel Hilbert Space mapped to the kernel k.
+Starting from this abstract scenario, one can deepen the HPO in supervised
+ML. Formally, HPO can be formulated as the problem of minimizing the dis-
+crepancy between A, trained on a given training dataset Dtr, and a validation
+dataset Dval [3], to find the optimal λ∗ such that
+λ∗ = argmin
+λ∈Λ
+V (A(λ, Dtr), Dval),
+where
+V : H × X → R.
+(3)
+In this study, we will address problem (3) through Gradient-based meth-
+ods (GB), by using a bi-level approach. Bi-level programming solves an outer
+optimization problem subject to the optimality of an inner optimization prob-
+lem, and it can be adopted to formalize HPO for any learning algorithm [4–7].
+We will work on Hilbert spaces for solving HPO in unsupervised problems,
+considering as hyperparameter the penalty coefficient. We already treat this
+aspect in the particular and more specific case of Nonnegative Matrix Factor-
+ization task, encurring in some generalization problems and restrictions of the
+theorems’ assumptions [8].
+To overcome the difficulties in ensuring the theoretical assumptions when
+real data domains are considered, this work extends existence and uniqueness
+theorems for the solution of the hyperparameters bi-level problem to the more
+general framework of infinite dimensional Hilbert space. This latter also allows
+the application of the Ekeland’s variational principle to state that whenever a
+functional is not guaranteed to have a minimum, under suitable assumptions, a
+“good” substitute can be found, namely the best one can get as an approximate
+minimum. One of the purposes of this paper is to use this theoretical tool as a
+stopping criterion for the update of the hyperparameters as we will see later.
+The outline of the paper is as follows. Section 2 introduces the classical
+bi-level formalization of HPO and some preliminary notions in a supervised
+context. Section 3 illustrates our proposal, extension on the unsupervised con-
+text. A general framework addressing HPO in Hilbert space is also set, and
+some general abstract tools are stated in Section 4. Section 5 presents a critical
+discussion and some practical considerations. Finally, Section 6 summarizes
+the obtained results and draws some conclusions.
+
+Springer Nature 2021 LATEX template
+4
+HPO: theoretical results
+2 Previous works and preliminaries
+As briefly mentioned in the introduction, in a supervised learning scenario,
+HPO can be addressed through a bi-level formulation. This approach looks for
+the hyperparameters λ such that the minimization of the regularized training
+leads to the best performance of the trained data-driven model on a validation
+set. Accordingly to the ideas introduced in [9, 10], the best hyperparameters
+for a data learning task can be selected as the solution of the following problem:
+min{J(λ) : λ ∈ Λ},
+(4)
+J(λ) = inf{E(wλ, λ) : wλ ∈ argmin
+u∈Rr
+Lλ(u)},
+(5)
+where w ∈ Rr are r parameters, J : Λ → R is the so-called Response Function
+of the outer problem E : Rr × Λ → R, and for every λ ∈ Λ ⊂ Rp, Lλ : Rr → R
+is the inner problem.
+A reformulation of HPO as a bi-level optimization problem is also solved via
+some GB algorithms. In particular, in GB methods HPO is addressed with
+classical procedure for continuous optimization, in which the hyperparameter
+update is given by
+λt+1 = λt − αht(λ)
+(6)
+where ht is an approximation of the gradient of function J and α is a step
+size. It is known that the main challenge in this context is the computation
+of ht, called hypergradient. In several cases, a numerical approximation of the
+hypergradient can be calculated for real-valued hyperparameters, although few
+learning algorithms are differentiable in the classical sense.
+There are two main strategies for computing the hypergradient: iterative dif-
+ferentiation [9, 11, 12] and implicit differentiation [13, 14]. The former requires
+calculating the exact gradient of an approximate objective. This is defined
+through the recursive application of an optimization dynamics that aims to
+replace and approximate the learning algorithm A; the latter involves the
+numerical application of the implicit function theorem to the solution mapping
+A(Dtr; ·), when it is expressible through an appropriate equation [2].
+In this study, we follow the iterative strategy, so that problem in (4)-(5)
+can be addressed through a dynamical system type approach.
+If the following hypothesis hold:
+Hypothesis 1
+1. the set Λ is a compact subset of R;
+2. the Error Function E : Rr × Λ → R is jointly continuous;
+3. the map (w, λ) → Lλ(w) is jointly continuous, and then the problem
+argmin Lλ is a singleton for every λ ∈ Λ;
+4. the problem wλ = argmin Lλ remains bounded as λ varies in Λ;
+
+Springer Nature 2021 LATEX template
+HPO: theoretical results
+5
+the problem in (4)-(5) becomes:
+min
+λ∈Λ J(λ) = E(w∗
+λ, λ),
+w∗
+λ = argmin
+u
+Lλ(u).
+(7)
+It can be proved that the optimal solution (wλ∗, λ∗) of problem (7) exists [11].
+The goal of HPO is to minimize the validation error of model gw : X → Y ,
+parameterized by a vector w ∈ Rr, with respect to hyperparameters λ.
+Considering the penalty optimization problems in which hyperparameter is
+the penalty coefficient λ ∈ R+, the Inner Problem is the penalized empirical
+error represented by L, defined as:
+Lλ(w) =
+�
+(x,y)∈Dtr
+ℓ(gw(x), y) + λr(w),
+(8)
+where ℓ is the loss function, Dtr = {(xi, yi)}n
+i=1 the training set, and r : Rr → R
+is a penalty function. While the Outer Problem is the generalized error of gw
+represented by E:
+E(w, λ) =
+�
+(x,y)∈Dval
+ℓ(gw(x), y),
+(9)
+where Dval = {(xi, yi)}n
+i=1 is the validation set. Note that E does not explicitly
+depend on λ.
+This work will allow overcoming some assumptions of Hypothesis 1 (such
+as compactness) that are difficult to satisfy in real data learning contexts, and
+also to use some theoretical result as the Ekeland’s variational principle, stated
+in the following, to improve iterative algorithms.
+Theorem 1 (Ekeland’s variational principle) [15] Let (V, d) be a complete metric
+space and J : V → ¯R be a lower semi-continuous function which is bounded from
+below. Suppose that ε > 0 and ˜v ∈ V exist such that
+J(˜v) ≤ inf
+V J + ε.
+Then, given any ρ > 0, vρ ∈ V exists such that
+J(vρ) ≤ J(˜v),
+d(vρ, ˜v) ≤ ε
+ρ,
+and
+J(vρ) < J(v) + ρ d(vρ, v)
+∀ v ̸= vρ.
+3 Our Proposal
+The bi-level HPO framework can be modified to include unsupervised learning
+paradigms, generally designed to detect some useful latent structure embedded
+in data. Tuning hyperparameters for unsupervised learning models is more
+complex than the supervised case due to the lack of the output space, which
+defines the ground truth collected in the validation set.
+
+Springer Nature 2021 LATEX template
+6
+HPO: theoretical results
+This section describes a general framework to address HPO in Hilbert
+spaces for the unsupervised case and a corollary of the Ekeland’s variational
+principle used to derive a useful stopping criterion for iterative algorithms
+solving this HPO.
+Let X ∈ Rn×m be a data matrix, with reference to the problem (4)-(5), where
+now J : Λ → R is a suitable functional and Λ a Hilbert space equipped with
+the scalar product (·, ·), the outer problem is:
+E : Rr × Λ → R
+E(w, λ) =
+�
+x∈X
+ℓ(gw(x)),
+(10)
+and for every λ ∈ Λ the inner problem is:
+L : Rr → R
+Lλ(w) =
+�
+x∈X
+ℓ(gw(x)) + R(λ, w),
+(11)
+where R : Λ × Rr → R is a penalty function. We want to emphasize the
+new formulation with respect to (8) regarding the function Lλ, in which each
+component of the parameter w is penalized independently, and all optimization
+is performed on the data matrix X.
+The bi-level problem associated to (10)-(11) can be solved with a dynamical
+system approach in which the hypergradient is computed. Once the hypergra-
+dient is achieved a gradient-based approach can be used to find the optimum
+λ∗. The Ekeland’s variational principle can be used to construct an appropriate
+stopping criterion for iterative algorithms, with the aim of justifying and set-
+ting the hyperparameters related to the stopping criterion more appropriately.
+Roughly speaking, this variational principle asserts that, under assumptions of
+lower semi-continuity and boundedness from below, if a point ˜λ is an “almost
+minimum point” for a function J, hence a small perturbation of J exists which
+attains its minimum at a point “near” to ˜λ. As a fruitful selection of ρ occurs
+when ρ = √ε and such a choice allows us to reduce the number of hyperparam-
+eters to the precision error only, thus we will use Theorem 1 in the following
+form.
+Corollary 1 Let (V, d) be a complete metric space and J : Λ → ¯R be a lower semi-
+continuous function which is bounded from below. Suppose that ε > 0 and ˜λ ∈ Λ exist
+such that
+J(˜λ) ≤ inf
+Λ J + ε.
+Then, ˜z ∈ Λ exists such that
+J(˜z) ≤ J(˜λ),
+d(˜z, ˜λ) ≤ √ε.
+and
+J(˜z) < J(λ) + √ε d(˜z, λ)
+∀ λ ̸= ˜z.
+
+Springer Nature 2021 LATEX template
+HPO: theoretical results
+7
+4 Main Abstract results
+In this section, we are ready to weaken the assumptions we discussed earlier
+and provide results related to the use of the Ekeland’s principle as a stopping
+criterion. We mention an abstract result of the existence of a minimizer in
+Hilbert spaces which has great importance and a wide range of applications in
+several fields. As just one example, the Riesz’s Representation Theorem, even
+if implicitly, makes use of the existence of a minimizer [16]. This is a widely rel-
+evant issue about Hilbert spaces, which makes them nicer than Banach spaces
+or other topological vector spaces. One can think, for example, that the whole
+Dirac Bra–ket formalism of quantum mechanics relies on this identification.
+4.1 Abstract Existence Theorem
+It is well known that each bounded sequence in a normed space Λ has a norm
+convergent subsequence if and only if it is a finite dimensional normed space.
+Thus, given a normed space Λ, as the strong topology (i.e., the one induced
+by the norm) is too strong to provide any widely appropriate subsequential
+extraction procedure, one can consider other weak topologies joined with the
+linear structure of the space and look for subsequential extraction processes
+therein.
+In Banach spaces, as well as in Hilbert spaces, the two most relevant weaker-
+than-norm topologies are the weak-star topology and the weak topology. As
+the former is established in dual spaces, the latter is set up in every normed
+space. The notions of these topologies are not self-contained but fulfill a leading
+role in many features of the Banach space theory. In this regard, here we state
+some results we will use shortly.
+Theorem 2 If Λ is a finite-dimensional space, the strong and weak topologies coin-
+cide. In particular, it follows that the weak topology is normable, and then clearly
+metrizable, too.
+If Λ is an infinite-dimensional space, the weak topology is strictly contained in the
+strong topology, namely open sets for the strong topology exist which are not open for
+the weak topology. Furthermore, the weak topology turns to be not metrizable in this
+case.
+Definition 1 A functional J : Λ → ¯R with Λ topological space, is said to be lower
+semi-continuous on Λ if for each a ∈ R, the sublevel sets
+J−1(] − ∞, a]) = {λ ∈ Λ : J(λ) ≤ a}
+are closed subsets of Λ.
+In the following we introduce a “generalized Weierstrass Theorem” which
+gives a criteria for the existence of a minimum for a functional defined on a
+Hilbert space. For this reason, the incoming results will be provided for the
+abstract framework of a Hilbert space although, in some cases, they apply in
+
+Springer Nature 2021 LATEX template
+8
+HPO: theoretical results
+the more general context of Banach spaces. Thus, throughout the remaining
+part of this section we denote by Λ any real infinite dimensional Hilbert space.
+In an infinite dimensional setting, the following definitions are strictly related
+to the different notions of weak and strong topology.
+Definition 2 A functional J : Λ → ¯R is said to be strongly (weakly, respectively)
+lower semi-continuous if J is lower semi-continuous when Λ is equipped with the
+strong (weak, respectively) topology.
+Definition 3 A functional J : Λ → ¯R is said to be strongly (weakly, respectively)
+sequentially lower semi-continuous if
+lim inf
+n→+∞ J(λn) ≥ J(λ)
+for any sequence (λn)n ⊂ Λ such that λn → λ (λn ⇀ λ, respectively).
+We proceed by providing some useful results.
+Proposition 3 The following statements are equivalent:
+i) J : Λ → R is sequentially weakly lower semi-continuous functional;
+ii) the epigraph of J is weakly sequentially closed, where, by definition, it is
+epi(J) = {(λ, t) ∈ dom(J) × R : J(λ) ≤ t}.
+Remark 1 As a further consequence of the preliminary Theorem 2, we have that
+sequential weak lower semi-continuity and weak lower semi-continuity do not match if
+Λ is infinite dimensional since weak topology is not metrizable. However, the weaker
+concept of sequential weak lower semi-continuity meets our needs.
+Proposition 4 Let C ⊆ Λ be a closed and convex subset. Then, C is weakly
+sequentially closed, too.
+Since a sequentially weakly closed set is also strongly closed, it follows that
+a sequentially weakly lower semi-continuous functional is also (strongly) lower
+semi-continuous. Instead, the converse holds under an additional assumption.
+In particular, Proposition 4 allows us to infer the following results.
+Proposition 5 If J : Λ → R is a strongly lower semi-continuous convex functional;
+thus J is weakly sequentially lower semi-continuous, too.
+Proof Since J is lower semi-continuous, thus epi(J) is closed. On the other hand,
+since J is convex, so it is epi(J), whence Proposition 4 ensures that epi(J) is weakly
+sequentially closed, i.e., J is weakly sequentially lower semi-continuous.
+□
+
+Springer Nature 2021 LATEX template
+HPO: theoretical results
+9
+Thus, we are able to state the main result of this section.
+Theorem 6 Let C ⊂ Λ be a non-empty, closed, bounded and convex subset. Let
+J : Λ → R be a lower semi-continuous and convex functional. Thus J achieves its
+minimum in C, i.e., ¯λ ∈ C exists such that J(¯λ) = inf
+λ∈C J(λ).
+Proof Let m := inf
+λ∈C J(λ); hence (λn)n ⊂ C exists such that
+J(λn) → m
+as n → +∞.
+(12)
+Now, our boundness assumption on C implies that, up to subsequences, λ ∈ H exists
+such that λn ⇀ λ as n → +∞. Actually, since C is a closed and convex subset of Λ,
+thus Proposition 4 applies, which guarantees that λ ∈ C.
+Finally, from (12), Proposition 5 and Definition 3 we infer that J(¯λ) ≤ m, which
+gives the desired result.
+□
+Remark 2 If every closed and bounded subset in a metric space is compact, the
+space is said to have the Heine–Borel property. This property holds in every finite
+dimensional normed space but, in general, may not be true.
+Remark 3 We observe that Theorem 6 still holds if the subset C is not bounded as
+long as we ask for an additional assumption on the functional J. In fact, requiring
+J to be coercive2 (and if at least ¯λ ∈ C exists such that J(¯λ) < +∞), then any
+minimizer of J on C necessarily lies in some closed ball of radius r > 0. In fact, since
+J(¯λ) < +∞, any minimizer λ of J must have J(λ) ≤ J(¯λ); furthermore, since J is
+coercive, a sufficient large radius r > 0 exists such that J(λ) > J(¯λ) for all λ ∈ C
+with ∥λ∥ > r. Thus, any minimizer, if exists, lies in the ball {λ ∈ C : ∥λ∥ ≤ r}.
+In particular, Theorem 6 applies to the intersection between C and a closed ball of
+suitable radius, since it turns to be convex if we formally require C to be closed and
+convex.
+Namely, the following result holds.
+Corollary 2 Let C ⊂ Λ be a non-empty, closed and convex subset. Let J : Λ →
+R be a lower semi-continuous, convex and coercive functional. Thus J achieves its
+minimum, i.e., ¯λ ∈ C exists such that J(¯λ) = inf
+λ∈C J(λ).
+Now we introduce a couple of results which are a direct consequence of
+Ekeland’s variational principle. For the sake of completeness, here we provide
+them with all the details (see [17] for the original statements). Let Λ be a
+complete metric space and J : Λ → R be the lower semicontinuous response
+function on Λ. Supposte that a point λ ∈ Λ exists such that J(λ) < +∞.
+Thus, the following results hold.
+2We say that a functional J : H → R is coercive if J(u) → ∞ as ∥u∥ → ∞, u ∈ H.
+
+Springer Nature 2021 LATEX template
+10
+HPO: theoretical results
+Theorem 7 (Perturbation Result) Let Jλ : Λ → ¯R be a lower semicontinuous
+function such that the inequality
+|Jλ(γ) − J(γ)| ≤ ζ(d(γ, λ))
+holds
+∀γ ∈ Λ,
+(13)
+where Jλ(·) denote model function3, ζ is some growth function4, and let λ+ be a
+minimizers of Jλ. If λ+ coincides with λ, then |∇J(λ)| = 0. On the other hand, if λ
+and λ+ are distinct, then a point ˆλ ∈ X exists which satisfies
+1. d(λ+, ˆλ) ≤ 2 · ζ(d(λ+,λ))
+ζ′(d(λ+,λ))
+(point proximity)
+2. J(ˆλ) ≤ J(λ+) + ζ(d(λ+, λ))
+(value proximity).
+Proof By Taylor’s theorem it is simple to verify that |∇Jλ|(λ) = |∇J|(λ).
+Now, since λ is a minimizer, we have |∇J(λ)| = 0 if λ+ = λ. On the other hand, if
+λ+ ̸= λ, from inequality (13) and the definition of λ+, it follows that
+J(γ) ≥ Jλ(λ+) − ζ(d(γ, λ)).
+Let us define the new function
+G(γ) := J(γ) + ζ(d(γ, λ)).
+Thus, from assumption (13) and inequality inf G ≥ Jλ(λ+) we infer that
+G(λ+) − inf G ≤ J(λ+) − Jλ(λ+) + ζ(d(λ+, λ)) ≤ 2ζ(d(λ+, λ)).
+Whence, Theorem 1 applies and, having ε := 2ζ(d(λ+, λ)), for all ρ > 0 λρ exists
+such that
+G(λρ) ≤ G(λ+)
+and
+d(λ+, λρ) ≤ ε
+ρ.
+The desired result follows simply by placing ρ = ζ′(d(λ+, λ)) with ˆλ = λρ.
+□
+An immediate consequence of Theorem 7 is the following subsequence
+convergence result.
+Corollary 3 (Subsequence convergence to stationary points) Consider a sequence
+of points λk and closed functions Jλk : Λ → ¯R satisfying λk+1 = argmin
+γ
+Jλk(γ) and
+d(λk+1, λk) → 0. Moreover suppose that the inequality
+|Jλk(γ) − J(γ)| ≤ ζ(d(λk, γ))
+holds
+∀k ∈ N
+and
+γ ∈ Λ,
+(14)
+where ζ is a proper growth function. If (λ∗, J(λ∗)) is a limit point of the sequence
+(λk, J(λk)), then λ∗ is stationary for J.
+Two interesting consequences for convergence analysis flow from there. Sup-
+pose that the models are chosen in such a way that the step-sizes ∥λk+1 − λk∥
+tend to zero. This assumption is often enforced by ensuring that J(λk+1) <
+J(λk) by at least a multiple of ∥λk+1 − λk∥2 (sufficient decrease condition).
+3As model function we mean the Taylor’s expansion of J in λ, stopped to the first order.
+4A differentiable univariate function ζ : R+ → R+ is called a growth function if it satisfies
+ζ(0) = ζ′(0) = 0 and ζ′ > 0 on (0, +∞). If in addition, equalities lim
+t→0 ζ′(t) = lim
+t→0 ζ(t)/ζ′(t) = 0
+hold, we say that ζ is a proper growth function.
+
+Springer Nature 2021 LATEX template
+HPO: theoretical results
+11
+Then, assuming for simplicity that J is continuous on its domain, any limit
+point λ∗ of the iterate sequence λk will be stationary for the problem (Corol-
+lary 3).
+Thus, by choosing an error ε, we can stop update (6) for GB algorithms in the
+context of bi-level HPO for penalty hyperparameter, according to the pseudo
+code in 1.
+Algorithm 1 Pseudo-code
+Require: Error ε. Some starting points λ0, λ1.
+Ensure: Optimum λ∗
+1: while ∥λt − λt−1∥ > ε do
+2:
+Compute h(λ);
+3:
+update λt+1 = λt − αht(λt);
+4:
+t+ = 1.
+5: end while
+5 Discussion and practical considerations
+We want to emphasize that moving to infinite dimensional Hilbert spaces are
+not a mere abstract pretense, but it is also important in some application
+contexts. For example, when Support Vector Machine (SVM) are taken into
+consideration, a well known “kernel trick” permits to interpret a Gaussian
+kernel as an inner product in a feature space. This is potentially infinite-
+dimensional, allowing to read the SVM classifier function as a linear function
+in the feature space [18]. Another example is provided by the quantum system
+possible states problem, in which the state of a free particle can be described
+as vectors residing in a complex separable Hilbert space [19].
+Indeed, the strength of this article lies in theory. Both the existence theorem
+and the stopping criterion allow us to build an approach based on solid
+mathematical foundations useful for future extensions and generalizations to
+other problems, too. For example, infinite-dimensional Covariance Descriptors
+(CovDs) for classification is a fertile application arena for the extensions devel-
+oped here. This finds motivation in the fact that CovDs could be mapped
+to Reproducing Kernel Hilbert Space (RKHS) via the use of SPD-specific
+kernels [20].
+6 Conclusions
+In this paper, we studied the task of penalty HPO and we provided a math-
+ematical formulation, based on Hilbert spaces, to address this issue in an
+unsupervised context.
+Focusing on bi-level formulation, we showed some relaxed theoretical results
+both to weaken the hypotheses necessary for the existence of the solution.
+
+Springer Nature 2021 LATEX template
+12
+HPO: theoretical results
+Our approach differs from the more standard techniques in reducing the
+random or black box strategies giving stronger mathematical generalization
+suitable also when it is not possible obtaining exact minimizer.
+We also propose to use the Ekeland’s principle as a stopping criterion, which
+fits well in the context of GB methods.
+Declarations
+• Funding: The author F. E. was funded by REFIN Project, grant number
+363BB1F4, Reference project idea UNIBA027 “Un modello numerico-
+matematico basato su metodologie di algebra lineare e multilineare per
+l’analisi di dati genomici”. The author C. S. was partially supported by
+PRIN project “Qualitative and quantitative aspects of nonlinear PDEs”
+(2017JPCAPN 005) funded by Ministero dell’Istruzione, dell’Universit`a e
+della Ricerca.
+• Conflict of interest: The authors have no relevant financial or non-
+financial interests to disclose.
+• Data availability: Data sharing not applicable to this article as no datasets
+were generated or analysed during the current study.
+References
+[1] Del Buono, N., Esposito, F., Selicato, L.: Methods for hyperparame-
+ters optimization in learning approaches: An overview. In: International
+Conference on Machine Learning, Optimization, and Data Science, pp.
+100–112 (2020). Springer
+[2] Franceschi, L.: A unified framework for gradient-based hyperparameter
+optimization and meta-learning. PhD thesis, UCL (University College
+London) (2021)
+[3] Feurer, M., Hutter, F.: Hyperparameter Optimization, pp. 3–33. Springer,
+USA (2019)
+[4] Bard, J.F.: Practical Bilevel Optimization: Algorithms and Applications
+vol. 30. Springer, Boston (2013)
+[5] Colson, B., Marcotte, P., Savard, G.: An overview of bilevel optimization.
+Annals of operations research 153(1), 235–256 (2007)
+[6] Bertrand, Q., Klopfenstein, Q., Blondel, M., Vaiter, S., Gramfort, A.,
+Salmon, J.: Implicit differentiation of lasso-type models for hyperparam-
+eter optimization. In: International Conference on Machine Learning, pp.
+810–821 (2020). PMLR
+[7] Franceschi, L.: A unified framework for gradient-based hyperparameter
+optimization and meta-learning. PhD thesis, University College London
+
+Springer Nature 2021 LATEX template
+HPO: theoretical results
+13
+(2021)
+[8] Del Buono, N., Esposito, F., Selicato, L., Zdunek, R.: Optimizing Penal-
+ization Hyperparameters in nonnegative matrix factorizations problems.
+Preprint at http://arxiv.org/abs/2203.13129 (2022)
+[9] Franceschi, L., Donini, M., Frasconi, P., Pontil, M.: Forward and reverse
+gradient-based hyperparameter optimization. In: International Confer-
+ence on Machine Learning, pp. 1165–1173 (2017). PMLR
+[10] Vincent, D., Gelly, S., Le Roux, N., Bousquet, O.: Online hyper-parameter
+optimization (2018)
+[11] Franceschi, L., Frasconi, P., Salzo, S., Grazzi, R., Pontil, M.: Bilevel
+programming for hyperparameter optimization and meta-learning. In:
+International Conference on Machine Learning, pp. 1568–1577 (2018).
+PMLR
+[12] Maclaurin, D., Duvenaud, D., Adams, R.: Gradient-based hyperparameter
+optimization through reversible learning. In: Proc. of ICML, pp. 2113–
+2122 (2015)
+[13] Pedregosa, F.: Hyperparameter optimization with approximate gradient.
+In: ICML, pp. 737–746 (2016). PMLR
+[14] Lorraine, J., Vicol, P., Duvenaud, D.: Optimizing millions of hyperparam-
+eters by implicit differentiation. In: International Conference on Artificial
+Intelligence and Statistics, pp. 1540–1552 (2020). PMLR
+[15] Ekeland, I.: On the variational principle. Journal of Mathematical Anal-
+ysis and Applications 47(2), 324–353 (1974)
+[16] Rudin, W.: Real and Complex Analysis, (1987)
+[17] Drusvyatskiy, D., Ioffe, A.D., Lewis, A.S.: Nonsmooth optimization using
+taylor-like models: error bounds, convergence, and termination criteria.
+Mathematical Programming, 1–27 (2019)
+[18] Rossi, F., Villa, N.: Classification in hilbert spaces with support vector
+machines. Proceedings of ASMDA, 635–642 (2005)
+[19] Ying, M.: Foundations of Quantum Programming, (2016)
+[20] Harandi, M., Salzmann, M., Porikli, F.: Bregman divergences for infinite
+dimensional covariance matrices. In: Proceedings of the IEEE Conference
+on Computer Vision and Pattern Recognition, pp. 1003–1010 (2014)
+
diff --git a/jNFRT4oBgHgl3EQfWDfG/content/tmp_files/load_file.txt b/jNFRT4oBgHgl3EQfWDfG/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..541e6a5260cba0766e2ea7b7bb5bd56388fac325
--- /dev/null
+++ b/jNFRT4oBgHgl3EQfWDfG/content/tmp_files/load_file.txt
@@ -0,0 +1,326 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf,len=325
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='13542v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='OC] 31 Jan 2023 Springer Nature 2021 LATEX template Theoretical aspects in penalty hyperparameters optimization Flavia Esposito1,2*†, Laura Selicato1,2† and Caterina Sportelli1,3† 1*Department of Mathematics, Universit`a degli Studi di Bari Aldo Moro, via Orabona, 4, Bari, 70125, Italy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 2Member of INDAM research group, GNCS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 3Member of INDAM research group, GNAMPA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Corresponding author(s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' E-mail(s): flavia.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='esposito@uniba.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='it;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Contributing authors: laura.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='selicato@uniba.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='it;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' caterina.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='sportelli@uniba.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='it;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' †These authors contributed equally to this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Abstract Learning processes are useful methodologies able to improve knowl- edge of real phenomena.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' These are often dependent on hyperparameters, variables set before the training process and regulating the learning procedure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Hyperparameters optimization problem is an open issue in learning approaches since it can strongly affect any real data analysis.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' They are usually selected using Grid-Search or Cross Validation tech- niques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' No automatic tuning procedure exists especially if we focus on an unsupervised learning scenario.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This study aims to assess some theoretical considerations for tun- ing penalty hyperparameters in optimization problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' It considers a bi-level formulation tuning problem in an unsupervised context, by using Gradient-based methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Suitable conditions for the exis- tence of a minimizer in an infinite-dimensional Hilbert space are outlined, together with some theoretical results, applicable in all those situations when it is unnecessary or not possible obtaining an exact minimizer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' An iterative algorithmic strategy is considered, equipped with a stopping criterion via Ekeland’s variational principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Keywords: Hyperparameters optimization, learning approaches, existence results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 1 Springer Nature 2021 LATEX template 2 HPO: theoretical results MSC Classification: 68Q32 , 46N10 , 90C46 , 49J27 , 90C48 Acknowledgment The authors would like to thank Prof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Candela and Prof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Del Buono from Universit`a degli Studi di Bari Aldo Moro for the deep discussions on the preliminary version of this manuscript.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This work was supported by INDAM- GNCS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 1 Introduction Training a Machine Learning (ML) algorithm is quite important to produce data-driven models, which can be successfully applied in real life applica- tions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' These processes often require to specify several variables by the users, namely hyperparameters, which must be set before the learning procedure starts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Hyperparameters govern the whole learning process and play a cru- cial role in guaranteeing good model performances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' They are often manually specified, and the lack of an automatic tuning procedure makes the field of Hyperparameter Optimization (HPO) an ever-evolving topic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The literature offers various solutions for hyperparameters tuning, from Gradient-based to Black-Box or Bayesian’s approaches, beside some naive but daily used meth- ods such as Grid and Random search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A brief overview on existing methods can be found in [1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Hyperparameters can be of different types (discrete, con- tinuous, categorical), and in most cases, the number of their configurations to explore is infinite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This paves the way for a mathematical formalization of the HPO in ML context with abstract spaces, such as Hilbert spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A supervised learning algorithm may be represented as a mapping that takes a configuration of hyperparameter and a dataset D and returns an hypothesis [2]: A : Λ × D → H;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A(λ, D) = h, with D = � N∈N (X × Y )N (1) where D is the space of finite dimensional dataset, representing a task, X and Y are the input and output spaces, Λ is an hyperparameter space, and H is an hypothesis space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A quite standard claim for the hypotheses set is to be a linear function space, endowed with a suitable norm (more binding arising from an inner product): two requirements satisfied when H is a Hilbert space of func- tions over X1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Assuming an Hilbert space structure on the hypothesis space has some advantages: (i) practical computations reduced to ordinary linear algebra operations and (ii) self duality;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' that is for any x ∈ X a representative 1If X is an infinite dimensional space the boundedness is needed, too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template HPO: theoretical results 3 of x can be found, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', kx ∈ H exists such that h(x) = ⟨kx, h⟩ for all h ∈ H, (2) where kx is a suitable positive definite “kernel”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This construction gives the chance to connect the abstract structure of H and what its elements actually are, flipping the construction of the hypotheses set from the kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Providing a suitable positive function k on X, H can be set as the minimal complete space of functions involving all {kx}x∈X equipped with the scalar product in (2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, H is outlined in a unique way, and it is named the Representing Kernel Hilbert Space mapped to the kernel k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Starting from this abstract scenario, one can deepen the HPO in supervised ML.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Formally, HPO can be formulated as the problem of minimizing the dis- crepancy between A, trained on a given training dataset Dtr, and a validation dataset Dval [3], to find the optimal λ∗ such that λ∗ = argmin λ∈Λ V (A(λ, Dtr), Dval), where V : H × X → R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' (3) In this study, we will address problem (3) through Gradient-based meth- ods (GB), by using a bi-level approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Bi-level programming solves an outer optimization problem subject to the optimality of an inner optimization prob- lem, and it can be adopted to formalize HPO for any learning algorithm [4–7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' We will work on Hilbert spaces for solving HPO in unsupervised problems, considering as hyperparameter the penalty coefficient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' We already treat this aspect in the particular and more specific case of Nonnegative Matrix Factor- ization task, encurring in some generalization problems and restrictions of the theorems’ assumptions [8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' To overcome the difficulties in ensuring the theoretical assumptions when real data domains are considered, this work extends existence and uniqueness theorems for the solution of the hyperparameters bi-level problem to the more general framework of infinite dimensional Hilbert space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This latter also allows the application of the Ekeland’s variational principle to state that whenever a functional is not guaranteed to have a minimum, under suitable assumptions, a “good” substitute can be found, namely the best one can get as an approximate minimum.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' One of the purposes of this paper is to use this theoretical tool as a stopping criterion for the update of the hyperparameters as we will see later.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The outline of the paper is as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Section 2 introduces the classical bi-level formalization of HPO and some preliminary notions in a supervised context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Section 3 illustrates our proposal, extension on the unsupervised con- text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A general framework addressing HPO in Hilbert space is also set, and some general abstract tools are stated in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Section 5 presents a critical discussion and some practical considerations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Finally, Section 6 summarizes the obtained results and draws some conclusions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template 4 HPO: theoretical results 2 Previous works and preliminaries As briefly mentioned in the introduction, in a supervised learning scenario, HPO can be addressed through a bi-level formulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This approach looks for the hyperparameters λ such that the minimization of the regularized training leads to the best performance of the trained data-driven model on a validation set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Accordingly to the ideas introduced in [9, 10], the best hyperparameters for a data learning task can be selected as the solution of the following problem: min{J(λ) : λ ∈ Λ}, (4) J(λ) = inf{E(wλ, λ) : wλ ∈ argmin u∈Rr Lλ(u)}, (5) where w ∈ Rr are r parameters, J : Λ → R is the so-called Response Function of the outer problem E : Rr × Λ → R, and for every λ ∈ Λ ⊂ Rp, Lλ : Rr → R is the inner problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' A reformulation of HPO as a bi-level optimization problem is also solved via some GB algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In particular, in GB methods HPO is addressed with classical procedure for continuous optimization, in which the hyperparameter update is given by λt+1 = λt − αht(λ) (6) where ht is an approximation of the gradient of function J and α is a step size.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' It is known that the main challenge in this context is the computation of ht, called hypergradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In several cases, a numerical approximation of the hypergradient can be calculated for real-valued hyperparameters, although few learning algorithms are differentiable in the classical sense.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' There are two main strategies for computing the hypergradient: iterative dif- ferentiation [9, 11, 12] and implicit differentiation [13, 14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The former requires calculating the exact gradient of an approximate objective.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This is defined through the recursive application of an optimization dynamics that aims to replace and approximate the learning algorithm A;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' the latter involves the numerical application of the implicit function theorem to the solution mapping A(Dtr;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' ·), when it is expressible through an appropriate equation [2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In this study, we follow the iterative strategy, so that problem in (4)-(5) can be addressed through a dynamical system type approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' If the following hypothesis hold: Hypothesis 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' the set Λ is a compact subset of R;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' the Error Function E : Rr × Λ → R is jointly continuous;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' the map (w, λ) → Lλ(w) is jointly continuous, and then the problem argmin Lλ is a singleton for every λ ∈ Λ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' the problem wλ = argmin Lλ remains bounded as λ varies in Λ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template HPO: theoretical results 5 the problem in (4)-(5) becomes: min λ∈Λ J(λ) = E(w∗ λ, λ), w∗ λ = argmin u Lλ(u).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' (7) It can be proved that the optimal solution (wλ∗, λ∗) of problem (7) exists [11].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The goal of HPO is to minimize the validation error of model gw : X → Y , parameterized by a vector w ∈ Rr, with respect to hyperparameters λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Considering the penalty optimization problems in which hyperparameter is the penalty coefficient λ ∈ R+, the Inner Problem is the penalized empirical error represented by L, defined as: Lλ(w) = � (x,y)∈Dtr ℓ(gw(x), y) + λr(w), (8) where ℓ is the loss function, Dtr = {(xi, yi)}n i=1 the training set, and r : Rr → R is a penalty function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' While the Outer Problem is the generalized error of gw represented by E: E(w, λ) = � (x,y)∈Dval ℓ(gw(x), y), (9) where Dval = {(xi, yi)}n i=1 is the validation set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Note that E does not explicitly depend on λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This work will allow overcoming some assumptions of Hypothesis 1 (such as compactness) that are difficult to satisfy in real data learning contexts, and also to use some theoretical result as the Ekeland’s variational principle, stated in the following, to improve iterative algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Theorem 1 (Ekeland’s variational principle) [15] Let (V, d) be a complete metric space and J : V → ¯R be a lower semi-continuous function which is bounded from below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Suppose that ε > 0 and ˜v ∈ V exist such that J(˜v) ≤ inf V J + ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Then, given any ρ > 0, vρ ∈ V exists such that J(vρ) ≤ J(˜v), d(vρ, ˜v) ≤ ε ρ, and J(vρ) < J(v) + ρ d(vρ, v) ∀ v ̸= vρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 3 Our Proposal The bi-level HPO framework can be modified to include unsupervised learning paradigms, generally designed to detect some useful latent structure embedded in data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Tuning hyperparameters for unsupervised learning models is more complex than the supervised case due to the lack of the output space, which defines the ground truth collected in the validation set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template 6 HPO: theoretical results This section describes a general framework to address HPO in Hilbert spaces for the unsupervised case and a corollary of the Ekeland’s variational principle used to derive a useful stopping criterion for iterative algorithms solving this HPO.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Let X ∈ Rn×m be a data matrix, with reference to the problem (4)-(5), where now J : Λ → R is a suitable functional and Λ a Hilbert space equipped with the scalar product (·, ·), the outer problem is: E : Rr × Λ → R E(w, λ) = � x∈X ℓ(gw(x)), (10) and for every λ ∈ Λ the inner problem is: L : Rr → R Lλ(w) = � x∈X ℓ(gw(x)) + R(λ, w), (11) where R : Λ × Rr → R is a penalty function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' We want to emphasize the new formulation with respect to (8) regarding the function Lλ, in which each component of the parameter w is penalized independently, and all optimization is performed on the data matrix X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The bi-level problem associated to (10)-(11) can be solved with a dynamical system approach in which the hypergradient is computed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Once the hypergra- dient is achieved a gradient-based approach can be used to find the optimum λ∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The Ekeland’s variational principle can be used to construct an appropriate stopping criterion for iterative algorithms, with the aim of justifying and set- ting the hyperparameters related to the stopping criterion more appropriately.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Roughly speaking, this variational principle asserts that, under assumptions of lower semi-continuity and boundedness from below, if a point ˜λ is an “almost minimum point” for a function J, hence a small perturbation of J exists which attains its minimum at a point “near” to ˜λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' As a fruitful selection of ρ occurs when ρ = √ε and such a choice allows us to reduce the number of hyperparam- eters to the precision error only, thus we will use Theorem 1 in the following form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Corollary 1 Let (V, d) be a complete metric space and J : Λ → ¯R be a lower semi- continuous function which is bounded from below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Suppose that ε > 0 and ˜λ ∈ Λ exist such that J(˜λ) ≤ inf Λ J + ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Then, ˜z ∈ Λ exists such that J(˜z) ≤ J(˜λ), d(˜z, ˜λ) ≤ √ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' and J(˜z) < J(λ) + √ε d(˜z, λ) ∀ λ ̸= ˜z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template HPO: theoretical results 7 4 Main Abstract results In this section, we are ready to weaken the assumptions we discussed earlier and provide results related to the use of the Ekeland’s principle as a stopping criterion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' We mention an abstract result of the existence of a minimizer in Hilbert spaces which has great importance and a wide range of applications in several fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' As just one example, the Riesz’s Representation Theorem, even if implicitly, makes use of the existence of a minimizer [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This is a widely rel- evant issue about Hilbert spaces, which makes them nicer than Banach spaces or other topological vector spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' One can think, for example, that the whole Dirac Bra–ket formalism of quantum mechanics relies on this identification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='1 Abstract Existence Theorem It is well known that each bounded sequence in a normed space Λ has a norm convergent subsequence if and only if it is a finite dimensional normed space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, given a normed space Λ, as the strong topology (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', the one induced by the norm) is too strong to provide any widely appropriate subsequential extraction procedure, one can consider other weak topologies joined with the linear structure of the space and look for subsequential extraction processes therein.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In Banach spaces, as well as in Hilbert spaces, the two most relevant weaker- than-norm topologies are the weak-star topology and the weak topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' As the former is established in dual spaces, the latter is set up in every normed space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The notions of these topologies are not self-contained but fulfill a leading role in many features of the Banach space theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In this regard, here we state some results we will use shortly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Theorem 2 If Λ is a finite-dimensional space, the strong and weak topologies coin- cide.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In particular, it follows that the weak topology is normable, and then clearly metrizable, too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' If Λ is an infinite-dimensional space, the weak topology is strictly contained in the strong topology, namely open sets for the strong topology exist which are not open for the weak topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Furthermore, the weak topology turns to be not metrizable in this case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Definition 1 A functional J : Λ → ¯R with Λ topological space, is said to be lower semi-continuous on Λ if for each a ∈ R, the sublevel sets J−1(] − ∞, a]) = {λ ∈ Λ : J(λ) ≤ a} are closed subsets of Λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In the following we introduce a “generalized Weierstrass Theorem” which gives a criteria for the existence of a minimum for a functional defined on a Hilbert space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' For this reason, the incoming results will be provided for the abstract framework of a Hilbert space although, in some cases, they apply in Springer Nature 2021 LATEX template 8 HPO: theoretical results the more general context of Banach spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, throughout the remaining part of this section we denote by Λ any real infinite dimensional Hilbert space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In an infinite dimensional setting, the following definitions are strictly related to the different notions of weak and strong topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Definition 2 A functional J : Λ → ¯R is said to be strongly (weakly, respectively) lower semi-continuous if J is lower semi-continuous when Λ is equipped with the strong (weak, respectively) topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Definition 3 A functional J : Λ → ¯R is said to be strongly (weakly, respectively) sequentially lower semi-continuous if lim inf n→+∞ J(λn) ≥ J(λ) for any sequence (λn)n ⊂ Λ such that λn → λ (λn ⇀ λ, respectively).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' We proceed by providing some useful results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proposition 3 The following statements are equivalent: i) J : Λ → R is sequentially weakly lower semi-continuous functional;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' ii) the epigraph of J is weakly sequentially closed, where, by definition, it is epi(J) = {(λ, t) ∈ dom(J) × R : J(λ) ≤ t}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Remark 1 As a further consequence of the preliminary Theorem 2, we have that sequential weak lower semi-continuity and weak lower semi-continuity do not match if Λ is infinite dimensional since weak topology is not metrizable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' However, the weaker concept of sequential weak lower semi-continuity meets our needs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proposition 4 Let C ⊆ Λ be a closed and convex subset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Then, C is weakly sequentially closed, too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Since a sequentially weakly closed set is also strongly closed, it follows that a sequentially weakly lower semi-continuous functional is also (strongly) lower semi-continuous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Instead, the converse holds under an additional assumption.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In particular, Proposition 4 allows us to infer the following results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proposition 5 If J : Λ → R is a strongly lower semi-continuous convex functional;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' thus J is weakly sequentially lower semi-continuous, too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proof Since J is lower semi-continuous, thus epi(J) is closed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' On the other hand, since J is convex, so it is epi(J), whence Proposition 4 ensures that epi(J) is weakly sequentially closed, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', J is weakly sequentially lower semi-continuous.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' □ Springer Nature 2021 LATEX template HPO: theoretical results 9 Thus, we are able to state the main result of this section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Theorem 6 Let C ⊂ Λ be a non-empty, closed, bounded and convex subset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Let J : Λ → R be a lower semi-continuous and convex functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus J achieves its minimum in C, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', ¯λ ∈ C exists such that J(¯λ) = inf λ∈C J(λ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proof Let m := inf λ∈C J(λ);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' hence (λn)n ⊂ C exists such that J(λn) → m as n → +∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' (12) Now, our boundness assumption on C implies that, up to subsequences, λ ∈ H exists such that λn ⇀ λ as n → +∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Actually, since C is a closed and convex subset of Λ, thus Proposition 4 applies, which guarantees that λ ∈ C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Finally, from (12), Proposition 5 and Definition 3 we infer that J(¯λ) ≤ m, which gives the desired result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' □ Remark 2 If every closed and bounded subset in a metric space is compact, the space is said to have the Heine–Borel property.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This property holds in every finite dimensional normed space but, in general, may not be true.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Remark 3 We observe that Theorem 6 still holds if the subset C is not bounded as long as we ask for an additional assumption on the functional J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In fact, requiring J to be coercive2 (and if at least ¯λ ∈ C exists such that J(¯λ) < +∞), then any minimizer of J on C necessarily lies in some closed ball of radius r > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In fact, since J(¯λ) < +∞, any minimizer λ of J must have J(λ) ≤ J(¯λ);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' furthermore, since J is coercive, a sufficient large radius r > 0 exists such that J(λ) > J(¯λ) for all λ ∈ C with ∥λ∥ > r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, any minimizer, if exists, lies in the ball {λ ∈ C : ∥λ∥ ≤ r}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In particular, Theorem 6 applies to the intersection between C and a closed ball of suitable radius, since it turns to be convex if we formally require C to be closed and convex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Namely, the following result holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Corollary 2 Let C ⊂ Λ be a non-empty, closed and convex subset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Let J : Λ → R be a lower semi-continuous, convex and coercive functional.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus J achieves its minimum, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', ¯λ ∈ C exists such that J(¯λ) = inf λ∈C J(λ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Now we introduce a couple of results which are a direct consequence of Ekeland’s variational principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' For the sake of completeness, here we provide them with all the details (see [17] for the original statements).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Let Λ be a complete metric space and J : Λ → R be the lower semicontinuous response function on Λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Supposte that a point λ ∈ Λ exists such that J(λ) < +∞.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, the following results hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 2We say that a functional J : H → R is coercive if J(u) → ∞ as ∥u∥ → ∞, u ∈ H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template 10 HPO: theoretical results Theorem 7 (Perturbation Result) Let Jλ : Λ → ¯R be a lower semicontinuous function such that the inequality |Jλ(γ) − J(γ)| ≤ ζ(d(γ, λ)) holds ∀γ ∈ Λ, (13) where Jλ(·) denote model function3, ζ is some growth function4, and let λ+ be a minimizers of Jλ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' If λ+ coincides with λ, then |∇J(λ)| = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' On the other hand, if λ and λ+ are distinct, then a point ˆλ ∈ X exists which satisfies 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' d(λ+, ˆλ) ≤ 2 · ζ(d(λ+,λ)) ζ′(d(λ+,λ)) (point proximity) 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' J(ˆλ) ≤ J(λ+) + ζ(d(λ+, λ)) (value proximity).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proof By Taylor’s theorem it is simple to verify that |∇Jλ|(λ) = |∇J|(λ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Now, since λ is a minimizer, we have |∇J(λ)| = 0 if λ+ = λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' On the other hand, if λ+ ̸= λ, from inequality (13) and the definition of λ+, it follows that J(γ) ≥ Jλ(λ+) − ζ(d(γ, λ)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Let us define the new function G(γ) := J(γ) + ζ(d(γ, λ)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, from assumption (13) and inequality inf G ≥ Jλ(λ+) we infer that G(λ+) − inf G ≤ J(λ+) − Jλ(λ+) + ζ(d(λ+, λ)) ≤ 2ζ(d(λ+, λ)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Whence, Theorem 1 applies and, having ε := 2ζ(d(λ+, λ)), for all ρ > 0 λρ exists such that G(λρ) ≤ G(λ+) and d(λ+, λρ) ≤ ε ρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The desired result follows simply by placing ρ = ζ′(d(λ+, λ)) with ˆλ = λρ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' □ An immediate consequence of Theorem 7 is the following subsequence convergence result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Corollary 3 (Subsequence convergence to stationary points) Consider a sequence of points λk and closed functions Jλk : Λ → ¯R satisfying λk+1 = argmin γ Jλk(γ) and d(λk+1, λk) → 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Moreover suppose that the inequality |Jλk(γ) − J(γ)| ≤ ζ(d(λk, γ)) holds ∀k ∈ N and γ ∈ Λ, (14) where ζ is a proper growth function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' If (λ∗, J(λ∗)) is a limit point of the sequence (λk, J(λk)), then λ∗ is stationary for J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Two interesting consequences for convergence analysis flow from there.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Sup- pose that the models are chosen in such a way that the step-sizes ∥λk+1 − λk∥ tend to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This assumption is often enforced by ensuring that J(λk+1) < J(λk) by at least a multiple of ∥λk+1 − λk∥2 (sufficient decrease condition).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 3As model function we mean the Taylor’s expansion of J in λ, stopped to the first order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 4A differentiable univariate function ζ : R+ → R+ is called a growth function if it satisfies ζ(0) = ζ′(0) = 0 and ζ′ > 0 on (0, +∞).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' If in addition, equalities lim t→0 ζ′(t) = lim t→0 ζ(t)/ζ′(t) = 0 hold, we say that ζ is a proper growth function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template HPO: theoretical results 11 Then, assuming for simplicity that J is continuous on its domain, any limit point λ∗ of the iterate sequence λk will be stationary for the problem (Corol- lary 3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Thus, by choosing an error ε, we can stop update (6) for GB algorithms in the context of bi-level HPO for penalty hyperparameter, according to the pseudo code in 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Algorithm 1 Pseudo-code Require: Error ε.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Some starting points λ0, λ1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Ensure: Optimum λ∗ 1: while ∥λt − λt−1∥ > ε do 2: Compute h(λ);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 3: update λt+1 = λt − αht(λt);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 4: t+ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 5: end while 5 Discussion and practical considerations We want to emphasize that moving to infinite dimensional Hilbert spaces are not a mere abstract pretense, but it is also important in some application contexts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' For example, when Support Vector Machine (SVM) are taken into consideration, a well known “kernel trick” permits to interpret a Gaussian kernel as an inner product in a feature space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This is potentially infinite- dimensional, allowing to read the SVM classifier function as a linear function in the feature space [18].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Another example is provided by the quantum system possible states problem, in which the state of a free particle can be described as vectors residing in a complex separable Hilbert space [19].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Indeed, the strength of this article lies in theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Both the existence theorem and the stopping criterion allow us to build an approach based on solid mathematical foundations useful for future extensions and generalizations to other problems, too.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' For example, infinite-dimensional Covariance Descriptors (CovDs) for classification is a fertile application arena for the extensions devel- oped here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' This finds motivation in the fact that CovDs could be mapped to Reproducing Kernel Hilbert Space (RKHS) via the use of SPD-specific kernels [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 6 Conclusions In this paper, we studied the task of penalty HPO and we provided a math- ematical formulation, based on Hilbert spaces, to address this issue in an unsupervised context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Focusing on bi-level formulation, we showed some relaxed theoretical results both to weaken the hypotheses necessary for the existence of the solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer Nature 2021 LATEX template 12 HPO: theoretical results Our approach differs from the more standard techniques in reducing the random or black box strategies giving stronger mathematical generalization suitable also when it is not possible obtaining exact minimizer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' We also propose to use the Ekeland’s principle as a stopping criterion, which fits well in the context of GB methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Declarations Funding: The author F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' was funded by REFIN Project, grant number 363BB1F4, Reference project idea UNIBA027 “Un modello numerico- matematico basato su metodologie di algebra lineare e multilineare per l’analisi di dati genomici”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' The author C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' was partially supported by PRIN project “Qualitative and quantitative aspects of nonlinear PDEs” (2017JPCAPN 005) funded by Ministero dell’Istruzione, dell’Universit`a e della Ricerca.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Conflict of interest: The authors have no relevant financial or non- financial interests to disclose.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Data availability: Data sharing not applicable to this article as no datasets were generated or analysed during the current study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' References [1] Del Buono, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Esposito, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Selicato, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Methods for hyperparame- ters optimization in learning approaches: An overview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: International Conference on Machine Learning, Optimization, and Data Science, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 100–112 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer [2] Franceschi, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': A unified framework for gradient-based hyperparameter optimization and meta-learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PhD thesis, UCL (University College London) (2021) [3] Feurer, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Hutter, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Hyperparameter Optimization, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 3–33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer, USA (2019) [4] Bard, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' : Practical Bilevel Optimization: Algorithms and Applications vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Springer, Boston (2013) [5] Colson, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Marcotte, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Savard, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': An overview of bilevel optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Annals of operations research 153(1), 235–256 (2007) [6] Bertrand, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Klopfenstein, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Blondel, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Vaiter, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Gramfort, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Salmon, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Implicit differentiation of lasso-type models for hyperparam- eter optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: International Conference on Machine Learning, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 810–821 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PMLR [7] Franceschi, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': A unified framework for gradient-based hyperparameter optimization and meta-learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PhD thesis, University College London Springer Nature 2021 LATEX template HPO: theoretical results 13 (2021) [8] Del Buono, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Esposito, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Selicato, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Zdunek, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Optimizing Penal- ization Hyperparameters in nonnegative matrix factorizations problems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Preprint at http://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='org/abs/2203.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='13129 (2022) [9] Franceschi, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Donini, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Frasconi, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Pontil, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Forward and reverse gradient-based hyperparameter optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: International Confer- ence on Machine Learning, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 1165–1173 (2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PMLR [10] Vincent, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Gelly, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Le Roux, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Bousquet, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Online hyper-parameter optimization (2018) [11] Franceschi, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Frasconi, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Salzo, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Grazzi, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Pontil, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Bilevel programming for hyperparameter optimization and meta-learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: International Conference on Machine Learning, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 1568–1577 (2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PMLR [12] Maclaurin, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Duvenaud, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Adams, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Gradient-based hyperparameter optimization through reversible learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: Proc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' of ICML, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 2113– 2122 (2015) [13] Pedregosa, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Hyperparameter optimization with approximate gradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: ICML, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 737–746 (2016).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PMLR [14] Lorraine, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Vicol, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Duvenaud, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Optimizing millions of hyperparam- eters by implicit differentiation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: International Conference on Artificial Intelligence and Statistics, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 1540–1552 (2020).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' PMLR [15] Ekeland, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': On the variational principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Journal of Mathematical Anal- ysis and Applications 47(2), 324–353 (1974) [16] Rudin, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Real and Complex Analysis, (1987) [17] Drusvyatskiy, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Ioffe, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Lewis, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Nonsmooth optimization using taylor-like models: error bounds, convergence, and termination criteria.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Mathematical Programming, 1–27 (2019) [18] Rossi, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Villa, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Classification in hilbert spaces with support vector machines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' Proceedings of ASMDA, 635–642 (2005) [19] Ying, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Foundations of Quantum Programming, (2016) [20] Harandi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Salzmann, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=', Porikli, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=': Bregman divergences for infinite dimensional covariance matrices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
+page_content=' 1003–1010 (2014)' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jNFRT4oBgHgl3EQfWDfG/content/2301.13542v1.pdf'}
diff --git a/jtFPT4oBgHgl3EQf1TVB/content/tmp_files/2301.13182v1.pdf.txt b/jtFPT4oBgHgl3EQf1TVB/content/tmp_files/2301.13182v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..0419b48607eac45a07389cef37b50ea8cf0856c0
--- /dev/null
+++ b/jtFPT4oBgHgl3EQf1TVB/content/tmp_files/2301.13182v1.pdf.txt
@@ -0,0 +1,2218 @@
+arXiv:2301.13182v1 [hep-th] 30 Jan 2023
+Open string field theory with stubs
+Martin Schnabl and Georg Stettinger
+January 31, 2023
+CEICO, Institute of Physics of the Czech Academy of Sciences,
+Na Slovance 2, 182 00 Prague 8, Czech Republic
+Abstract
+There are various reasons why adding stubs to the vertices of open string field theory (OSFT) is
+interesting: Not only the stubs can tame certain singularities and make the theory more well-behaved,
+but also the new theory shares a lot of similarities with closed string field theory, which helps to improve
+our understanding of its structure and possible solutions. In this paper we explore two natural ways
+of implementing stubs into the framework of OSFT, resulting in an A∞-algebra giving rise to infinitely
+many vertices. We find two distinct consistent actions, both generated by a field redefinition, interestingly
+sharing the same equations of motion. In the last section we illustrate their relationship and physical
+meaning by applying our construction to nearly marginal solutions.
+Contents
+1
+Introduction and motivation
+2
+2
+Deforming the Witten action using stubs
+4
+2.1
+Elements of homological perturbation theory
+. . . . . . . . . . . . . . . . . . . . . . . . . . .
+5
+2.2
+Transferring algebraic structure: SDR-case . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+5
+2.3
+Transferring algebraic structure: Non-SDR-case . . . . . . . . . . . . . . . . . . . . . . . . . .
+6
+2.4
+Higher products using tree diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+8
+2.5
+Proof of cyclicity to all orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+9
+2.6
+Geometric picture
+. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+10
+3
+Analytic solutions and action(s)
+10
+3.1
+Projection cohomomorphism from the HPL . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+10
+3.2
+The action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+12
+3.3
+Elements from closed string field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+13
+3.4
+Constructing the field redefinition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+14
+3.5
+Proof of the ansatz for the infinitesimal field redefinition . . . . . . . . . . . . . . . . . . . . .
+16
+3.6
+Finite field redefinition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+19
+1
+
+4
+Physical interpretation
+19
+4.1
+Nearly marginal solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+20
+4.2
+The A∞-action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+21
+5
+Conclusion and outlook
+22
+1
+Introduction and motivation
+It is a well-known fact that the algebraic and geometric structures of open and closed string field theory are
+fundamentally different. The OSFT action (Witten action) consists of a standard kinetic term and a single
+cubic interaction and reads [1]
+S (Ψ) = 1
+2ω (Ψ, QΨ) + 1
+3ω (Ψ, Ψ ∗ Ψ) .
+(1)
+The algebraic ingredients are a nilpotent differential given by the BRST-operator, the star product and a
+symplectic form ω, which all together form a cyclic differential graded algebra. In the following, We will
+use the coalgebra notation of [2] with m2 (Ψ1, Ψ2) := (−)d(Ψ1) Ψ1 ∗ Ψ2 and the shifted degree given by
+d (Ψ) = gh (Ψ) + 1. Now the defining algebraic properties are
+Q2 = 0,
+(2)
+Qm2 (Ψ1, Ψ2) = −m2 (QΨ1, Ψ2) − (−)d(Ψ1) m2 (Ψ1, QΨ2) ,
+(3)
+m2 (Ψ1, m2 (Ψ2, Ψ3)) = − (−)d(Ψ1) m2 (m2 (Ψ1, Ψ2) , Ψ3) ,
+(4)
+ω (Ψ1, m2 (Ψ2, Ψ3)) = − (−)d(Ψ1) ω (m2 (Ψ1, Ψ2) , Ψ3) ,
+(5)
+ω (Ψ1, QΨ2) = − (−)d(Ψ1) ω (QΨ1, Ψ2) .
+(6)
+In contrast, the classical closed string field theory (CSFT) action [3] contains infinitely many vertices and
+reads
+S (Ψ) = 1
+2ω (Ψ, QΨ) +
+∞
+�
+n=3
+κn−2
+n!
+{Ψn} ,
+(7)
+where the vertices
+{Ψn} =
+�
+V0,n
+Ω(0)0,n
+Ψn
+(8)
+are given by integrating certain differential forms over vertex regions V0,n in the moduli space of n-punctured
+Riemann surfaces. At quantum level there are also vertices of higher genus to consider, i. e. intrinsic loops
+already present within the elementary vertices. However, they will not be of interest here for now.
+The vertices also can be decomposed into a symplectic form ω and higher products ln:
+�
+Ψn+1�
+= ω (Ψ, ln (Ψ, Ψ, ..., Ψ)) .
+(9)
+In contrast to the open string star product, the ln are all totally symmetric. The objects Q =: l1, all higher
+ln and ω give rise to a cyclic L∞-algebra:
+lnl1 (Ψ1, ..., Ψn) + ln−1l2 (Ψ1, ..., Ψn) + ... + l1ln (Ψ1, ..., Ψn) = 0,
+∀n
+(10)
+ω (Ψ1, ln (Ψ2, ..., Ψn+1)) = − (−)d(Ψ1) ω (ln (Ψ1, ..., Ψn) , Ψn+1) .
+∀n
+(11)
+2
+
+The multiplication of the multilinear products in the first line is defined as
+lkll (Ψ1, ..., Ψk+l−1) =
+�
+σ
+(−1)ǫ(σ)
+l! (k − 1)!lk
+�
+ll
+�
+Ψσ(1), ..., Ψσ(l)
+�
+, Ψσ(l+1), ..., Ψσ(k+l−1)
+�
+,
+(12)
+where the sum runs over all permutations of k + l − 1 elements and the factor (−1)ǫ(σ) is just the sign picked
+up when permuting the Ψs.
+The different algebraic structures have a geometric origin: For a unitary theory it is necessary that the
+Feynman string diagrams computed from the action cover the full moduli space of n-punctured Riemann
+surfaces exactly once. For the open string where one needs to consider surfaces with boundaries it has been
+shown by Zwiebach [4] that this is indeed the case for the action (1) with only one cubic interaction. In
+contrast, for the closed string, i. e. surfaces without boundaries, such a simple cubic representation of the
+action seems not to be possible, see [5].
+The simplicity of the OSFT action allowed for the discovery of analytic classical solutions, most impor-
+tantly the open string tachyon vacuum [6]. In classical CSFT, no such analytic solutions have been found yet
+[7, 8]. To make some effort into this direction, we want to propose the following strategy: If CSFT cannot be
+simplified easily, maybe we can make OSFT “more complicated” in the sense that its algebraic and geometric
+structure resembles that of CSFT? There are two motivations for that: First, it could help to make the very
+abstract language of CSFT more intuitive and tractable. Second, if we can translate the known analytic
+solutions to the new, deformed OSFT, it could give an idea how CSFT solutions might look like. In total
+there are three steps to do:
+1. Find a consistent deformation of the Witten OSFT such that its mathematical structure resembles
+CSFT
+2. Find analytic solutions of this deformed theory
+3. Make an educated guess for analytic solutions of CSFT
+The first two tasks will be worked out in this paper while the third one is left for the future.
+The deformation we will consider has many additional interesting aspects: First, OSFT exhibits some
+issues with singularities or ill-defined quantities: For example, there exist “identity-based” solutions [9] (for a
+recent discussion see [10]) which solve the equations of motion but do not have a well defined action. There are
+hints that these solutions could be better behaved in the deformed theory. Moreover, the deformed products
+M2, M3, . . . defined in the next section could allow for a natural definition of a Banach type A∞-algebra over
+the space of string fields.
+The paper is organized as follows: In the second section the deformation using stubs [11] is discussed in
+detail, resulting in an explicit definition of the higher products of the A∞-algebra. The necessary mathemat-
+ical ingredients, namely homological perturbation theory and homotopy transfer, are introduced as well. As
+a side result, we give a consistent method of applying homotopy transfer to general homotopy equivalences
+which are not deformation retracts. Section three deals with analytical solutions and the action of the stubbed
+theory. Surprisingly, we find two consistent actions with the same equations of motion, both generated by
+a field redefinition. The two field redefinitions, one coming from homological perturbation theory and one
+motivated by closed string field theory, are derived explicitly. The last section is devoted to the physical
+interpretation of our results. We apply the whole construction to a suitable class of solutions and compute
+the action up to finite order.
+3
+
+2
+Deforming the Witten action using stubs
+We have seen that the algebraic structure of CSFT is that of an L∞-algebra, so naively we should try to
+find an L∞-based deformation of OSFT. However, this would imply a commutative product of open strings,
+which would hence be fundamentally different from the Witten product. We still want the deformation be
+equivalent to Witten theory, so the best we can do is to aim for an A∞-algebra with products Mn obeying
+MnM1 (Ψ1, . . . , Ψn) + Mn−1M2 (Ψ1, . . . , Ψn) + . . . + M1Mn (Ψ1, . . . , Ψn) = 0
+∀n
+(13)
+and the multiplication given by
+MkMl (Ψ1, . . . , Ψk+l−1) = Mk (Ml (Ψ1, . . . , Ψl) , Ψl+1, . . . , Ψk+l−1) +
++ (−)d(Ψ1) Mk (Ψ1, Ml (Ψ2, . . . , Ψl+1) , Ψl+2, . . . , Ψk+l−1) + · · · +
++ (−)d(Ψ1)+...+d(Ψk−1) Mk (Ψ1, . . . , Ψk−1, Ml (Ψk, . . . , Ψk+l−1)) .
+(14)
+From now on we will use the tensor coalgebra notation of [2] where the A∞-relations take the simple form
+n
+�
+k=1
+MkMn+1−k = 0.
+(15)
+Geometrically, to mimic the situation of CSFT, we would like to have a three-vertex that does not give rise
+to a full moduli space cover, such that higher vertices are necessary. One modification that achieves both of
+these requirements is to attach stubs on the three vertex, which means, small pieces of propagating strings
+on each input. In this way, there will appear some regions of moduli space that are not covered by Feynman
+diagrams and hence have to be taken over by elementary vertices. The higher products which will give rise
+to those vertices will indeed form an A∞-algebra.
+The Hamiltonian of our CFT which generates time evolution is L0, so the operator which inserts a strip
+of length λ into a string diagram is e−λL0. The natural modification of the star product which attaches stubs
+symmetrically on all three inputs is then given by [12]:
+m2 (·, ·) → M2 (·, ·) = e−λL0m2
+�
+e−λL0·, e−λL0·
+�
+.
+(16)
+M2 is cyclic with respect to the simplectic form ω and Q-invariant:
+ω (Ψ1, M2 (Ψ2, Ψ3)) = − (−)d(Ψ1) ω (M2 (Ψ1, Ψ2) , Ψ3) ,
+(17)
+QM2 (Ψ1, Ψ2) = −M2 (QΨ1, Ψ2) − (−)d(Ψ1) M2 (Ψ1, QΨ2) .
+(18)
+Those relations follow from [Q, L0] = 0 and the fact that L0 is BPZ-even. It is easy to see that M3 is not
+associative as expected for an A∞-algebra:
+e−λL0m2
+�
+e−λL0Ψ1, e−2λL0m2
+�
+e−λL0Ψ2, e−λL0Ψ3
+��
++ e−λL0m2
+�
+e−2λL0m2
+�
+e−λL0Ψ1, e−λL0Ψ2
+�
+, e−λL0Ψ3
+�
+̸= 0.
+(19)
+The task is now to explicitly determine all the higher products and prove that they satisfy all relevant
+properties. This will be done in two ways, first in a formal algebraic way using homological perturbation
+theory and second in a more heuristic way using tree diagrams.
+4
+
+2.1
+Elements of homological perturbation theory
+The starting point of homological perturbation theory (HPT) is a chain homotopy equivalence. Let V , W
+be two chain complexes, i. e. graded vector spaces with a nilpotent differential of degree one denoted by
+dV (dW ). Furthermore, we are given two chain maps p : V → W and i : W → V of degree zero as well as a
+homotopy map h : V → V of degree minus one obeying the relations
+pdV = dW p
+(20)
+idW = dV i
+(21)
+ip − 1 = hdV + dV h.
+(22)
+This ensures that i and p as well as the combination ip leave cohomology classes invariant. The most common
+case in literature is that of a special deformation retract (SDR) where W is isomorphic to a subspace of V ,
+see e.g. [13]. The maps i and p are then given by the canonical inclusion and projection, respectively. It
+follows trivially that pi = 1 and additionally the following annihilation conditions are demanded:
+hh = 0,
+hi = 0,
+ph = 0
+(23)
+It can be shown that any deformation retract with pi = 1 can be made an SDR by a redefinition of the maps
+[14].
+Lets assume that the differential dV is perturbed by some δ of degree one such that (dV + δ)2 = 0 again.
+The homological perturbation lemma (HPL) now states that one can construct a new SDR using the perturbed
+differential with the other maps given by
+DV = dV + δ
+(24)
+DW = dW + pδ (1 − hδ)−1 i
+(25)
+I = (1 − hδ)−1 i
+(26)
+P = p (1 − hδ)−1
+(27)
+H = (1 − hδ)−1 h
+(28)
+The expression (1 − hδ)−1 requires some explanation: Usually it should be understood as a geometric series
+(1 − hδ)−1 = 1 + hδ + hδhδ + hδhδhδ + ...
+(29)
+We demand for the validity of the HPL that this series either converges or terminates, which will be the case
+in our examples. The proof of the lemma including all relevant relations is straight forward but tedious and
+can be found in [14, 2].
+2.2
+Transferring algebraic structure: SDR-case
+Our main purpose for using the HPL will be transferring some algebraic structure from one side of the
+homotopy equivalence to the other. Lets consider an SDR with an associative multiplication defined on V . It
+is a natural question to ask if there exists some “induced product” defined on W: For example, for a, b ∈ W
+the product ia · ib in V need not be in W anymore. The only thing that could be done is to project it to W,
+i. e. define a · b |W = p (ia · ib), but this product will in general not be associative anymore. What happens
+5
+
+is that an associative algebra on one side of the equivalence becomes an A∞-algebra after transferring to the
+other side. A formal way to construct all the higher products is given by the so-called tensor trick:
+One defines a new SDR over the tensor algebra of V (W) with the tensorial versions of the maps given
+by
+dV(W) =
+∞
+�
+n=1
+n−1
+�
+k=0
+1⊗k ⊗ dV (W) ⊗ 1⊗n−k−1
+(30)
+i =
+∞
+�
+n=1
+i⊗n
+(31)
+p =
+∞
+�
+n=1
+p⊗n
+(32)
+h =
+∞
+�
+n=1
+n−1
+�
+k=0
+1⊗k ⊗ h ⊗ (ip)⊗n−k−1
+(33)
+dV(W) is the standard tensor coderivation associated to dV (W), see (138), i and p are cohomomorphisms
+while h is defined somewhat asymmetrical which will turn out to be crucial for the formalism to work. One
+can check directly that those definitions indeed give rise to an SDR again. The product on V , denoted by
+m2, can now be treated as a perturbation δ of the coderivative dV,
+DV = dV + m2
+(34)
+where m2 is the coderivation associated to m2. DV
+2 = 0 will be fulfilled as a consequence of m2 being
+associative and dV obeying the Leibniz rule.
+According to the HPL we will be given a new, perturbed
+complex with the maps
+DW = dW + pm2 (1 − hm2)−1 i
+(35)
+I = (1 − hm2)−1 i
+(36)
+P = p (1 − hm2)−1
+(37)
+H = h (1 − hm2)−1
+(38)
+where DW squares to zero and therefore defines an A∞-algebra on W. The higher products can be obtained
+by expanding DW, i. e. projecting onto n inputs and one output:
+Mn = π1DWπn
+(39)
+For example we get
+M2 (·, ·) = pm2 (i·, i·)
+(40)
+as already anticipated above.
+2.3
+Transferring algebraic structure: Non-SDR-case
+We want to apply those concepts now to the problem of attaching stubs in OSFT. The space V should be
+given by the space of string fields HBCF T with its grading d (Ψ) = gh (Ψ) + 1 and the differential Q. By
+comparing (16) with (40) it seems natural to define
+i = p = e−λL0.
+(41)
+6
+
+However, we see immediately that this definition does not give rise to an SDR: i and p are not an inclusion
+and projection anymore and pi ̸= 1. Still, in principle the HPL holds for arbitrary homotopy equivalences,
+so there is a chance to succeed anyway.
+Our choice is so far completely symmetric, so lets define W = V = HBCF T , dW = dV = Q and
+hW = hV = h where we have to solve
+e−2λL0 − 1 = hQ + Qh
+(42)
+The simplest (although not unique) solution for h is
+h = e−2λL0 − 1
+L0
+b0.
+(43)
+It is important to stress that h is well-behaved also for L0 = 0 and does not have any pole.
+One could now try to proceed with the tensor trick as above and eventually compute the map DW but
+this runs into problems: Although DW
+2 = 0 still, as guaranteed by the HPL, DW is not a coderivation
+anymore! This means, it does not obey the co-Leibniz rule (136) anymore. As explained in the Appendix
+of [15], the condition pi = 1 as well as the annihilation relations are necessary (and sufficient) for the tensor
+trick to work. So there has to be some modification to account for that and it will actually turn out to be
+surprisingly simple: We can just take the expression we get for DW and while expanding to calculate the
+higher products, pretend that all SDR-relations are satisfied. More precisely, we define an operator PSDR
+acting on the space of maps from T H → T H which projects on maps in which all SDR-relations hold. This
+means, every time that pi occurs, it will be replaced by 1 and every time hh, hi or ph occurs, the term will
+be discarded:
+PSDR (....hh....) = 0,
+PSDR (....hi....) = 0,
+PSDR (....ph....) = 0,
+PSDR (....pi....) = PSDR (....1....)
+Now our new higher products will be given as
+Mn = PSDRπ1DWπn.
+(44)
+Their associated coderivations can be added together to form a total map called M,
+M =
+∞
+�
+n=1
+∞
+�
+m=1
+m−n
+�
+k=0
+1⊗k ⊗ Mn ⊗ 1⊗m−k−n
+(45)
+It is a coderivation by construction and moreover it squares to zero because in the proof of the HPL, where
+it is shown that D2
+W = 0 (Eq. 1.2.16 of [2]), the SDR-relations are never used. Since we know that DW
+2 = 0
+independently of the SDR-relations, also M2 = 0 and M defines the desired A∞-algebra.
+As an example, lets explicitly calculate M3 :
+PSDRπ1DWπ3 = PSDRπ1pm2hm2iπ3 = PSDR
+�
+pm2 (1 ⊗ h + h ⊗ ip) (1 ⊗ m2 + m2 ⊗ 1) i⊗3�
+= PSDR (pm2 (i·, hm2 (i·, i·)) − pm2 (m2 (i·, i·) , hi·) + pm2 (hi·, ipm2 (i·, i·)) + pm2 (hm2 (i·, i·) , ipi·))
+= pm2 (i·, hm2 (i·, i·)) + pm2 (hm2 (i·, i·) , i·)
+(46)
+The second and third term in the second line contained an hi and were deleted whereas in the last term ipi
+was replaced by i.
+This whole procedure including the proof of the statement may seem quite handwavy, however, there
+exists a more formal and precise way of arriving at the same result using operad theory [16]. A combinatorial
+proof using tree diagrams will be given in the next section.
+7
+
+2.4
+Higher products using tree diagrams
+The proposal is that Mn is equal to the sum of all distinct, rooted, full binary trees with n leaves such that
+every leaf represents one input and the root is the output. With every leaf there is one factor of i associated,
+with every node the product m2, with every internal line h and with the root p. In [17] it is argued that this
+is true for SDRs, since we can construct the products in the same way as for an SDR, we conclude that the
+proposal also holds for our non-SDR case.
+In the tree language, the A∞-relations can be proven directly: The commutator of Q with an n-leaved
+tree gives a sum of n − 2 terms in which one of the n − 2 internal lines h is replaced by 1 − ip. The 1-terms
+actually cancel away because of the associativity of m2: The propagator associated with unity connects two
+nodes m2 which leads to an expression of the form m2 (m2 (A, B) , C) where A, B, C are three subtrees. In the
+sum there always exists a second tree with another propagator turned into unity giving rise to the expression
+m2 (A, m2 (B, C)). These two trees cancel away such that only the -ip-factors remain in total. Now the other
+terms occuring in the relation
+− [Q, Mn] = M2Mn−1 + M3Mn−2 + ... + Mn−1M2
+(47)
+can be interpreted as follows: If we project on one output, π1MkMn+1−k gives a sum of trees where one
+of the terms in Mn+1−k is connected with its root to one of the k leaves of one of the trees in Mk and this
+is done in all possible combinations. The result is a sum of trees with in total n leaves where one of the
+internal lines does not contain h but ip; the i from the leaf of the left tree and the p from the root of the
+right tree. But that is exactly the same sum of terms we have on the l.h.s., indeed it is easy to see that each
+tree occuring on the r.h.s. must also occur on the l.h.s. and vice versa.
+An interesting crosscheck of the π1-projection of (47) can be done by comparing the total number of trees
+on both sides: The number of full binary trees with n + 1 leaves is given by the Catalan number
+Cn =
+1
+n + 1
+�
+2n
+n
+�
+.
+(48)
+This means that on the l.h.s. there are Cn−1 trees in Mn and n − 2 internal lines that can be changed, hence
+a total of Cn−1 (n − 2) trees. On the r.h.s. we have 2 · C1Cn−2 + 3 · C2Cn−3 + ... + (n − 1) Cn−2C2 trees in
+total, leading to the equation
+n−1
+�
+k=2
+(n + 1 − k) Cn−kCk−1 = (n − 2) Cn−1.
+(49)
+The Catalan numbers fulfill the following useful recursive relations:
+Cn+1 = 2 (2n + 1)
+n + 2
+Cn,
+Cn+1 =
+n
+�
+k=0
+Cn−kCk
+(50)
+8
+
+Using them one can proceed by induction: Assuming equation (49) is valid for some n, then
+n
+�
+k=2
+(n + 2 − k) Cn+1−kCk−1
+=
+n
+�
+k=2
+(n + 2 − k) 2 (2n − 2k + 1)
+n − k + 2
+Cn−kCk−1
+= 4
+n−1
+�
+k=2
+(n + 1 − k) Cn−kCk−1 − 2
+n−1
+�
+k=2
+Cn−kCk−1 + 2C0Cn−1
+= (4n − 8) Cn−1 − 2
+n−2
+�
+k=1
+Cn−1−kCk + 2Cn−1
+= (4n − 6) Cn−1 − 2 (Cn − C0Cn−1 − Cn−1C0)
+= (4n − 2)
+n + 1
+2 (2n − 1)Cn − 2Cn
+= (n − 1) Cn
+(51)
+as it should be to complete the induction. This shows that the number of terms in the equation (47) is the
+same on both sides.
+2.5
+Proof of cyclicity to all orders
+Using the tree language it is possible to prove that all higher products Mn are cyclic with respect to the
+BPZ-product. We have to show
+ω (Ψ1, Mn (Ψ2, ..., Ψn+1)) = − (−)d(Ψ1) ω (Mn (Ψ1, ..., Ψn) , Ψn+1) ,
+(52)
+hence we start with a sum of trees on the l.h.s. and use the BPZ-properties of m2 and h as well as p = i† = i
+to rewrite it as the sum of trees on the r.h.s.. The explicit steps are:
+1. Take the p from the root of the tree and write it to the left side of the product where it can be interpreted
+as i, acting on Ψ1.
+2. Take the m2 from the root of the tree and use cyclicity of m2 to apply it on the first two arguments
+inside of ω. This gives a sign factor of − (−)d(Ψ1). In general one will be left with two subtrees then
+with n + 1 leaves in total.
+3. Take the h from the right subtree and use that it is BPZ-even to apply it on the left subtree. This gives
+an additional sign factor of (−)d(left subtree).
+4. Take the m2 from the root of the right subtree and apply it on the first two arguments inside of ω. It
+gives a sign factor of − (−)d(left subtree)+1 (the +1 comes from the h that was shifted in step 3) which
+cancels the sign factor of step 3. Again, one is left with two subtrees.
+5. Repeat steps 3 and 4 until the right subtree only consist of i acting on one input. The total sign factor
+remains − (−)d(Ψ1).
+9
+
+6. Remove the i acting on Ψn+1 and let it act as a p on the left subtree.
+Now the left subtree fulfills all requirements to be an element of Mn and since we also have the right sign
+factor, we have obtained a term contained in the r.h.s. of (52). The manipulations are all uniquely invertible
+so we can conclude that the map between the trees is one-to-one and all terms we need are constructed
+exactly once. As a result, Eq. (52) holds and all higher products are cyclic. Moreover, as already suggested
+by the name, Eq. (52) together with the antisymmetry of ω implies invariance of the vertices under cyclic
+permutations.
+2.6
+Geometric picture
+We have now shown that the higher products fulfill all the algebraic requirements but we do not know
+anything yet about the geometric picture, if they indeed give rise to a full single cover of the moduli space.
+To answer this question, the tree description of the products turns out to be very useful. Lets consider an
+arbitrary string tree diagram using the stubbed three vertex: As long as the external states are on-shell, the
+stubs make no difference because the external legs consist of a semiinfinite strip anyway. On the internal
+lines instead, the stubs make a difference because all internal strips with a length smaller than 2λ do not
+appear. We can conclude that the additional elemantary vertices we need should consist of all tree diagrams
+with all internal strips having a length smaller than 2λ. The Siegel-gauge string propagator in the Schwinger
+parametrization is given by
+� ∞
+0
+dt e−tL0b0 = b0
+L0
+.
+(53)
+The integral over t can be thought of an integral over strips of propagating strings of all different lengths.
+Following this logic, the propagators in our new vertices should be given as
+� 2λ
+0
+dt e−tL0b0 = −e−2λL0 − 1
+L0
+b0 = −h
+(54)
+and hence be equal to minus the homotopy!1 We have constructed the higher products by drawing all binary
+tree diagrams with Witten vertices, h as propagators and e−λL0 on the leaves and the root. Those are in
+one-to-one correspondence with all the Feynman tree diagrams that should make up the new elementary
+vertices. This shows that our higher products Mn indeed define a set of vertices which gives rise to a full
+cover of the moduli space.
+3
+Analytic solutions and action(s)
+3.1
+Projection cohomomorphism from the HPL
+Using the definition
+m = Q + m2
+(55)
+the equations of motion of the original Witten theory can be written in coalgebra language as
+m
+1
+1 − Ψ = 0,
+(56)
+1It is interesting to notice at this point that the simple but not unique choice of h corresponds to choosing Siegel gauge for
+the propagator. The minus sign is just a convention and can be absorbed in the definition of h.
+10
+
+i. e. solutions are Maurer-Cartan elements of the A2-algebra defining the theory. In the same spirit, to find
+solutions of the deformed theory, we have to solve the equation
+M
+1
+1 − Ψ = 0.
+(57)
+In fact, the homological perturbation lemma already gave us an operator which maps solutions of the Witten
+theory to solutions of the stubbed theory. This can be seen as follows: The perturbed projection P is a chain
+map and hence obeys
+Pm = MP.
+(58)
+Now lets assume Ψ∗ is a solution of the Witten theory, then
+(Ψ∗)′ = π1P
+1
+1 − Ψ∗
+(59)
+obeys
+M
+1
+1 − (Ψ∗)′ = M
+1
+1 − π1P
+1
+1−Ψ∗
+= MP
+1
+1 − Ψ∗ = Pm
+1
+1 − Ψ∗ = 0
+(60)
+where Eq. (146) was used. It remains to determine the cohomorphism P for the case where the homotopy
+equivalence is not an SDR. Again, as for the higher products above, the simplest way is to take the expression
+from the HPL
+PHP L = p (1 − m2h)−1 ,
+(61)
+and apply the operator PSDR on its components to get
+Pn = PSDRπ1PHP Lπn.
+(62)
+The resulting maps can then be packaged into a cohomorphism P again. More explicitly, we get for the first
+few orders
+P1 = p
+P2 = pm2 (·, h·) + pm2 (h·, ip·)
+P3 = pm2 (·, hm2 (·, h·)) + pm2 (·, hm2 (h·, ip·)) + pm2 (h·, hm2 (ip·, ip·)) + pm2 (hm2 (·, h·) , ip·)
++ pm2 (hm2 (h·, ip·) , ip·) + pm2 (h·, ipm2 (·, h·)) + pm2 (h·, ipm2 (h·, ip·))
+(63)
+...
+One can now check the equation π1Pm = π1MP order by order:
+π1Pmπ1 = pQ = Qp = π1MPπ1
+π1Pmπ2 = P2 (Q·, ·) + P2 (·, Q·) + pm2
+= −pm2 (Q·, h·) + pm2 (hQ·, ip·) + pm2 (·, hQ·) + pm2 (h·, ipQ·) + pm2
+= −pm2 (Q·, h·) − pm2 (Qh·, ip·) + pm2 ((ip − 1) ·, ip·)
+− pm2 (·, Qh·) + pm2 (·, (ip − 1) ·) + pm2 (h·, Qip·) + pm2
+= Qpm2 (·, h·) + Qpm2 (h·, ip·) + pm2 (ip·, ip·)
+= QP2 + M2 (p·, p·) = π1MPπ2
+(64)
+11
+
+For order three the calculation is already very tedious but it also turns out to work. The important point
+is that in the manipulations that are necessary, the SDR-relations were never used. This implies, since we
+know that (63) works for SDRs, it also works in our case and (63) is a valid definition. Now we have a
+cohomorphism by construction that obeys the chain map relation (58) such that we can construct analytic
+solutions of the deformed theory.2
+3.2
+The action
+The on-shell action is one of the most important observables in OSFT, for example for the tachyon vacuum
+its value is equal to minus the energy of the D-brane which has decayed. Since the stubbed theory should be
+physically equivalent to the original Witten theory, we expect that the values for the on-shell action we get
+in the two theories coincide. The Witten action can be written in coalgebra notation [2] as
+S (Ψ) =
+� 1
+0
+dt ω
+�
+π1∂t
+1
+1 − Ψ (t), π1m
+1
+1 − Ψ (t)
+�
+(65)
+where Ψ (t) is any smooth interpolation between Ψ (0) = 0 and Ψ (1) = Ψ and ∂t the coderivation associated
+to ∂t. Similarly, the stubbed action reads
+S′ (Ψ) =
+� 1
+0
+dt ω
+�
+π1∂t
+1
+1 − Ψ (t), π1M
+1
+1 − Ψ (t)
+�
+(66)
+We would expect now a relation
+S′ �
+(Ψ∗)′� ?= S (Ψ∗)
+(67)
+with Ψ∗ a MC-element of m. However, this relation turns out not to be true: Instead
+S (Ψ) =
+� 1
+0
+dt ω
+�
+π1∂t
+1
+1 − Ψ (t), π1m
+1
+1 − Ψ (t)
+�
+=
+� 1
+0
+dt ω
+�
+π1∂tP−1P
+1
+1 − Ψ (t), π1P−1MP
+1
+1 − Ψ (t)
+�
+=
+� 1
+0
+dt ω
+�
+π1∂tP−1
+1
+1 − π1P
+1
+1−Ψ(t)
+, π1P−1M
+1
+1 − π1P
+1
+1−Ψ(t)
+�
+=
+� 1
+0
+dt ω
+�
+π1P−1∂t
+1
+1 − (Ψ)′ (t), π1P−1M
+1
+1 − (Ψ)′ (t)
+�
+=: ˜S (Ψ′)
+(68)
+where the last line differs from S′ (Ψ′) by the insertions of P−1 on both inputs of ω.
+The invertibility of P is actually a delicate question: In general, a cohomomorphism is invertible iff its
+linear component, i. e. P1 = p is invertible. Now p = e−λL0 inserts a strip of length λ, so one would expect
+the inverse eλL0 to remove a strip of length λ from the world sheet, which is not always possible. On the
+other hand, eλL0 makes sense on a string field expanded in eigenstates of L0 as long as the eigenvalues are
+finite. From now on, we shall assume that this is the case and eλL0 is well-defined on all string fields in
+question.
+Explicitly, the inversion of P works as follows: PP−1 = P−1P should be equal to the identity cohomo-
+morphism, which is identity in its linear component and zero in all higher components. One can now solve
+2It is actually a non-trivial counting problem to determine the number of terms of Pn. It grows faster than the Catalan
+numbers because at order three we have seven terms and at order four already 33.
+12
+
+the components P −1
+n
+= π1P−1πn order by order:
+P −1
+1
+= p−1
+P −1
+2
+= −m2
+�
+p−1·, hp−1·
+�
+− m2
+�
+hp−1·, i·
+�
+P −1
+3
+= −m2
+�
+hp−1·, hm2 (i·, i·)
+�
++ m2
+�
+m2
+�
+p−1·, hp−1·
+�
+, hp−1·
+�
++ m2
+�
+m2
+�
+hp−1·, i·
+�
+, hp−1·
+�
+...
+(69)
+One can now write out in more detail
+˜S (Ψ) = 1
+2ω
+�
+p−1Ψ, p−1QΨ
+�
++ 1
+3ω
+�
+p−1Ψ, p−1M2 (Ψ, Ψ)
+�
++ 1
+3ω
+�
+p−1Ψ, P −1
+2
+((QΨ, Ψ) + (Ψ, QΨ))
+�
++ 1
+3ω
+�
+P −1
+2
+(Ψ, Ψ) , p−1QΨ
+�
++ 1
+4ω
+�
+p−1Ψ, p−1M3 (Ψ, Ψ, Ψ)
+�
++ 1
+4ω
+�
+p−1Ψ, P −1
+2
+((M2 (Ψ, Ψ) , Ψ) + (Ψ, M2 (Ψ, Ψ)))
+�
++ 1
+4ω
+�
+p−1Ψ, P −1
+3
+((QΨ, Ψ, Ψ) + (Ψ, QΨ, Ψ) + (Ψ, Ψ, QΨ))
+�
++ 1
+4ω
+�
+P −1
+2
+(Ψ, Ψ) , p−1M2 (Ψ, Ψ)
+�
++ 1
+4ω
+�
+P −1
+2
+(Ψ, Ψ) , P −1
+2
+((QΨ, Ψ) + (Ψ, QΨ))
+�
++ 1
+4ω
+�
+P −1
+3
+(Ψ, Ψ, Ψ) , p−1QΨ
+�
++ O
+�
+Ψ⊗5�
+(70)
+It seems that the cohomomorphism P does not define the field redefinition we were looking for, instead
+it relates the Witten theory to a theory defined by ˜S (Ψ). The equations of motion derived from ˜S are the
+same as for S′, namely M
+1
+1−Ψ = 0, but it is not at all obvious that the two actions agree even on-shell.
+The reason is that the HPL does not know anything about the symplectic form ω: To get an invariant
+action with S′ (Ψ′) = S (Ψ), not only m has to transform accordingly, but also ω would have to go to
+ω′ (Ψ1, Ψ2) = ω
+�
+π1P−1
+1
+1 − Ψ1
+, π1P−1
+1
+1 − Ψ2
+�
+,
+(71)
+otherwise P would fail to be cyclic. As it can be seen from (68), ˜S (Ψ) is just a fancy rewriting of the Witten
+action and therefore defines an equivalent theory. But since we would like to keep the original ω, the field
+redefinition induced by P from the HPL and giving rise to ˜S (Ψ) is not exactly what we want. However, since
+it shares the same equations of motion as S′ (Ψ), it might provide a new family of gauge-invariant observables
+for solutions of S′ (Ψ), parametrized by λ. In the last section we will check this explicitly on a special class of
+solutions. For now the next task is to derive the originally desired field redefinition which relates the actions
+S (Ψ) and S′ (Ψ). 3
+3.3
+Elements from closed string field theory
+In [18], Zwiebach and Hata have shown how to relate slightly different, consistent sets of vertices in CSFT
+via an infinitesimal field redefinition. Our strategy is now to apply their method to our problem in OSFT
+and integrate the result to the finite case. This will not only provide us the field redefinition we are looking
+for, but also give some insight into the rather abstract formalism of CSFT. First, it is useful to collect some
+basic information about the structure of CSFT.
+3It would also be an interesting direction to examine if there exists some kind of “dual” HPL that directly yields this correct
+field redefinition.
+13
+
+As already explained in the introduction, the vertices are given by integrating basic differential forms
+Ωg,n
+Ψ1Ψ2...Ψn defined by
+Ωg,n
+Ψ1Ψ2...Ψn
+�
+ˆV1, ˆV2, ..., ˆV6g−6+2n
+�
+= (2πi)−(3g−3+n) ⟨Σ|b (v1)...b (v6g−6+2n)|Ψ1⟩...|Ψn⟩.
+(72)
+They are living in the tangent space of the fibre bundle ˆPg,n over the moduli space Mg,n, with the fiber
+being the space of local coordinates around the punctures modulo phase rotations. The dimension of this
+bundle is infinite, but the degree of Ωg,n
+Ψ1Ψ2...Ψn is just the real dimension of the base space Mg,n. It takes as
+arguments tangent vectors ˆVi ∈ T ˆPg,n, which represent deformations of the world sheet Riemann surface Σ,
+either by changing the moduli or the local coordinates. The vi are Schiffer vectors on Σ, supported around
+the punctures, which generate those deformations. This means that the local coordinate around the nth
+puncture transforms as
+z(n) → z(n) + ǫv(n) �
+z(n)�
+(73)
+for some small ǫ. The b-ghost insertions are then defined as
+b (v) =
+n
+�
+i=1
+�� dzi
+2πib (zi) v(i) (zi) +
+� d¯zi
+2πi
+¯b (¯zi) ¯v(i) (¯zi)
+�
+.
+(74)
+We will be only interested in the classical action without the loop vertices, so the genus g shall be zero from
+now on. The basic forms can now be integrated over sections of ˆP0,n defining the vertices V0,n.
+The quantization procedure makes use of the Batalin-Vilkovisky formalism; although we are not interested
+in quantum effects, the BV-antibracket is used in constructing the symmetry generator. It is defined as
+{A, B} = ∂rA
+∂Ψi
+∂lB
+∂Ψ∗
+i
+− ∂rA
+∂Ψ∗
+i
+∂lB
+∂Ψi
+(75)
+where the Ψ∗
+i are antifields of opposite parity associated to each basis element of H. The BV-master action
+takes the same form as (7) with the only difference that the Ψ are not restricted in ghost number and run
+over fields as well as antifields.
+3.4
+Constructing the field redefinition
+We are looking for a non-linear field redefinition of the form
+Ψ′ = F (Ψ) =
+∞
+�
+n=1
+Fn
+�
+Ψ⊗n�
+= π1F
+1
+1 − Ψ
+(76)
+that relates the Witten action to the stubbed A∞-action in the form (66). To be consistent with the results
+of Zwiebach and Hata [18] we demand
+S (Ψ′) = S′ (Ψ) .
+(77)
+Since the kinetic term is identical we immediately find
+F1 (Ψ) = Ψ
+(78)
+In [18] it is shown that under an infinitesimal field redefinition of the form
+Ψ → Ψ + t {Ψ, e} + O
+�
+t2�
+(79)
+14
+
+the classical action transforms as
+S (Ψ) → S (Ψ) + t {S, e} + O
+�
+t2�
+,
+(80)
+where {} denotes the BV-antibracket. In the paper it is now argued that for any small change of vertices,
+the change of the action indeed takes this form and the generator e is constructed explicitly:
+e (u0) = −
+�
+n
+κn−2 1
+n!
+�
+V0,n(u0)
+Ω(0)0,n
+b(u)Ψ⊗n
+(81)
+Here we assume that there exists some family of consistent vertex sets V0,n (u) parametrized by some real
+number u and everything is evaluated at the point u0. The vector u is some Schiffer vector which generates
+a deformation of the V0,n (u0) in the direction of u, i. e. it generates diffeomorphisms which push V0,n (u0)
+into V0,n (u0 + δu). For the case of varying the stub length, this Schiffer vector takes a particular simple
+form: First, lets notice that the stub length λ for closed strings is defined as the geodesic distance from the
+location | z |= 1 of the local coordinate to the begin of the semiinfinite cylinder associated with the puncture.
+This implies that λ can be changed by just rescaling the coordinate: Sending z to z′ = z + ǫz, the location
+of | z′ |= 1 corresponds to | z |= 1 − ǫ, such that λ is increased by ǫ. By comparing with Eq. (73) we read off
+u(i) = z(i).
+(82)
+The b-ghost insertion is then given by
+b (u) =
+n
+�
+i=1
+�� dzi
+2πizib (zi) +
+� d¯zi
+2πi ¯zi¯b (¯zi)
+�
+=
+n
+�
+i=1
+b(i)
+0 + b
+(i)
+0
+=
+n
+�
+i=1
+b+(i)
+0
+(83)
+We want to use the above expression for e in the context of changing the stub length for open strings, hence
+a few modifications and simplifications are necessary: First, the combinatorial factor n! originates from total
+symmetrization of the vertices and is not necessary for open strings. Second, the insertions of b+
+0 should get
+replaced simply by b0 since there is no antiholomorphic sector. Moreover, the string coupling κ will be set
+to one. Now the generator simplifies to
+e (λ) = −
+�
+n
+�
+V(λ)0,n
+Ω0,n
+b0Ψ⊗n.
+(84)
+If we make the ansatz
+Ψ′ = Ψ + δλ
+∞
+�
+n=2
+fn
+�
+Ψ⊗n�
+(85)
+as the infinitesimal version of (76), then the fn are determined as
+fn
+�
+Ψ⊗n�
+=
+�
+Ψ, e(n)�
+(86)
+where δλ plays the role of t in (79).
+To find f2 (Ψ, Ψ) we need to consider e (λ) for n = 3: The vertex V (λ)0,3 is zero dimensional, so there is
+no integral and the surface state ⟨Σ| is just the Witten vertex with stubs of length λ,
+⟨V3 (λ) | = ω (·, M2 (·, ·)) = ω
+�
+e−λL0·, e−λL0 · ∗e−λL0·
+�
+(87)
+15
+
+Inserting into (84) yields
+e(3) (λ, Ψ) = − (ω (b0Ψ, M2 (Ψ, Ψ)) + ω (Ψ, M2 (b0Ψ, Ψ)) + ω (Ψ, M2 (Ψ, b0Ψ))) .
+(88)
+The BV-bracket with Ψ can be straightforwardly evaluated; after carefully checking the signs the result is
+f2 (Ψ, Ψ) = −b0M2 (Ψ, Ψ) + M2 (b0Ψ, Ψ) + M2 (Ψ, b0Ψ) .
+(89)
+We expect now the relation
+S′ (Ψ + δλf2 (Ψ, Ψ) , λ) = S′ (Ψ, λ + δλ)
+(90)
+to hold up to order 3 in Ψ; by directly inserting we can compute explicitly
+S′ (Ψ + δλf2 (Ψ, Ψ) , λ) = 1
+2ω (Ψ, QΨ) + 1
+3ω (Ψ, M2 (Ψ, Ψ)) + δλ ω (Ψ, Qf2 (Ψ, Ψ)) + O
+�
+Ψ⊗4�
+(91)
+The last and most interesting term yields
+(−ω (Ψ, Qb0M2 (Ψ, Ψ)) + ω (Ψ, QM2 (b0Ψ, Ψ)) + ω (Ψ, QM2 (Ψ, b0Ψ))) δλ
+= (−ω (Ψ, L0M2 (Ψ, Ψ)) + ω (Ψ, b0QM2 (Ψ, Ψ)) − ω (QΨ, M2 (b0Ψ, Ψ)) − ω (QΨ, M2 (Ψ, b0Ψ))) δλ
+= − ω (Ψ, L0M2 (Ψ, Ψ)) δλ
+(92)
+where the last three terms in the second line cancel after applying the Leibniz rule and cyclicity. On the
+other hand,
+S′ (Ψ, λ + δλ) = S′ (Ψ, λ) + δλ d
+dλS′ (Ψ, λ) = 1
+2ω (Ψ, QΨ) + 1
+3ω (Ψ, M2 (Ψ, Ψ))
+− 1
+3δλ ω (L0Ψ, M2 (Ψ, Ψ)) − 1
+3δλ ω (Ψ, M2 (L0Ψ, Ψ)) − 1
+3δλ ω (Ψ, M2 (Ψ, L0Ψ)) + O
+�
+Ψ⊗4�
+= 1
+2ω (Ψ, QΨ) + 1
+3 ω (Ψ, M2 (Ψ, Ψ)) − δλ ω (L0Ψ, M2 (Ψ, Ψ)) + O
+�
+Ψ⊗4�
+(93)
+which is the same expression up to order Ψ⊗3 (Again, cyclicity was used in the last line.).
+The explicit form of f2 (Ψ, Ψ) suggests the following general structure: We can guess the ansatz
+fn
+�
+Ψ⊗n�
+= −b0Mn
+�
+Ψ⊗n�
++ Mn
+�
+b0
+�
+Ψ⊗n��
+(94)
+where b0 again denotes the coderivation associated to b0. At first sight, Eq.( 94 ) looks a bit strange now from
+the coalgebra perspective because it is a commutator of two odd objects, so one would more naturally expect
+an anticommutator. However, the first term in (94) stems from the application of b0 on the first argument
+of the symplectic form ω, hence the sign contains implicit information about ω. From the discussion about
+P from the HPL we could anticipate that ω has to enter the calculation at some point. The more natural
+looking expression π1 [b0, Mn] would have been independent of ω.
+One can prove now that the ansatz ( 94 ) is indeed correct by directly inserting into the action.
+3.5
+Proof of the ansatz for the infinitesimal field redefinition
+If we focus solely on terms of order n + 1 in Ψ we get
+S′(n+1) (Ψ′) = ω
+�
+Ψ, Qfn
+�
+Ψ⊗n��
+δλ+
+n−1
+�
+k=2
+ω
+�
+Ψ, Mk
+�
+fn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+δλ+
+1
+n + 1ω
+�
+Ψ, Mn
+�
+Ψ⊗n��
+(95)
+16
+
+The last term is the contribution from the original S (Ψ), so the first two terms denoted by δS(n+1) should
+yield the infinitesimal variation
+δS′(n+1) ?= δλ d
+dλ
+1
+n + 1ω
+�
+Ψ, Mn
+�
+Ψ⊗n��
+(96)
+Inserting the ansatz and performing some straight forward manipulations gives
+δS′(n+1)
+δλ
+= ω
+�
+Ψ, Q (−b0Mn + Mnb0)
+�
+Ψ⊗n��
++
+n−1
+�
+k=2
+ω
+�
+Ψ, Mk
+�
+(−b0Mn+1−k + Mn+1−kb0)
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+= − ω
+�
+Ψ, L0Mn
+�
+Ψ⊗n��
++ ω
+�
+Ψ, b0 [Q, Mn]
+�
+Ψ⊗n��
+− ω
+�
+Ψ, b0MnQ
+�
+Ψ⊗n��
++ ω
+�
+Ψ, QMnb0
+�
+Ψ⊗n��
++
+n−1
+�
+k=2
+�
+−ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
++ ω
+�
+Ψ, Mk
+�
+Mn+1−kb0
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1���
+(97)
+The last two terms of the second line actually cancel each other:
+− ω
+�
+Ψ, b0MnQ
+�
+Ψ⊗n��
++ ω
+�
+Ψ, QMnb0
+�
+Ψ⊗n��
+= − ω
+�
+b0Ψ, MnQ
+�
+Ψ⊗n��
+− ω
+�
+QΨ, Mnb0
+�
+Ψ⊗n��
+= − ω (b0Ψ, Mn (QΨ, Ψ, ..., Ψ)) − ω (b0Ψ, Mn (Ψ, QΨ, ..., Ψ)) − ω (b0Ψ, Mn (Ψ, Ψ, ..., QΨ))
+− ω (QΨ, Mn (b0Ψ, Ψ, ..., Ψ)) − ω (QΨ, Mn (Ψ, b0Ψ, ..., Ψ)) − ω (QΨ, Mn (Ψ, Ψ, ..., b0Ψ))
+(98)
+Because of cyclicity of the (n + 1)-vertex the last two lines contain of the same terms, just differing by a sign
+which comes from commuting Q with b0. Therefore they add to zero. The second term in of the second line
+of (97) can be further manipulated using the A∞-relations:
+δS′(n+1)
+δλ
+= − ω
+�
+Ψ, L0Mn
+�
+Ψ⊗n��
+−
+n−1
+�
+k=2
+ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
++
+n−1
+�
+k=2
+�
+ω
+�
+Ψ, Mk
+�
+Mn+1−k
+�
+b0
+�
+Ψ⊗n+1−k��
+, Ψ⊗k−1��
+− ω
+�
+Ψ, b0MkMn+1−k
+�
+Ψ⊗n���
+(99)
+Now again, the terms in the last line cancel after using cyclicity:
+ω
+�
+Ψ, b0MkMn+1−k
+�
+Ψ⊗n��
+= ω
+�
+b0Ψ, MkMn+1−k
+�
+Ψ⊗n��
+= ω
+�
+Mk
+�
+b0Ψ, Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−2�
+, Ψ
+�
++ ω
+�
+Mk
+�
+b0Ψ, Ψ, Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−3�
+, Ψ
+�
++ ... + ω
+�
+Mk
+�
+b0Ψ, Ψ⊗k−2, Mn+1−k
+�
+Ψ⊗n+1−k��
+, Ψ
+�
++ ω
+�
+Mk
+�
+b0Ψ, Ψ⊗k−1�
+, Mn+1−k
+�
+Ψ⊗n+1−k��
+(100)
+All terms except for the last one can be further manipulated using the antisymmetry of ω and cyclicity of
+Mk. For example,
+ω
+�
+Mk
+�
+b0Ψ, Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−2�
+, Ψ
+�
+= −ω
+�
+Ψ, Mk
+�
+b0Ψ, Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−2��
+= ω
+�
+Mk
+�
+Ψ, b0Ψ, Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−3�
+, Ψ
+�
+= ... = ω
+�
+Mk
+�
+Ψ⊗k−1, b0Ψ
+�
+, Mn+1−k
+�
+Ψ⊗n+1−k��
+(101)
+17
+
+so in all of the terms the Mn+1−k can be moved to the outermost right. After all, the terms can be summed
+up as
+ω
+�
+Mk
+�
+Ψ⊗k−2, b0Ψ, Mn+1−k
+�
+Ψ⊗n+1−k���
++ ω
+�
+Mk
+�
+Ψ⊗k−3, b0Ψ, Ψ, Mn+1−k
+�
+Ψ⊗n+1−k���
++ ... + ω
+�
+Mk
+�
+b0Ψ, Ψ⊗k−1�
+, Mn+1−k
+�
+Ψ⊗n+1−k��
+= ω
+�
+Mk
+�
+b0
+�
+Ψ⊗k��
+, Mn+1−k
+�
+Ψ⊗n+1−k��
+= ω
+�
+Ψ, Mn+1−k
+�
+Mk
+�
+b0
+�
+Ψ⊗k��
+, Ψ⊗n−k��
+(102)
+which is after the summation over k identical to the first term in the second line of (99), just with opposite
+sign. So we arrive at the expression
+δS′(n+1)
+δλ
+= −ω
+�
+Ψ, L0Mn
+�
+Ψ⊗n��
+−
+n−1
+�
+k=2
+ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+(103)
+which should now be compared to the result of formula (96).
+The derivative with respect to λ can act on the stubs as well as on the homotopy h. The action on e−λL0
+inserts a factor of −L0 on every input string field of the (n + 1)-vertex. Since the vertices are cyclically
+symmetric, we get n + 1 identical terms, which cancels the prefactor
+1
+n+1. The result is
+d
+dλ
+1
+n + 1ω
+�
+Ψ, Mn
+�
+Ψ⊗n��
+⊃ −ω
+�
+L0Ψ, Mn
+�
+Ψ⊗n��
+(104)
+which is equal to the first term of (103). To compute the action on h, the tree representation turns out to
+be useful again: First of all,
+d
+dλh = −2b0e−2λL0
+(105)
+hence we get a sum of all possible tree diagrams with one propagator replaced by −2b0e−2λL0. We can
+cut through the diagram along this replaced propagator and think of the factor e−2λL0 as arising from two
+e−λL0-stubs from the leaf and the root of the two subtrees. Both subtrees are now part of a higher product
+Mk for some k, 2 ≤ k ≤ n − 1. So the whole expression can be written as a combination of two higher
+products Mk, Mn+1−k with a factor −2b0 inserted:
+d
+dλ
+1
+n + 1ω
+�
+Ψ, Mn
+�
+Ψ⊗n��
+⊃ −
+1
+n + 1
+n−1
+�
+k=2
+ω
+�
+Ψ, Mk
+�
+2b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
++
+ω
+�
+Ψ, Mk
+�
+Ψ, 2b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−2��
++ ...+
+ω
+�
+Ψ, Mk
+�
+Ψ⊗k−1, 2b0Mn+1−k
+�
+Ψ⊗n+1−k���
+(106)
+Because of cyclicity of the k + 1-vertex, the different lines contain the same terms so we have
+d
+dλ
+1
+n + 1ω
+�
+Ψ, Mn
+�
+Ψ⊗n��
+⊃ −
+1
+n + 1
+n−1
+�
+k=2
+2k · ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+(107)
+The last bracket can be further manipulated:
+ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+= −ω
+�
+Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1�
+, Ψ
+�
+= ω
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Mk
+�
+Ψ⊗k��
+= −ω
+�
+Mn+1−k
+�
+Ψ⊗n+1−k�
+, b0Mk
+�
+Ψ⊗k��
+= ω
+�
+Ψ, Mn+1−k
+�
+Ψ⊗n−k, b0Mk
+�
+Ψ⊗k���
+= ω
+�
+Ψ, Mn+1−k
+�
+b0Mk
+�
+Ψ⊗k�
+, Ψ⊗n−k��
+(108)
+18
+
+In the last step cyclicity of the (n + 1 − k)-vertex was used again. We see that in the sum of (107) the kth
+term and the (n + 1 − k)th term are identical so the sum can be rewritten as
+−
+1
+n + 1
+n−1
+�
+k=2
+(n + 1) · ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+=
+n−1
+�
+k=2
+ω
+�
+Ψ, Mk
+�
+b0Mn+1−k
+�
+Ψ⊗n+1−k�
+, Ψ⊗k−1��
+(109)
+which is precisely the second term in (103).
+This completes the proof that the ansatz
+fn
+�
+Ψ⊗n�
+= −b0Mn
+�
+Ψ⊗n�
++ Mn
+�
+b0
+�
+Ψ⊗n��
+(110)
+indeed yields the correct infinitesimal field redefinition.
+3.6
+Finite field redefinition
+So far we have only been concerned with infinitesimal variations of λ, now we want to generalize the results
+to finite changes. We know
+Ψλ+δλ = Ψλ + δλf λ
+2 (Ψλ, Ψλ) + δλf λ
+3 (Ψλ, Ψλ, Ψλ) + ... = Ψλ + δλ d
+dλΨλ
+(111)
+where we have written the superscript λ to indicate that the fn also depend on λ explicitly. This equation
+can be integrated to
+Ψλ = Ψ0 +
+� λ
+0
+dtf t
+2 (Ψt, Ψt) +
+� λ
+0
+dtf t
+3 (Ψt, Ψt, Ψt) + ...
+(112)
+Inserting Ψλ back we get a perturbative expansion in the original solution Ψ0 :
+Ψλ =Ψ0 +
+� λ
+0
+dtf t
+2 (Ψ0, Ψ0) +
+� λ
+0
+dtf t
+3 (Ψ0, Ψ0, Ψ0) +
+� λ
+0
+dtf t
+2
+�� t
+0
+dsf s
+2 (Ψ0, Ψ0) , Ψ0
+�
++
+� λ
+0
+dtf t
+2
+�
+Ψ0,
+� t
+0
+dsf s
+2 (Ψ0, Ψ0)
+�
++ O
+�
+Ψ⊗4
+0
+�
+(113)
+This formula provides an algorithm to find the associated A∞-solution to each known solution of the Witten
+OSFT.
+4
+Physical interpretation
+To summarize, we found two distinct field redefinitions
+˜Ψ =
+∞
+�
+n=1
+Pn
+�
+Ψ⊗n�
+,
+Ψ′ =
+∞
+�
+n=1
+Fn
+�
+Ψ⊗n�
+(114)
+19
+
+which generate two different actions ˜S and S′ via
+˜S
+�
+˜Ψ
+�
+= S (Ψ) ,
+S′ (Ψ) = S (Ψ′) .
+(115)
+However, both actions share the same equations of motion, namely the Maurer-Cartan equation of the stubbed
+A∞-algebra
+π1M
+1
+1 − Ψ = 0.
+(116)
+It remains to examine what the physical meaning of those two actions is, most importantly, if they yield the
+same on-shell value for a given solution Ψ∗.
+In [19, 20], a special class of solutions containing a nearly marginal vertex operator is introduced which
+serves as a useful playground to analyze this question.
+4.1
+Nearly marginal solutions
+Consider a matter conformal primary field V with weight h smaller but very close to one. The string field
+Ψ1 = µ · cV (0) |0⟩,
+(117)
+with some real coupling constant µ obeys the Siegel gauge condition
+b0Ψ1 = 0
+(118)
+and will serve as a starting point to find the full solution Ψ = � Ψn as a perturbative series in the expansion
+parameter y = 1 − h. One can solve for the string coupling µ using the Witten equations of motion
+QΨ + Ψ ∗ Ψ = 0
+(119)
+to obtain [20]
+µ =
+y
+CV V V
++ O
+�
+y3�
+,
+(120)
+where CV V V denotes the three-point function constant of V . We can deduce that Ψ1 = O (y) and from the
+perturbative algorithm for the full solution one can also show that in general Ψn = O (yn).
+The on-shell action in Witten theory can be written compactly as
+S (Ψ) = −1
+6 ⟨ Ψ, QΨ ⟩ .
+(121)
+From
+QΨ1 = µy · c∂cV (0) |0⟩
+(122)
+we see that the action will be of leading order y3 and given by
+S (Ψ) = −1
+6
+y3
+C2
+V V V
+⟨ cV, c∂cV ⟩ = 1
+6
+y3
+C2
+V V V
++ O
+�
+y4�
+(123)
+if we assume that V is conveniently normalized, ⟨ V (z1) , V (z2) ⟩ = z−2h
+12 .
+20
+
+4.2
+The A∞-action
+The first important observation is that the cohomomorphism P simplifies significantly for string fields in
+Siegel gauge: Since h is proportional to b0, it annihilates Ψ and (63) collapses to
+˜Ψ =
+∞
+�
+n=1
+Pn
+�
+Ψ⊗n�
+= pΨ.
+(124)
+We already know that ˜S
+�
+˜Ψ
+�
+yields the original value S (Ψ), so now we want check the expression S′ �
+˜Ψ
+�
+:
+Ψ1 is an L0-eigenstate so we straightforwardly get
+pΨ1 = e−λL0Ψ1 = eλyΨ1.
+(125)
+For cubic order in y we just have to insert this into the kinetic term (121) and get
+S′ �
+˜Ψ
+�
+|y3= −1
+6 ⟨ pΨ1, QpΨ1 ⟩ = 1
+6
+y3
+C2
+V V V
+e2λy |y3= 1
+6
+y3
+C2
+V V V
+(126)
+which agrees with the result above up to terms of O
+�
+y4�
+.
+For a more non-trivial check we can collect the terms of quartic order in y: In the action we have to
+consider the first three terms
+S′ �
+˜Ψ
+�
+|y4=
+�
+−1
+2
+�
+˜Ψ, Q˜Ψ
+�
+− 1
+3
+�
+˜Ψ, M2
+�
+˜Ψ, ˜Ψ
+� �
+− 1
+4
+�
+˜Ψ, M3
+�
+˜Ψ, ˜Ψ, ˜Ψ
+� ��
+|y4 .
+(127)
+The equations of motion however tell us that
+�
+˜Ψ, Q˜Ψ
+�
++
+�
+˜Ψ, M2
+�
+˜Ψ, ˜Ψ
+� �
++
+�
+˜Ψ, M3
+�
+˜Ψ, ˜Ψ, ˜Ψ
+� �
+= O
+�
+y5�
+,
+(128)
+so the expression simplifies to
+S′ �
+˜Ψ
+�
+|y4=
+�
+−1
+4
+�
+˜Ψ, Q˜Ψ
+�
+− 1
+12
+�
+˜Ψ, M2
+�
+˜Ψ, ˜Ψ
+� ��
+|y4 .
+(129)
+Plugging in ˜Ψ = eλyΨ1 and isolating y4-terms yields
+S′ �
+˜Ψ
+�
+|y4=
+�
+−1
+4
+�
+eλyΨ1, QeλyΨ1
+�
+− 1
+12
+�
+eλyΨ1, e−λL0 �
+e−λL0eλyΨ1 ∗ e−λL0eλyΨ1
+� ��
+|y4
+=
+�
+−1
+4e2λy ⟨ Ψ1, QΨ1 ⟩ − 1
+12e6λy ⟨ Ψ1, (Ψ1 ∗ Ψ1) ⟩
+�
+|y4
+= − 1
+2
+λy4
+C2
+V V V
+⟨ cV, c∂cV ⟩ − 1
+2
+λy4
+C3
+V V V
+⟨ cV, (cV ∗ cV ) ⟩ .
+(130)
+The correlation functions can be calculated by using standard CFT methods, see e.g. [21, 20]; the result is
+⟨ cV, c∂cV ⟩ = −1,
+⟨ cV, (cV ∗ cV ) ⟩ = CV V V
+�
+3
+√
+3
+4
+�3y
+= CV V V
+�
+1 + 3y · ln
+�
+3
+√
+3
+4
+��
++O
+�
+y2�
+. (131)
+21
+
+We see by inserting into (130) that S′ �
+˜Ψ
+�
+|y4 indeed vanishes and the value of the on-shell action is the
+same as the original S (Ψ) to order y4. In principle, terms containing Ψ2 ∝ y2 also contribute at this order.
+However, since any appearance of λ automatically comes with a factor y, there are no terms of order y4
+containing λ as well as Ψ2. All Ψ2-contributions are just the ones already present in S (Ψ) and were studied
+in detail in [20].
+We see that to the first two leading orders, the actions S′ �
+˜Ψ
+�
+and ˜S
+�
+˜Ψ
+�
+= S (Ψ) give the same on-shell
+result. Since S′ �
+˜Ψ
+�
+= S
+�
+˜Ψ′�
+, where
+˜Ψ′ =
+∞
+�
+n=1
+Fn
+�
+˜Ψ⊗n�
+= π1FP
+1
+1 − Ψ,
+(132)
+this suggests that the combination FP gives rise to a gauge transformation rather than a physically distinct
+solution. However, a full proof of this statement will be left for future publications.
+5
+Conclusion and outlook
+We succeeded in providing an explicit consistent description of OSFT with stubs and interestingly found
+two possible actions with the same equations of motion. The field redefinitions used to convert solutions of
+Witten OSFT to the new theory are given in explicit form. We hope that the analysis of solutions to the
+stubbed equations of motion can teach us more general properties of solutions to Maurer-Cartan equations
+for A∞- or L∞-algebras, in particular about the solutions of closed string field theory. One way to proceed
+would be to transform the whole construction to the sliver frame, where many analytic solutions of OSFT
+are formulated.
+Another possible future direction is to examine wether the stubbed theory is “more well-behaved” in the
+sense that some typical singularities and ambiguities, for example connected to identity-like solutions, are
+ameliorated.
+Acknowledgements
+We thank Ted Erler, Jakub Vošmera, Branislav Jurčo, Igor Khavkine and Martin Markl for useful discussions.
+Our work has been funded by the Grant Agency of Czech Republic under the grant EXPRO 20-25775X.
+Appendix
+Tensor coalgebras
+The tensor coalgebra T V associated to a (graded) vector space V is defined as the Fock space
+V ⊗0 + V ⊗1 + V ⊗2 + ...
+(133)
+together with the comultiplication ∆ : T V → T V ⊗′ T V given by
+∆ (v1 ⊗ ... ⊗ vn) =
+n
+�
+k=0
+(v1 ⊗ ... ⊗ vk) ⊗′ (vk+1 ⊗ ... ⊗ vn)
+(134)
+22
+
+on homogeneous elements and extended by linearity. Here the vi are elements of V and ⊗′ denotes the
+tensor product arising from a comultiplication, in contrast to the usual ⊗. We define the projection operator
+πn : T V → T V to project any element on its nth tensor power component,
+πnT V = V ⊗n
+(135)
+A linear map d : T V → T V is called a coderivation if it satisfies the co-Leibniz rule:
+∆d = (d ⊗′ 1 + 1 ⊗′ d) ∆
+(136)
+Linear combinations of coderivations are again coderivations as well as their graded commutator
+[d1, d2] = d1d2 − (−1)deg(d1)deg(d2) d2d1.
+(137)
+The product d1d2 is in general not a coderivation. For any m-linear map dm : V ⊗m → V one can construct
+an associated coderivation by the formula
+d =
+∞
+�
+n=1
+n−m
+�
+k=0
+1⊗k ⊗ dm ⊗ 1⊗n−k−m.
+(138)
+The co-Leibniz rule guarantees that any coderivation is a sum of terms of this form for different m. The
+individual m-products can be recovered as
+dm = π1dπm
+(139)
+If an odd coderivation d obeys
+d2 = 0
+(140)
+then its components dm form an A∞-algebra.
+A linear map f is called a cohomomorphism if it fulfills
+∆f = (f ⊗′ f) ∆.
+(141)
+Linear combinations and products of cohomomorphisms are again cohomomorphisms. Given a family of
+m-products fm one can construct a unique cohomomorphism via
+f =
+∞
+�
+j=1
+∞
+�
+k=1
+�
+m1+...+mj=k
+fm1 ⊗ ... ⊗ fmj.
+(142)
+Again, the individual products can be recovered from f as
+fm = π1fπm
+(143)
+Of special importance are elements of T V of the form
+1 + v + v ⊗ v + v ⊗ v ⊗ v + ... =:
+1
+1 − v
+(144)
+for some v ∈ V. They fulfill the following useful properties:
+π1f
+1
+1 − v =
+∞
+�
+m=1
+fm
+�
+v⊗m�
+(145)
+23
+
+f
+1
+1 − v =
+1
+1 − π1f
+1
+1−v
+(146)
+for any cohomomorphism f.
+A bilinear map ⟨ω|: T V × T V → C is called a symplectic form if it satisfies
+⟨ω|v1 ⊗ v2 =: ω (v1, v2) = − (−1)deg(v1)deg(v2) ω (v2, v1) .
+(147)
+A multilinear product mk is called cyclic with respect to ω if it fulfills
+ω (v1, mk (v2, ..., vk+1)) = − (−1)deg(v1)deg(mk) ω (mk (v1, ..., vk) , vk+1)
+(148)
+A coderivation d is cyclic if all of its components dm = π1dπm are cyclic or equivalently
+⟨ω|π2d = 0.
+(149)
+Given two symplectic forms ⟨ω|, ⟨ω′|, a cohomomorphism f is cyclic if
+⟨ω′|π2f = ⟨ω|π2
+(150)
+References
+[1] E. Witten, “Noncommutative Geometry and String Field Theory,” Nucl. Phys. B 268 (1986), 253-294
+doi:10.1016/0550-3213(86)90155-0
+[2] J. Vošmera, “Selected topics in string field theory and physics of D-branes,”
+[3] B. Zwiebach, “Closed string field theory: Quantum action and the B-V master equation,” Nucl. Phys. B
+390 (1993), 33-152 doi:10.1016/0550-3213(93)90388-6 [arXiv:hep-th/9206084 [hep-th]].
+[4] B. Zwiebach, “A Proof that Witten’s open string theory gives a single cover of moduli space,” Commun.
+Math. Phys. 142 (1991), 193-216 doi:10.1007/BF02099176
+[5] H. Sonoda and B. Zwiebach, “COVARIANT CLOSED STRING THEORY CANNOT BE CUBIC,” Nucl.
+Phys. B 336 (1990), 185-221 doi:10.1016/0550-3213(90)90108-P
+[6] M. Schnabl, “Analytic solution for tachyon condensation in open string field theory,” Adv. Theor. Math.
+Phys. 10 (2006) no.4, 433-501 doi:10.4310/ATMP.2006.v10.n4.a1 [arXiv:hep-th/0511286 [hep-th]].
+[7] T.
+Erler,
+“Four
+Lectures
+on
+Closed
+String
+Field
+Theory,”
+Phys.
+Rept.
+851
+(2020),
+1-36
+doi:10.1016/j.physrep.2020.01.003 [arXiv:1905.06785 [hep-th]].
+[8] T.
+Erler,
+“The
+closed
+string
+field
+theory
+action
+vanishes,”
+JHEP
+10
+(2022),
+055
+doi:10.1007/JHEP10(2022)055 [arXiv:2204.12863 [hep-th]].
+[9] T. Takahashi and S. Tanimoto, “Marginal and scalar solutions in cubic open string field theory,” JHEP
+03 (2002), 033 doi:10.1088/1126-6708/2002/03/033 [arXiv:hep-th/0202133 [hep-th]].
+[10] T. Erler, “Four lectures on analytic solutions in open string field theory,” Phys. Rept. 980 (2022), 1-95
+doi:10.1016/j.physrep.2022.06.004 [arXiv:1912.00521 [hep-th]].
+24
+
+[11] B. Zwiebach, “Oriented open - closed string theory revisited,” Annals Phys. 267 (1998), 193-248
+doi:10.1006/aphy.1998.5803 [arXiv:hep-th/9705241 [hep-th]].
+[12] T. Takezaki, “Open superstring field theory including the Ramond sector based on the supermoduli
+space,” [arXiv:1901.02176 [hep-th]].
+[13] H. Erbin, C. Maccaferri, M. Schnabl and J. Vošmera, “Classical algebraic structures in string theory
+effective actions,” JHEP 11 (2020), 123 doi:10.1007/JHEP11(2020)123 [arXiv:2006.16270 [hep-th]].
+[14] M. Crainic; “On the perturbation lemma, and deformations,” [arXiv:math/0403266]
+[15] S.
+Konopka,
+“The
+S-Matrix
+of
+superstring
+field
+theory,”
+JHEP
+11
+(2015),
+187
+doi:10.1007/JHEP11(2015)187 [arXiv:1507.08250 [hep-th]].
+[16] J. L. Loday, B. Vallette; “Algebraic operads,” Springer Berlin, Heidelberg, 2012 doi:10.1007/978-3-642-
+30362-3
+[17] S. Li and K. Zeng, “Homotopy Algebras in Higher Spin Theory,” Adv. Theor. Math. Phys. 24 (2020)
+no.3, 757-819 doi:10.4310/ATMP.2020.v24.n3.a5 [arXiv:1807.06037 [hep-th]].
+[18] H. Hata and B. Zwiebach, “Developing the covariant Batalin-Vilkovisky approach to string theory,”
+Annals Phys. 229 (1994), 177-216 doi:10.1006/aphy.1994.1006 [arXiv:hep-th/9301097 [hep-th]].
+[19] S. Mukherji and A. Sen, “Some all order classical solutions in nonpolynomial closed string field theory,”
+Nucl. Phys. B 363 (1991), 639-664 doi:10.1016/0550-3213(91)80037-M
+[20] J. Scheinpflug and M. Schnabl, “Conformal perturbation theory from open string field theory,”
+[arXiv:2301.05216 [hep-th]].
+[21] Y. Okawa, “Analytic methods in open string field theory,” Prog. Theor. Phys. 128 (2012), 1001-1060
+doi:10.1143/PTP.128.1001
+25
+
diff --git a/jtFPT4oBgHgl3EQf1TVB/content/tmp_files/load_file.txt b/jtFPT4oBgHgl3EQf1TVB/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..3a4764d1f7d1231a130d0861446897ee04c1fa1b
--- /dev/null
+++ b/jtFPT4oBgHgl3EQf1TVB/content/tmp_files/load_file.txt
@@ -0,0 +1,1217 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf,len=1216
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='13182v1 [hep-th] 30 Jan 2023 Open string field theory with stubs Martin Schnabl and Georg Stettinger January 31,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2023 CEICO,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Institute of Physics of the Czech Academy of Sciences,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Na Slovance 2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 182 00 Prague 8,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Czech Republic Abstract There are various reasons why adding stubs to the vertices of open string field theory (OSFT) is interesting: Not only the stubs can tame certain singularities and make the theory more well-behaved,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' but also the new theory shares a lot of similarities with closed string field theory,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' which helps to improve our understanding of its structure and possible solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In this paper we explore two natural ways of implementing stubs into the framework of OSFT, resulting in an A∞-algebra giving rise to infinitely many vertices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We find two distinct consistent actions, both generated by a field redefinition, interestingly sharing the same equations of motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the last section we illustrate their relationship and physical meaning by applying our construction to nearly marginal solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Contents 1 Introduction and motivation 2 2 Deforming the Witten action using stubs 4 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 Elements of homological perturbation theory .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 Transferring algebraic structure: SDR-case .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='3 Transferring algebraic structure: Non-SDR-case .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 6 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4 Higher products using tree diagrams .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 8 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='5 Proof of cyclicity to all orders .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 9 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='6 Geometric picture .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 10 3 Analytic solutions and action(s) 10 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 Projection cohomomorphism from the HPL .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 10 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 The action .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 12 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='3 Elements from closed string field theory .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 13 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4 Constructing the field redefinition .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 14 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='5 Proof of the ansatz for the infinitesimal field redefinition .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 16 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='6 Finite field redefinition .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 19 1 4 Physical interpretation 19 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 Nearly marginal solutions .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 20 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 The A∞-action .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 21 5 Conclusion and outlook 22 1 Introduction and motivation It is a well-known fact that the algebraic and geometric structures of open and closed string field theory are fundamentally different.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The OSFT action (Witten action) consists of a standard kinetic term and a single cubic interaction and reads [1] S (Ψ) = 1 2ω (Ψ, QΨ) + 1 3ω (Ψ, Ψ ∗ Ψ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (1) The algebraic ingredients are a nilpotent differential given by the BRST-operator, the star product and a symplectic form ω, which all together form a cyclic differential graded algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the following, We will use the coalgebra notation of [2] with m2 (Ψ1, Ψ2) := (−)d(Ψ1) Ψ1 ∗ Ψ2 and the shifted degree given by d (Ψ) = gh (Ψ) + 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Now the defining algebraic properties are Q2 = 0, (2) Qm2 (Ψ1, Ψ2) = −m2 (QΨ1, Ψ2) − (−)d(Ψ1) m2 (Ψ1, QΨ2) , (3) m2 (Ψ1, m2 (Ψ2, Ψ3)) = − (−)d(Ψ1) m2 (m2 (Ψ1, Ψ2) , Ψ3) , (4) ω (Ψ1, m2 (Ψ2, Ψ3)) = − (−)d(Ψ1) ω (m2 (Ψ1, Ψ2) , Ψ3) , (5) ω (Ψ1, QΨ2) = − (−)d(Ψ1) ω (QΨ1, Ψ2) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (6) In contrast, the classical closed string field theory (CSFT) action [3] contains infinitely many vertices and reads S (Ψ) = 1 2ω (Ψ, QΨ) + ∞ � n=3 κn−2 n!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' {Ψn} , (7) where the vertices {Ψn} = � V0,n Ω(0)0,n Ψn (8) are given by integrating certain differential forms over vertex regions V0,n in the moduli space of n-punctured Riemann surfaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' At quantum level there are also vertices of higher genus to consider, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' intrinsic loops already present within the elementary vertices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, they will not be of interest here for now.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The vertices also can be decomposed into a symplectic form ω and higher products ln: � Ψn+1� = ω (Ψ, ln (Ψ, Ψ, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψ)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (9) In contrast to the open string star product, the ln are all totally symmetric.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The objects Q =: l1, all higher ln and ω give rise to a cyclic L∞-algebra: lnl1 (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn) + ln−1l2 (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn) + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' + l1ln (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn) = 0, ∀n (10) ω (Ψ1, ln (Ψ2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn+1)) = − (−)d(Ψ1) ω (ln (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn) , Ψn+1) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ∀n (11) 2 The multiplication of the multilinear products in the first line is defined as lkll (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψk+l−1) = � σ (−1)ǫ(σ) l!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (k − 1)!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='lk � ll � Ψσ(1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψσ(l) � , Ψσ(l+1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψσ(k+l−1) � , (12) where the sum runs over all permutations of k + l − 1 elements and the factor (−1)ǫ(σ) is just the sign picked up when permuting the Ψs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The different algebraic structures have a geometric origin: For a unitary theory it is necessary that the Feynman string diagrams computed from the action cover the full moduli space of n-punctured Riemann surfaces exactly once.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' For the open string where one needs to consider surfaces with boundaries it has been shown by Zwiebach [4] that this is indeed the case for the action (1) with only one cubic interaction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In contrast, for the closed string, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' surfaces without boundaries, such a simple cubic representation of the action seems not to be possible, see [5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The simplicity of the OSFT action allowed for the discovery of analytic classical solutions, most impor- tantly the open string tachyon vacuum [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In classical CSFT, no such analytic solutions have been found yet [7, 8].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' To make some effort into this direction, we want to propose the following strategy: If CSFT cannot be simplified easily, maybe we can make OSFT “more complicated” in the sense that its algebraic and geometric structure resembles that of CSFT?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' There are two motivations for that: First, it could help to make the very abstract language of CSFT more intuitive and tractable.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Second, if we can translate the known analytic solutions to the new, deformed OSFT, it could give an idea how CSFT solutions might look like.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In total there are three steps to do: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Find a consistent deformation of the Witten OSFT such that its mathematical structure resembles CSFT 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Find analytic solutions of this deformed theory 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Make an educated guess for analytic solutions of CSFT The first two tasks will be worked out in this paper while the third one is left for the future.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The deformation we will consider has many additional interesting aspects: First, OSFT exhibits some issues with singularities or ill-defined quantities: For example, there exist “identity-based” solutions [9] (for a recent discussion see [10]) which solve the equations of motion but do not have a well defined action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' There are hints that these solutions could be better behaved in the deformed theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Moreover, the deformed products M2, M3, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' defined in the next section could allow for a natural definition of a Banach type A∞-algebra over the space of string fields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The paper is organized as follows: In the second section the deformation using stubs [11] is discussed in detail, resulting in an explicit definition of the higher products of the A∞-algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The necessary mathemat- ical ingredients, namely homological perturbation theory and homotopy transfer, are introduced as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' As a side result, we give a consistent method of applying homotopy transfer to general homotopy equivalences which are not deformation retracts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Section three deals with analytical solutions and the action of the stubbed theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Surprisingly, we find two consistent actions with the same equations of motion, both generated by a field redefinition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The two field redefinitions, one coming from homological perturbation theory and one motivated by closed string field theory, are derived explicitly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The last section is devoted to the physical interpretation of our results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We apply the whole construction to a suitable class of solutions and compute the action up to finite order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3 2 Deforming the Witten action using stubs We have seen that the algebraic structure of CSFT is that of an L∞-algebra, so naively we should try to find an L∞-based deformation of OSFT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, this would imply a commutative product of open strings, which would hence be fundamentally different from the Witten product.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We still want the deformation be equivalent to Witten theory, so the best we can do is to aim for an A∞-algebra with products Mn obeying MnM1 (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψn) + Mn−1M2 (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψn) + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' + M1Mn (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψn) = 0 ∀n (13) and the multiplication given by MkMl (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψk+l−1) = Mk (Ml (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψl) , Ψl+1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψk+l−1) + + (−)d(Ψ1) Mk (Ψ1, Ml (Ψ2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψl+1) , Ψl+2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψk+l−1) + · · · + + (−)d(Ψ1)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='+d(Ψk−1) Mk (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψk−1, Ml (Ψk, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' , Ψk+l−1)) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (14) From now on we will use the tensor coalgebra notation of [2] where the A∞-relations take the simple form n � k=1 MkMn+1−k = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (15) Geometrically, to mimic the situation of CSFT, we would like to have a three-vertex that does not give rise to a full moduli space cover, such that higher vertices are necessary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One modification that achieves both of these requirements is to attach stubs on the three vertex, which means, small pieces of propagating strings on each input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In this way, there will appear some regions of moduli space that are not covered by Feynman diagrams and hence have to be taken over by elementary vertices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The higher products which will give rise to those vertices will indeed form an A∞-algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The Hamiltonian of our CFT which generates time evolution is L0, so the operator which inserts a strip of length λ into a string diagram is e−λL0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The natural modification of the star product which attaches stubs symmetrically on all three inputs is then given by [12]: m2 (·, ·) → M2 (·, ·) = e−λL0m2 � e−λL0·, e−λL0· � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (16) M2 is cyclic with respect to the simplectic form ω and Q-invariant: ω (Ψ1, M2 (Ψ2, Ψ3)) = − (−)d(Ψ1) ω (M2 (Ψ1, Ψ2) , Ψ3) , (17) QM2 (Ψ1, Ψ2) = −M2 (QΨ1, Ψ2) − (−)d(Ψ1) M2 (Ψ1, QΨ2) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (18) Those relations follow from [Q, L0] = 0 and the fact that L0 is BPZ-even.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It is easy to see that M3 is not associative as expected for an A∞-algebra: e−λL0m2 � e−λL0Ψ1, e−2λL0m2 � e−λL0Ψ2, e−λL0Ψ3 �� + e−λL0m2 � e−2λL0m2 � e−λL0Ψ1, e−λL0Ψ2 � , e−λL0Ψ3 � ̸= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (19) The task is now to explicitly determine all the higher products and prove that they satisfy all relevant properties.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This will be done in two ways, first in a formal algebraic way using homological perturbation theory and second in a more heuristic way using tree diagrams.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 4 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 Elements of homological perturbation theory The starting point of homological perturbation theory (HPT) is a chain homotopy equivalence.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Let V , W be two chain complexes, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' graded vector spaces with a nilpotent differential of degree one denoted by dV (dW ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Furthermore, we are given two chain maps p : V → W and i : W → V of degree zero as well as a homotopy map h : V → V of degree minus one obeying the relations pdV = dW p (20) idW = dV i (21) ip − 1 = hdV + dV h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (22) This ensures that i and p as well as the combination ip leave cohomology classes invariant.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The most common case in literature is that of a special deformation retract (SDR) where W is isomorphic to a subspace of V , see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [13].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The maps i and p are then given by the canonical inclusion and projection, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It follows trivially that pi = 1 and additionally the following annihilation conditions are demanded: hh = 0, hi = 0, ph = 0 (23) It can be shown that any deformation retract with pi = 1 can be made an SDR by a redefinition of the maps [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Lets assume that the differential dV is perturbed by some δ of degree one such that (dV + δ)2 = 0 again.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The homological perturbation lemma (HPL) now states that one can construct a new SDR using the perturbed differential with the other maps given by DV = dV + δ (24) DW = dW + pδ (1 − hδ)−1 i (25) I = (1 − hδ)−1 i (26) P = p (1 − hδ)−1 (27) H = (1 − hδ)−1 h (28) The expression (1 − hδ)−1 requires some explanation: Usually it should be understood as a geometric series (1 − hδ)−1 = 1 + hδ + hδhδ + hδhδhδ + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (29) We demand for the validity of the HPL that this series either converges or terminates, which will be the case in our examples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The proof of the lemma including all relevant relations is straight forward but tedious and can be found in [14, 2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 Transferring algebraic structure: SDR-case Our main purpose for using the HPL will be transferring some algebraic structure from one side of the homotopy equivalence to the other.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Lets consider an SDR with an associative multiplication defined on V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It is a natural question to ask if there exists some “induced product” defined on W: For example, for a, b ∈ W the product ia · ib in V need not be in W anymore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The only thing that could be done is to project it to W, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' define a · b |W = p (ia · ib), but this product will in general not be associative anymore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' What happens 5 is that an associative algebra on one side of the equivalence becomes an A∞-algebra after transferring to the other side.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' A formal way to construct all the higher products is given by the so-called tensor trick: One defines a new SDR over the tensor algebra of V (W) with the tensorial versions of the maps given by dV(W) = ∞ � n=1 n−1 � k=0 1⊗k ⊗ dV (W) ⊗ 1⊗n−k−1 (30) i = ∞ � n=1 i⊗n (31) p = ∞ � n=1 p⊗n (32) h = ∞ � n=1 n−1 � k=0 1⊗k ⊗ h ⊗ (ip)⊗n−k−1 (33) dV(W) is the standard tensor coderivation associated to dV (W),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' see (138),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' i and p are cohomomorphisms while h is defined somewhat asymmetrical which will turn out to be crucial for the formalism to work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One can check directly that those definitions indeed give rise to an SDR again.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The product on V , denoted by m2, can now be treated as a perturbation δ of the coderivative dV, DV = dV + m2 (34) where m2 is the coderivation associated to m2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' DV 2 = 0 will be fulfilled as a consequence of m2 being associative and dV obeying the Leibniz rule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' According to the HPL we will be given a new, perturbed complex with the maps DW = dW + pm2 (1 − hm2)−1 i (35) I = (1 − hm2)−1 i (36) P = p (1 − hm2)−1 (37) H = h (1 − hm2)−1 (38) where DW squares to zero and therefore defines an A∞-algebra on W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The higher products can be obtained by expanding DW, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' projecting onto n inputs and one output: Mn = π1DWπn (39) For example we get M2 (·, ·) = pm2 (i·, i·) (40) as already anticipated above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='3 Transferring algebraic structure: Non-SDR-case We want to apply those concepts now to the problem of attaching stubs in OSFT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The space V should be given by the space of string fields HBCF T with its grading d (Ψ) = gh (Ψ) + 1 and the differential Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' By comparing (16) with (40) it seems natural to define i = p = e−λL0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (41) 6 However, we see immediately that this definition does not give rise to an SDR: i and p are not an inclusion and projection anymore and pi ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Still, in principle the HPL holds for arbitrary homotopy equivalences, so there is a chance to succeed anyway.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Our choice is so far completely symmetric, so lets define W = V = HBCF T , dW = dV = Q and hW = hV = h where we have to solve e−2λL0 − 1 = hQ + Qh (42) The simplest (although not unique) solution for h is h = e−2λL0 − 1 L0 b0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (43) It is important to stress that h is well-behaved also for L0 = 0 and does not have any pole.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One could now try to proceed with the tensor trick as above and eventually compute the map DW but this runs into problems: Although DW 2 = 0 still, as guaranteed by the HPL, DW is not a coderivation anymore!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This means, it does not obey the co-Leibniz rule (136) anymore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' As explained in the Appendix of [15], the condition pi = 1 as well as the annihilation relations are necessary (and sufficient) for the tensor trick to work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' So there has to be some modification to account for that and it will actually turn out to be surprisingly simple: We can just take the expression we get for DW and while expanding to calculate the higher products, pretend that all SDR-relations are satisfied.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' More precisely, we define an operator PSDR acting on the space of maps from T H → T H which projects on maps in which all SDR-relations hold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This means, every time that pi occurs, it will be replaced by 1 and every time hh, hi or ph occurs, the term will be discarded: PSDR (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.hh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.) = 0, PSDR (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.hi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.) = 0, PSDR (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.ph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.) = 0, PSDR (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.pi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.) = PSDR (.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='.) Now our new higher products will be given as Mn = PSDRπ1DWπn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (44) Their associated coderivations can be added together to form a total map called M, M = ∞ � n=1 ∞ � m=1 m−n � k=0 1⊗k ⊗ Mn ⊗ 1⊗m−k−n (45) It is a coderivation by construction and moreover it squares to zero because in the proof of the HPL, where it is shown that D2 W = 0 (Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='16 of [2]), the SDR-relations are never used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Since we know that DW 2 = 0 independently of the SDR-relations, also M2 = 0 and M defines the desired A∞-algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' As an example, lets explicitly calculate M3 : PSDRπ1DWπ3 = PSDRπ1pm2hm2iπ3 = PSDR � pm2 (1 ⊗ h + h ⊗ ip) (1 ⊗ m2 + m2 ⊗ 1) i⊗3� = PSDR (pm2 (i·, hm2 (i·, i·)) − pm2 (m2 (i·, i·) , hi·) + pm2 (hi·, ipm2 (i·, i·)) + pm2 (hm2 (i·, i·) , ipi·)) = pm2 (i·, hm2 (i·, i·)) + pm2 (hm2 (i·, i·) , i·) (46) The second and third term in the second line contained an hi and were deleted whereas in the last term ipi was replaced by i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This whole procedure including the proof of the statement may seem quite handwavy, however, there exists a more formal and precise way of arriving at the same result using operad theory [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' A combinatorial proof using tree diagrams will be given in the next section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 7 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4 Higher products using tree diagrams The proposal is that Mn is equal to the sum of all distinct, rooted, full binary trees with n leaves such that every leaf represents one input and the root is the output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' With every leaf there is one factor of i associated, with every node the product m2, with every internal line h and with the root p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In [17] it is argued that this is true for SDRs, since we can construct the products in the same way as for an SDR, we conclude that the proposal also holds for our non-SDR case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the tree language, the A∞-relations can be proven directly: The commutator of Q with an n-leaved tree gives a sum of n − 2 terms in which one of the n − 2 internal lines h is replaced by 1 − ip.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The 1-terms actually cancel away because of the associativity of m2: The propagator associated with unity connects two nodes m2 which leads to an expression of the form m2 (m2 (A, B) , C) where A, B, C are three subtrees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the sum there always exists a second tree with another propagator turned into unity giving rise to the expression m2 (A, m2 (B, C)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' These two trees cancel away such that only the -ip-factors remain in total.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Now the other terms occuring in the relation − [Q, Mn] = M2Mn−1 + M3Mn−2 + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' + Mn−1M2 (47) can be interpreted as follows: If we project on one output, π1MkMn+1−k gives a sum of trees where one of the terms in Mn+1−k is connected with its root to one of the k leaves of one of the trees in Mk and this is done in all possible combinations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The result is a sum of trees with in total n leaves where one of the internal lines does not contain h but ip;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' the i from the leaf of the left tree and the p from the root of the right tree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' But that is exactly the same sum of terms we have on the l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', indeed it is easy to see that each tree occuring on the r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' must also occur on the l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' and vice versa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' An interesting crosscheck of the π1-projection of (47) can be done by comparing the total number of trees on both sides: The number of full binary trees with n + 1 leaves is given by the Catalan number Cn = 1 n + 1 � 2n n � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (48) This means that on the l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' there are Cn−1 trees in Mn and n − 2 internal lines that can be changed, hence a total of Cn−1 (n − 2) trees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' On the r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' we have 2 · C1Cn−2 + 3 · C2Cn−3 + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' + (n − 1) Cn−2C2 trees in total, leading to the equation n−1 � k=2 (n + 1 − k) Cn−kCk−1 = (n − 2) Cn−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (49) The Catalan numbers fulfill the following useful recursive relations: Cn+1 = 2 (2n + 1) n + 2 Cn,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Cn+1 = n � k=0 Cn−kCk (50) 8 Using them one can proceed by induction: Assuming equation (49) is valid for some n,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' then n � k=2 (n + 2 − k) Cn+1−kCk−1 = n � k=2 (n + 2 − k) 2 (2n − 2k + 1) n − k + 2 Cn−kCk−1 = 4 n−1 � k=2 (n + 1 − k) Cn−kCk−1 − 2 n−1 � k=2 Cn−kCk−1 + 2C0Cn−1 = (4n − 8) Cn−1 − 2 n−2 � k=1 Cn−1−kCk + 2Cn−1 = (4n − 6) Cn−1 − 2 (Cn − C0Cn−1 − Cn−1C0) = (4n − 2) n + 1 2 (2n − 1)Cn − 2Cn = (n − 1) Cn (51) as it should be to complete the induction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This shows that the number of terms in the equation (47) is the same on both sides.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='5 Proof of cyclicity to all orders Using the tree language it is possible to prove that all higher products Mn are cyclic with respect to the BPZ-product.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We have to show ω (Ψ1, Mn (Ψ2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn+1)) = − (−)d(Ψ1) ω (Mn (Ψ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψn) , Ψn+1) , (52) hence we start with a sum of trees on the l.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' and use the BPZ-properties of m2 and h as well as p = i† = i to rewrite it as the sum of trees on the r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='. The explicit steps are: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Take the p from the root of the tree and write it to the left side of the product where it can be interpreted as i, acting on Ψ1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Take the m2 from the root of the tree and use cyclicity of m2 to apply it on the first two arguments inside of ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This gives a sign factor of − (−)d(Ψ1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In general one will be left with two subtrees then with n + 1 leaves in total.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Take the h from the right subtree and use that it is BPZ-even to apply it on the left subtree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This gives an additional sign factor of (−)d(left subtree).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Take the m2 from the root of the right subtree and apply it on the first two arguments inside of ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It gives a sign factor of − (−)d(left subtree)+1 (the +1 comes from the h that was shifted in step 3) which cancels the sign factor of step 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Again, one is left with two subtrees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Repeat steps 3 and 4 until the right subtree only consist of i acting on one input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The total sign factor remains − (−)d(Ψ1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 9 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Remove the i acting on Ψn+1 and let it act as a p on the left subtree.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Now the left subtree fulfills all requirements to be an element of Mn and since we also have the right sign factor, we have obtained a term contained in the r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' of (52).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The manipulations are all uniquely invertible so we can conclude that the map between the trees is one-to-one and all terms we need are constructed exactly once.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' As a result, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (52) holds and all higher products are cyclic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Moreover, as already suggested by the name, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (52) together with the antisymmetry of ω implies invariance of the vertices under cyclic permutations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='6 Geometric picture We have now shown that the higher products fulfill all the algebraic requirements but we do not know anything yet about the geometric picture, if they indeed give rise to a full single cover of the moduli space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' To answer this question, the tree description of the products turns out to be very useful.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Lets consider an arbitrary string tree diagram using the stubbed three vertex: As long as the external states are on-shell, the stubs make no difference because the external legs consist of a semiinfinite strip anyway.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' On the internal lines instead, the stubs make a difference because all internal strips with a length smaller than 2λ do not appear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We can conclude that the additional elemantary vertices we need should consist of all tree diagrams with all internal strips having a length smaller than 2λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The Siegel-gauge string propagator in the Schwinger parametrization is given by � ∞ 0 dt e−tL0b0 = b0 L0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (53) The integral over t can be thought of an integral over strips of propagating strings of all different lengths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Following this logic, the propagators in our new vertices should be given as � 2λ 0 dt e−tL0b0 = −e−2λL0 − 1 L0 b0 = −h (54) and hence be equal to minus the homotopy!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 We have constructed the higher products by drawing all binary tree diagrams with Witten vertices, h as propagators and e−λL0 on the leaves and the root.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Those are in one-to-one correspondence with all the Feynman tree diagrams that should make up the new elementary vertices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This shows that our higher products Mn indeed define a set of vertices which gives rise to a full cover of the moduli space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3 Analytic solutions and action(s) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 Projection cohomomorphism from the HPL Using the definition m = Q + m2 (55) the equations of motion of the original Witten theory can be written in coalgebra language as m 1 1 − Ψ = 0, (56) 1It is interesting to notice at this point that the simple but not unique choice of h corresponds to choosing Siegel gauge for the propagator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The minus sign is just a convention and can be absorbed in the definition of h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 10 i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' solutions are Maurer-Cartan elements of the A2-algebra defining the theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the same spirit, to find solutions of the deformed theory, we have to solve the equation M 1 1 − Ψ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (57) In fact, the homological perturbation lemma already gave us an operator which maps solutions of the Witten theory to solutions of the stubbed theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This can be seen as follows: The perturbed projection P is a chain map and hence obeys Pm = MP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (58) Now lets assume Ψ∗ is a solution of the Witten theory, then (Ψ∗)′ = π1P 1 1 − Ψ∗ (59) obeys M 1 1 − (Ψ∗)′ = M 1 1 − π1P 1 1−Ψ∗ = MP 1 1 − Ψ∗ = Pm 1 1 − Ψ∗ = 0 (60) where Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (146) was used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It remains to determine the cohomorphism P for the case where the homotopy equivalence is not an SDR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Again, as for the higher products above, the simplest way is to take the expression from the HPL PHP L = p (1 − m2h)−1 , (61) and apply the operator PSDR on its components to get Pn = PSDRπ1PHP Lπn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (62) The resulting maps can then be packaged into a cohomorphism P again.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' More explicitly, we get for the first few orders P1 = p P2 = pm2 (·, h·) + pm2 (h·, ip·) P3 = pm2 (·, hm2 (·, h·)) + pm2 (·, hm2 (h·, ip·)) + pm2 (h·, hm2 (ip·, ip·)) + pm2 (hm2 (·, h·) , ip·) + pm2 (hm2 (h·, ip·) , ip·) + pm2 (h·, ipm2 (·, h·)) + pm2 (h·, ipm2 (h·, ip·)) (63) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One can now check the equation π1Pm = π1MP order by order: π1Pmπ1 = pQ = Qp = π1MPπ1 π1Pmπ2 = P2 (Q·, ·) + P2 (·, Q·) + pm2 = −pm2 (Q·, h·) + pm2 (hQ·, ip·) + pm2 (·, hQ·) + pm2 (h·, ipQ·) + pm2 = −pm2 (Q·, h·) − pm2 (Qh·, ip·) + pm2 ((ip − 1) ·, ip·) − pm2 (·, Qh·) + pm2 (·, (ip − 1) ·) + pm2 (h·, Qip·) + pm2 = Qpm2 (·, h·) + Qpm2 (h·, ip·) + pm2 (ip·, ip·) = QP2 + M2 (p·, p·) = π1MPπ2 (64) 11 For order three the calculation is already very tedious but it also turns out to work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The important point is that in the manipulations that are necessary, the SDR-relations were never used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This implies, since we know that (63) works for SDRs, it also works in our case and (63) is a valid definition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Now we have a cohomorphism by construction that obeys the chain map relation (58) such that we can construct analytic solutions of the deformed theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 The action The on-shell action is one of the most important observables in OSFT, for example for the tachyon vacuum its value is equal to minus the energy of the D-brane which has decayed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Since the stubbed theory should be physically equivalent to the original Witten theory, we expect that the values for the on-shell action we get in the two theories coincide.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The Witten action can be written in coalgebra notation [2] as S (Ψ) = � 1 0 dt ω � π1∂t 1 1 − Ψ (t), π1m 1 1 − Ψ (t) � (65) where Ψ (t) is any smooth interpolation between Ψ (0) = 0 and Ψ (1) = Ψ and ∂t the coderivation associated to ∂t.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Similarly, the stubbed action reads S′ (Ψ) = � 1 0 dt ω � π1∂t 1 1 − Ψ (t), π1M 1 1 − Ψ (t) � (66) We would expect now a relation S′ � (Ψ∗)′� ?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='= S (Ψ∗) (67) with Ψ∗ a MC-element of m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, this relation turns out not to be true: Instead S (Ψ) = � 1 0 dt ω � π1∂t 1 1 − Ψ (t), π1m 1 1 − Ψ (t) � = � 1 0 dt ω � π1∂tP−1P 1 1 − Ψ (t), π1P−1MP 1 1 − Ψ (t) � = � 1 0 dt ω � π1∂tP−1 1 1 − π1P 1 1−Ψ(t) , π1P−1M 1 1 − π1P 1 1−Ψ(t) � = � 1 0 dt ω � π1P−1∂t 1 1 − (Ψ)′ (t), π1P−1M 1 1 − (Ψ)′ (t) � =: ˜S (Ψ′) (68) where the last line differs from S′ (Ψ′) by the insertions of P−1 on both inputs of ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The invertibility of P is actually a delicate question: In general, a cohomomorphism is invertible iff its linear component, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' P1 = p is invertible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Now p = e−λL0 inserts a strip of length λ, so one would expect the inverse eλL0 to remove a strip of length λ from the world sheet, which is not always possible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' On the other hand, eλL0 makes sense on a string field expanded in eigenstates of L0 as long as the eigenvalues are finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' From now on, we shall assume that this is the case and eλL0 is well-defined on all string fields in question.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Explicitly, the inversion of P works as follows: PP−1 = P−1P should be equal to the identity cohomo- morphism, which is identity in its linear component and zero in all higher components.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One can now solve 2It is actually a non-trivial counting problem to determine the number of terms of Pn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It grows faster than the Catalan numbers because at order three we have seven terms and at order four already 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 12 the components P −1 n = π1P−1πn order by order: P −1 1 = p−1 P −1 2 = −m2 � p−1·, hp−1· � − m2 � hp−1·, i· � P −1 3 = −m2 � hp−1·, hm2 (i·, i·) � + m2 � m2 � p−1·, hp−1· � , hp−1· � + m2 � m2 � hp−1·, i· � , hp−1· � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (69) One can now write out in more detail ˜S (Ψ) = 1 2ω � p−1Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' p−1QΨ � + 1 3ω � p−1Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' p−1M2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) � + 1 3ω � p−1Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' P −1 2 ((QΨ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) + (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' QΨ)) � + 1 3ω � P −1 2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' p−1QΨ � + 1 4ω � p−1Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' p−1M3 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) � + 1 4ω � p−1Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' P −1 2 ((M2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) + (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' M2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ))) � + 1 4ω � p−1Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' P −1 3 ((QΨ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) + (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' QΨ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) + (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' QΨ)) � + 1 4ω � P −1 2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' p−1M2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) � + 1 4ω � P −1 2 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' P −1 2 ((QΨ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) + (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' QΨ)) � + 1 4ω � P −1 3 (Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ) ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' p−1QΨ � + O � Ψ⊗5� (70) It seems that the cohomomorphism P does not define the field redefinition we were looking for,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' instead it relates the Witten theory to a theory defined by ˜S (Ψ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The equations of motion derived from ˜S are the same as for S′, namely M 1 1−Ψ = 0, but it is not at all obvious that the two actions agree even on-shell.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The reason is that the HPL does not know anything about the symplectic form ω: To get an invariant action with S′ (Ψ′) = S (Ψ), not only m has to transform accordingly, but also ω would have to go to ω′ (Ψ1, Ψ2) = ω � π1P−1 1 1 − Ψ1 , π1P−1 1 1 − Ψ2 � , (71) otherwise P would fail to be cyclic.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' As it can be seen from (68), ˜S (Ψ) is just a fancy rewriting of the Witten action and therefore defines an equivalent theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' But since we would like to keep the original ω, the field redefinition induced by P from the HPL and giving rise to ˜S (Ψ) is not exactly what we want.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, since it shares the same equations of motion as S′ (Ψ), it might provide a new family of gauge-invariant observables for solutions of S′ (Ψ), parametrized by λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the last section we will check this explicitly on a special class of solutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' For now the next task is to derive the originally desired field redefinition which relates the actions S (Ψ) and S′ (Ψ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='3 Elements from closed string field theory In [18], Zwiebach and Hata have shown how to relate slightly different, consistent sets of vertices in CSFT via an infinitesimal field redefinition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Our strategy is now to apply their method to our problem in OSFT and integrate the result to the finite case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This will not only provide us the field redefinition we are looking for, but also give some insight into the rather abstract formalism of CSFT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' First, it is useful to collect some basic information about the structure of CSFT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3It would also be an interesting direction to examine if there exists some kind of “dual” HPL that directly yields this correct field redefinition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 13 As already explained in the introduction, the vertices are given by integrating basic differential forms Ωg,n Ψ1Ψ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='Ψn defined by Ωg,n Ψ1Ψ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='Ψn � ˆV1, ˆV2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', ˆV6g−6+2n � = (2πi)−(3g−3+n) ⟨Σ|b (v1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='b (v6g−6+2n)|Ψ1⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='|Ψn⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (72) They are living in the tangent space of the fibre bundle ˆPg,n over the moduli space Mg,n, with the fiber being the space of local coordinates around the punctures modulo phase rotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The dimension of this bundle is infinite, but the degree of Ωg,n Ψ1Ψ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='Ψn is just the real dimension of the base space Mg,n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It takes as arguments tangent vectors ˆVi ∈ T ˆPg,n, which represent deformations of the world sheet Riemann surface Σ, either by changing the moduli or the local coordinates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The vi are Schiffer vectors on Σ, supported around the punctures, which generate those deformations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This means that the local coordinate around the nth puncture transforms as z(n) → z(n) + ǫv(n) � z(n)� (73) for some small ǫ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The b-ghost insertions are then defined as b (v) = n � i=1 �� dzi 2πib (zi) v(i) (zi) + � d¯zi 2πi ¯b (¯zi) ¯v(i) (¯zi) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (74) We will be only interested in the classical action without the loop vertices, so the genus g shall be zero from now on.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The basic forms can now be integrated over sections of ˆP0,n defining the vertices V0,n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The quantization procedure makes use of the Batalin-Vilkovisky formalism;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' although we are not interested in quantum effects, the BV-antibracket is used in constructing the symmetry generator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' It is defined as {A, B} = ∂rA ∂Ψi ∂lB ∂Ψ∗ i − ∂rA ∂Ψ∗ i ∂lB ∂Ψi (75) where the Ψ∗ i are antifields of opposite parity associated to each basis element of H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The BV-master action takes the same form as (7) with the only difference that the Ψ are not restricted in ghost number and run over fields as well as antifields.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4 Constructing the field redefinition We are looking for a non-linear field redefinition of the form Ψ′ = F (Ψ) = ∞ � n=1 Fn � Ψ⊗n� = π1F 1 1 − Ψ (76) that relates the Witten action to the stubbed A∞-action in the form (66).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' To be consistent with the results of Zwiebach and Hata [18] we demand S (Ψ′) = S′ (Ψ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (77) Since the kinetic term is identical we immediately find F1 (Ψ) = Ψ (78) In [18] it is shown that under an infinitesimal field redefinition of the form Ψ → Ψ + t {Ψ, e} + O � t2� (79) 14 the classical action transforms as S (Ψ) → S (Ψ) + t {S, e} + O � t2� , (80) where {} denotes the BV-antibracket.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In the paper it is now argued that for any small change of vertices, the change of the action indeed takes this form and the generator e is constructed explicitly: e (u0) = − � n κn−2 1 n!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' � V0,n(u0) Ω(0)0,n b(u)Ψ⊗n (81) Here we assume that there exists some family of consistent vertex sets V0,n (u) parametrized by some real number u and everything is evaluated at the point u0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The vector u is some Schiffer vector which generates a deformation of the V0,n (u0) in the direction of u, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' it generates diffeomorphisms which push V0,n (u0) into V0,n (u0 + δu).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' For the case of varying the stub length, this Schiffer vector takes a particular simple form: First, lets notice that the stub length λ for closed strings is defined as the geodesic distance from the location | z |= 1 of the local coordinate to the begin of the semiinfinite cylinder associated with the puncture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This implies that λ can be changed by just rescaling the coordinate: Sending z to z′ = z + ǫz, the location of | z′ |= 1 corresponds to | z |= 1 − ǫ, such that λ is increased by ǫ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' By comparing with Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (73) we read off u(i) = z(i).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (82) The b-ghost insertion is then given by b (u) = n � i=1 �� dzi 2πizib (zi) + � d¯zi 2πi ¯zi¯b (¯zi) � = n � i=1 b(i) 0 + b (i) 0 = n � i=1 b+(i) 0 (83) We want to use the above expression for e in the context of changing the stub length for open strings, hence a few modifications and simplifications are necessary: First, the combinatorial factor n!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' originates from total symmetrization of the vertices and is not necessary for open strings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Second, the insertions of b+ 0 should get replaced simply by b0 since there is no antiholomorphic sector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Moreover, the string coupling κ will be set to one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Now the generator simplifies to e (λ) = − � n � V(λ)0,n Ω0,n b0Ψ⊗n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (84) If we make the ansatz Ψ′ = Ψ + δλ ∞ � n=2 fn � Ψ⊗n� (85) as the infinitesimal version of (76), then the fn are determined as fn � Ψ⊗n� = � Ψ, e(n)� (86) where δλ plays the role of t in (79).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' To find f2 (Ψ, Ψ) we need to consider e (λ) for n = 3: The vertex V (λ)0,3 is zero dimensional, so there is no integral and the surface state ⟨Σ| is just the Witten vertex with stubs of length λ, ⟨V3 (λ) | = ω (·, M2 (·, ·)) = ω � e−λL0·, e−λL0 · ∗e−λL0· � (87) 15 Inserting into (84) yields e(3) (λ, Ψ) = − (ω (b0Ψ, M2 (Ψ, Ψ)) + ω (Ψ, M2 (b0Ψ, Ψ)) + ω (Ψ, M2 (Ψ, b0Ψ))) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (88) The BV-bracket with Ψ can be straightforwardly evaluated;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' after carefully checking the signs the result is f2 (Ψ, Ψ) = −b0M2 (Ψ, Ψ) + M2 (b0Ψ, Ψ) + M2 (Ψ, b0Ψ) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (89) We expect now the relation S′ (Ψ + δλf2 (Ψ, Ψ) , λ) = S′ (Ψ, λ + δλ) (90) to hold up to order 3 in Ψ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' by directly inserting we can compute explicitly S′ (Ψ + δλf2 (Ψ, Ψ) , λ) = 1 2ω (Ψ, QΨ) + 1 3ω (Ψ, M2 (Ψ, Ψ)) + δλ ω (Ψ, Qf2 (Ψ, Ψ)) + O � Ψ⊗4� (91) The last and most interesting term yields (−ω (Ψ, Qb0M2 (Ψ, Ψ)) + ω (Ψ, QM2 (b0Ψ, Ψ)) + ω (Ψ, QM2 (Ψ, b0Ψ))) δλ = (−ω (Ψ, L0M2 (Ψ, Ψ)) + ω (Ψ, b0QM2 (Ψ, Ψ)) − ω (QΨ, M2 (b0Ψ, Ψ)) − ω (QΨ, M2 (Ψ, b0Ψ))) δλ = − ω (Ψ, L0M2 (Ψ, Ψ)) δλ (92) where the last three terms in the second line cancel after applying the Leibniz rule and cyclicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' On the other hand, S′ (Ψ, λ + δλ) = S′ (Ψ, λ) + δλ d dλS′ (Ψ, λ) = 1 2ω (Ψ, QΨ) + 1 3ω (Ψ, M2 (Ψ, Ψ)) − 1 3δλ ω (L0Ψ, M2 (Ψ, Ψ)) − 1 3δλ ω (Ψ, M2 (L0Ψ, Ψ)) − 1 3δλ ω (Ψ, M2 (Ψ, L0Ψ)) + O � Ψ⊗4� = 1 2ω (Ψ, QΨ) + 1 3 ω (Ψ, M2 (Ψ, Ψ)) − δλ ω (L0Ψ, M2 (Ψ, Ψ)) + O � Ψ⊗4� (93) which is the same expression up to order Ψ⊗3 (Again, cyclicity was used in the last line.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The explicit form of f2 (Ψ, Ψ) suggests the following general structure: We can guess the ansatz fn � Ψ⊗n� = −b0Mn � Ψ⊗n� + Mn � b0 � Ψ⊗n�� (94) where b0 again denotes the coderivation associated to b0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' At first sight, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ( 94 ) looks a bit strange now from the coalgebra perspective because it is a commutator of two odd objects, so one would more naturally expect an anticommutator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, the first term in (94) stems from the application of b0 on the first argument of the symplectic form ω, hence the sign contains implicit information about ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' From the discussion about P from the HPL we could anticipate that ω has to enter the calculation at some point.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The more natural looking expression π1 [b0, Mn] would have been independent of ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One can prove now that the ansatz ( 94 ) is indeed correct by directly inserting into the action.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='5 Proof of the ansatz for the infinitesimal field redefinition If we focus solely on terms of order n + 1 in Ψ we get S′(n+1) (Ψ′) = ω � Ψ, Qfn � Ψ⊗n�� δλ+ n−1 � k=2 ω � Ψ, Mk � fn+1−k � Ψ⊗n+1−k� , Ψ⊗k−1�� δλ+ 1 n + 1ω � Ψ, Mn � Ψ⊗n�� (95) 16 The last term is the contribution from the original S (Ψ), so the first two terms denoted by δS(n+1) should yield the infinitesimal variation δS′(n+1) ?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='= δλ d dλ 1 n + 1ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn � Ψ⊗n�� (96) Inserting the ansatz and performing some straight forward manipulations gives δS′(n+1) δλ = ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Q (−b0Mn + Mnb0) � Ψ⊗n�� + n−1 � k=2 ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � (−b0Mn+1−k + Mn+1−kb0) � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1�� = − ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' L0Mn � Ψ⊗n�� + ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0 [Q,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn] � Ψ⊗n�� − ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0MnQ � Ψ⊗n�� + ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' QMnb0 � Ψ⊗n�� + n−1 � k=2 � −ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � b0Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1�� + ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � Mn+1−kb0 � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1��� (97) The last two terms of the second line actually cancel each other: − ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0MnQ � Ψ⊗n�� + ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' QMnb0 � Ψ⊗n�� = − ω � b0Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' MnQ � Ψ⊗n�� − ω � QΨ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mnb0 � Ψ⊗n�� = − ω (b0Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn (QΨ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψ)) − ω (b0Ψ, Mn (Ψ, QΨ, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψ)) − ω (b0Ψ, Mn (Ψ, Ψ, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', QΨ)) − ω (QΨ, Mn (b0Ψ, Ψ, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψ)) − ω (QΨ, Mn (Ψ, b0Ψ, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', Ψ)) − ω (QΨ, Mn (Ψ, Ψ, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', b0Ψ)) (98) Because of cyclicity of the (n + 1)-vertex the last two lines contain of the same terms, just differing by a sign which comes from commuting Q with b0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Therefore they add to zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The second term in of the second line of (97) can be further manipulated using the A∞-relations: δS′(n+1) δλ = − ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' L0Mn � Ψ⊗n�� − n−1 � k=2 ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � b0Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1�� + n−1 � k=2 � ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � Mn+1−k � b0 � Ψ⊗n+1−k�� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1�� − ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0MkMn+1−k � Ψ⊗n��� (99) Now again,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' the terms in the last line cancel after using cyclicity: ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0MkMn+1−k � Ψ⊗n�� = ω � b0Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' MkMn+1−k � Ψ⊗n�� = ω � Mk � b0Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−2� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ � + ω � Mk � b0Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−3� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ � + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' + ω � Mk � b0Ψ, Ψ⊗k−2, Mn+1−k � Ψ⊗n+1−k�� , Ψ � + ω � Mk � b0Ψ, Ψ⊗k−1� , Mn+1−k � Ψ⊗n+1−k�� (100) All terms except for the last one can be further manipulated using the antisymmetry of ω and cyclicity of Mk.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' For example, ω � Mk � b0Ψ, Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−2� , Ψ � = −ω � Ψ, Mk � b0Ψ, Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−2�� = ω � Mk � Ψ, b0Ψ, Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−3� , Ψ � = .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' = ω � Mk � Ψ⊗k−1, b0Ψ � , Mn+1−k � Ψ⊗n+1−k�� (101) 17 so in all of the terms the Mn+1−k can be moved to the outermost right.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' After all, the terms can be summed up as ω � Mk � Ψ⊗k−2, b0Ψ, Mn+1−k � Ψ⊗n+1−k��� + ω � Mk � Ψ⊗k−3, b0Ψ, Ψ, Mn+1−k � Ψ⊗n+1−k��� + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' + ω � Mk � b0Ψ, Ψ⊗k−1� , Mn+1−k � Ψ⊗n+1−k�� = ω � Mk � b0 � Ψ⊗k�� , Mn+1−k � Ψ⊗n+1−k�� = ω � Ψ, Mn+1−k � Mk � b0 � Ψ⊗k�� , Ψ⊗n−k�� (102) which is after the summation over k identical to the first term in the second line of (99), just with opposite sign.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' So we arrive at the expression δS′(n+1) δλ = −ω � Ψ, L0Mn � Ψ⊗n�� − n−1 � k=2 ω � Ψ, Mk � b0Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−1�� (103) which should now be compared to the result of formula (96).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The derivative with respect to λ can act on the stubs as well as on the homotopy h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The action on e−λL0 inserts a factor of −L0 on every input string field of the (n + 1)-vertex.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Since the vertices are cyclically symmetric, we get n + 1 identical terms, which cancels the prefactor 1 n+1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The result is d dλ 1 n + 1ω � Ψ, Mn � Ψ⊗n�� ⊃ −ω � L0Ψ, Mn � Ψ⊗n�� (104) which is equal to the first term of (103).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' To compute the action on h, the tree representation turns out to be useful again: First of all, d dλh = −2b0e−2λL0 (105) hence we get a sum of all possible tree diagrams with one propagator replaced by −2b0e−2λL0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We can cut through the diagram along this replaced propagator and think of the factor e−2λL0 as arising from two e−λL0-stubs from the leaf and the root of the two subtrees.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Both subtrees are now part of a higher product Mk for some k, 2 ≤ k ≤ n − 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' So the whole expression can be written as a combination of two higher products Mk, Mn+1−k with a factor −2b0 inserted: d dλ 1 n + 1ω � Ψ, Mn � Ψ⊗n�� ⊃ − 1 n + 1 n−1 � k=2 ω � Ψ, Mk � 2b0Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−1�� + ω � Ψ, Mk � Ψ, 2b0Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−2�� + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='+ ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � Ψ⊗k−1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 2b0Mn+1−k � Ψ⊗n+1−k��� (106) Because of cyclicity of the k + 1-vertex,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' the different lines contain the same terms so we have d dλ 1 n + 1ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn � Ψ⊗n�� ⊃ − 1 n + 1 n−1 � k=2 2k · ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � b0Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1�� (107) The last bracket can be further manipulated: ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � b0Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1�� = −ω � Mk � b0Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗k−1� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ � = ω � b0Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mk � Ψ⊗k�� = −ω � Mn+1−k � Ψ⊗n+1−k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0Mk � Ψ⊗k�� = ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn+1−k � Ψ⊗n−k,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' b0Mk � Ψ⊗k��� = ω � Ψ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mn+1−k � b0Mk � Ψ⊗k� ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Ψ⊗n−k�� (108) 18 In the last step cyclicity of the (n + 1 − k)-vertex was used again.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We see that in the sum of (107) the kth term and the (n + 1 − k)th term are identical so the sum can be rewritten as − 1 n + 1 n−1 � k=2 (n + 1) · ω � Ψ, Mk � b0Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−1�� = n−1 � k=2 ω � Ψ, Mk � b0Mn+1−k � Ψ⊗n+1−k� , Ψ⊗k−1�� (109) which is precisely the second term in (103).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This completes the proof that the ansatz fn � Ψ⊗n� = −b0Mn � Ψ⊗n� + Mn � b0 � Ψ⊗n�� (110) indeed yields the correct infinitesimal field redefinition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='6 Finite field redefinition So far we have only been concerned with infinitesimal variations of λ, now we want to generalize the results to finite changes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We know Ψλ+δλ = Ψλ + δλf λ 2 (Ψλ, Ψλ) + δλf λ 3 (Ψλ, Ψλ, Ψλ) + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' = Ψλ + δλ d dλΨλ (111) where we have written the superscript λ to indicate that the fn also depend on λ explicitly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' This equation can be integrated to Ψλ = Ψ0 + � λ 0 dtf t 2 (Ψt, Ψt) + � λ 0 dtf t 3 (Ψt, Ψt, Ψt) + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (112) Inserting Ψλ back we get a perturbative expansion in the original solution Ψ0 : Ψλ =Ψ0 + � λ 0 dtf t 2 (Ψ0, Ψ0) + � λ 0 dtf t 3 (Ψ0, Ψ0, Ψ0) + � λ 0 dtf t 2 �� t 0 dsf s 2 (Ψ0, Ψ0) , Ψ0 � + � λ 0 dtf t 2 � Ψ0, � t 0 dsf s 2 (Ψ0, Ψ0) � + O � Ψ⊗4 0 � (113) This formula provides an algorithm to find the associated A∞-solution to each known solution of the Witten OSFT.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 4 Physical interpretation To summarize, we found two distinct field redefinitions ˜Ψ = ∞ � n=1 Pn � Ψ⊗n� , Ψ′ = ∞ � n=1 Fn � Ψ⊗n� (114) 19 which generate two different actions ˜S and S′ via ˜S � ˜Ψ � = S (Ψ) , S′ (Ψ) = S (Ψ′) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (115) However, both actions share the same equations of motion, namely the Maurer-Cartan equation of the stubbed A∞-algebra π1M 1 1 − Ψ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (116) It remains to examine what the physical meaning of those two actions is, most importantly, if they yield the same on-shell value for a given solution Ψ∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In [19, 20], a special class of solutions containing a nearly marginal vertex operator is introduced which serves as a useful playground to analyze this question.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1 Nearly marginal solutions Consider a matter conformal primary field V with weight h smaller but very close to one.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The string field Ψ1 = µ · cV (0) |0⟩, (117) with some real coupling constant µ obeys the Siegel gauge condition b0Ψ1 = 0 (118) and will serve as a starting point to find the full solution Ψ = � Ψn as a perturbative series in the expansion parameter y = 1 − h.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One can solve for the string coupling µ using the Witten equations of motion QΨ + Ψ ∗ Ψ = 0 (119) to obtain [20] µ = y CV V V + O � y3� , (120) where CV V V denotes the three-point function constant of V .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We can deduce that Ψ1 = O (y) and from the perturbative algorithm for the full solution one can also show that in general Ψn = O (yn).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The on-shell action in Witten theory can be written compactly as S (Ψ) = −1 6 ⟨ Ψ, QΨ ⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (121) From QΨ1 = µy · c∂cV (0) |0⟩ (122) we see that the action will be of leading order y3 and given by S (Ψ) = −1 6 y3 C2 V V V ⟨ cV, c∂cV ⟩ = 1 6 y3 C2 V V V + O � y4� (123) if we assume that V is conveniently normalized, ⟨ V (z1) , V (z2) ⟩ = z−2h 12 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 20 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2 The A∞-action The first important observation is that the cohomomorphism P simplifies significantly for string fields in Siegel gauge: Since h is proportional to b0, it annihilates Ψ and (63) collapses to ˜Ψ = ∞ � n=1 Pn � Ψ⊗n� = pΨ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (124) We already know that ˜S � ˜Ψ � yields the original value S (Ψ), so now we want check the expression S′ � ˜Ψ � : Ψ1 is an L0-eigenstate so we straightforwardly get pΨ1 = e−λL0Ψ1 = eλyΨ1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (125) For cubic order in y we just have to insert this into the kinetic term (121) and get S′ � ˜Ψ � |y3= −1 6 ⟨ pΨ1, QpΨ1 ⟩ = 1 6 y3 C2 V V V e2λy |y3= 1 6 y3 C2 V V V (126) which agrees with the result above up to terms of O � y4� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' For a more non-trivial check we can collect the terms of quartic order in y: In the action we have to consider the first three terms S′ � ˜Ψ � |y4= � −1 2 � ˜Ψ, Q˜Ψ � − 1 3 � ˜Ψ, M2 � ˜Ψ, ˜Ψ � � − 1 4 � ˜Ψ, M3 � ˜Ψ, ˜Ψ, ˜Ψ � �� |y4 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (127) The equations of motion however tell us that � ˜Ψ, Q˜Ψ � + � ˜Ψ, M2 � ˜Ψ, ˜Ψ � � + � ˜Ψ, M3 � ˜Ψ, ˜Ψ, ˜Ψ � � = O � y5� , (128) so the expression simplifies to S′ � ˜Ψ � |y4= � −1 4 � ˜Ψ, Q˜Ψ � − 1 12 � ˜Ψ, M2 � ˜Ψ, ˜Ψ � �� |y4 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (129) Plugging in ˜Ψ = eλyΨ1 and isolating y4-terms yields S′ � ˜Ψ � |y4= � −1 4 � eλyΨ1, QeλyΨ1 � − 1 12 � eλyΨ1, e−λL0 � e−λL0eλyΨ1 ∗ e−λL0eλyΨ1 � �� |y4 = � −1 4e2λy ⟨ Ψ1, QΨ1 ⟩ − 1 12e6λy ⟨ Ψ1, (Ψ1 ∗ Ψ1) ⟩ � |y4 = − 1 2 λy4 C2 V V V ⟨ cV, c∂cV ⟩ − 1 2 λy4 C3 V V V ⟨ cV, (cV ∗ cV ) ⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (130) The correlation functions can be calculated by using standard CFT methods, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [21, 20];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' the result is ⟨ cV, c∂cV ⟩ = −1, ⟨ cV, (cV ∗ cV ) ⟩ = CV V V � 3 √ 3 4 �3y = CV V V � 1 + 3y · ln � 3 √ 3 4 �� +O � y2� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (131) 21 We see by inserting into (130) that S′ � ˜Ψ � |y4 indeed vanishes and the value of the on-shell action is the same as the original S (Ψ) to order y4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' In principle, terms containing Ψ2 ∝ y2 also contribute at this order.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, since any appearance of λ automatically comes with a factor y, there are no terms of order y4 containing λ as well as Ψ2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' All Ψ2-contributions are just the ones already present in S (Ψ) and were studied in detail in [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We see that to the first two leading orders, the actions S′ � ˜Ψ � and ˜S � ˜Ψ � = S (Ψ) give the same on-shell result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Since S′ � ˜Ψ � = S � ˜Ψ′� , where ˜Ψ′ = ∞ � n=1 Fn � ˜Ψ⊗n� = π1FP 1 1 − Ψ, (132) this suggests that the combination FP gives rise to a gauge transformation rather than a physically distinct solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' However, a full proof of this statement will be left for future publications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 5 Conclusion and outlook We succeeded in providing an explicit consistent description of OSFT with stubs and interestingly found two possible actions with the same equations of motion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The field redefinitions used to convert solutions of Witten OSFT to the new theory are given in explicit form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We hope that the analysis of solutions to the stubbed equations of motion can teach us more general properties of solutions to Maurer-Cartan equations for A∞- or L∞-algebras, in particular about the solutions of closed string field theory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' One way to proceed would be to transform the whole construction to the sliver frame, where many analytic solutions of OSFT are formulated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Another possible future direction is to examine wether the stubbed theory is “more well-behaved” in the sense that some typical singularities and ambiguities, for example connected to identity-like solutions, are ameliorated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Acknowledgements We thank Ted Erler, Jakub Vošmera, Branislav Jurčo, Igor Khavkine and Martin Markl for useful discussions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Our work has been funded by the Grant Agency of Czech Republic under the grant EXPRO 20-25775X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Appendix Tensor coalgebras The tensor coalgebra T V associated to a (graded) vector space V is defined as the Fock space V ⊗0 + V ⊗1 + V ⊗2 + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (133) together with the comultiplication ∆ : T V → T V ⊗′ T V given by ∆ (v1 ⊗ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ⊗ vn) = n � k=0 (v1 ⊗ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ⊗ vk) ⊗′ (vk+1 ⊗ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ⊗ vn) (134) 22 on homogeneous elements and extended by linearity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Here the vi are elements of V and ⊗′ denotes the tensor product arising from a comultiplication, in contrast to the usual ⊗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' We define the projection operator πn : T V → T V to project any element on its nth tensor power component, πnT V = V ⊗n (135) A linear map d : T V → T V is called a coderivation if it satisfies the co-Leibniz rule: ∆d = (d ⊗′ 1 + 1 ⊗′ d) ∆ (136) Linear combinations of coderivations are again coderivations as well as their graded commutator [d1, d2] = d1d2 − (−1)deg(d1)deg(d2) d2d1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (137) The product d1d2 is in general not a coderivation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' For any m-linear map dm : V ⊗m → V one can construct an associated coderivation by the formula d = ∞ � n=1 n−m � k=0 1⊗k ⊗ dm ⊗ 1⊗n−k−m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (138) The co-Leibniz rule guarantees that any coderivation is a sum of terms of this form for different m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' The individual m-products can be recovered as dm = π1dπm (139) If an odd coderivation d obeys d2 = 0 (140) then its components dm form an A∞-algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' A linear map f is called a cohomomorphism if it fulfills ∆f = (f ⊗′ f) ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (141) Linear combinations and products of cohomomorphisms are again cohomomorphisms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Given a family of m-products fm one can construct a unique cohomomorphism via f = ∞ � j=1 ∞ � k=1 � m1+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='+mj=k fm1 ⊗ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' ⊗ fmj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (142) Again, the individual products can be recovered from f as fm = π1fπm (143) Of special importance are elements of T V of the form 1 + v + v ⊗ v + v ⊗ v ⊗ v + .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' =: 1 1 − v (144) for some v ∈ V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' They fulfill the following useful properties: π1f 1 1 − v = ∞ � m=1 fm � v⊗m� (145) 23 f 1 1 − v = 1 1 − π1f 1 1−v (146) for any cohomomorphism f.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' A bilinear map ⟨ω|: T V × T V → C is called a symplectic form if it satisfies ⟨ω|v1 ⊗ v2 =: ω (v1, v2) = − (−1)deg(v1)deg(v2) ω (v2, v1) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (147) A multilinear product mk is called cyclic with respect to ω if it fulfills ω (v1, mk (v2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', vk+1)) = − (−1)deg(v1)deg(mk) ω (mk (v1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=', vk) , vk+1) (148) A coderivation d is cyclic if all of its components dm = π1dπm are cyclic or equivalently ⟨ω|π2d = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' (149) Given two symplectic forms ⟨ω|, ⟨ω′|, a cohomomorphism f is cyclic if ⟨ω′|π2f = ⟨ω|π2 (150) References [1] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Witten, “Noncommutative Geometry and String Field Theory,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' B 268 (1986), 253-294 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1016/0550-3213(86)90155-0 [2] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Vošmera, “Selected topics in string field theory and physics of D-branes,” [3] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Zwiebach, “Closed string field theory: Quantum action and the B-V master equation,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' B 390 (1993), 33-152 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1016/0550-3213(93)90388-6 [arXiv:hep-th/9206084 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [4] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Zwiebach, “A Proof that Witten’s open string theory gives a single cover of moduli space,” Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 142 (1991), 193-216 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1007/BF02099176 [5] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Sonoda and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Zwiebach, “COVARIANT CLOSED STRING THEORY CANNOT BE CUBIC,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' B 336 (1990), 185-221 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1016/0550-3213(90)90108-P [6] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Schnabl, “Analytic solution for tachyon condensation in open string field theory,” Adv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Theor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 10 (2006) no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4, 433-501 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4310/ATMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='v10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='n4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='a1 [arXiv:hep-th/0511286 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [7] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Erler, “Four Lectures on Closed String Field Theory,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Rept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 851 (2020), 1-36 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='physrep.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='01.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='003 [arXiv:1905.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='06785 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [8] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Erler, “The closed string field theory action vanishes,” JHEP 10 (2022), 055 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1007/JHEP10(2022)055 [arXiv:2204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='12863 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [9] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Takahashi and S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Tanimoto, “Marginal and scalar solutions in cubic open string field theory,” JHEP 03 (2002), 033 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1088/1126-6708/2002/03/033 [arXiv:hep-th/0202133 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [10] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Erler, “Four lectures on analytic solutions in open string field theory,” Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Rept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 980 (2022), 1-95 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='physrep.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='06.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='004 [arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='00521 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 24 [11] B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Zwiebach, “Oriented open - closed string theory revisited,” Annals Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 267 (1998), 193-248 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1006/aphy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1998.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='5803 [arXiv:hep-th/9705241 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [12] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Takezaki, “Open superstring field theory including the Ramond sector based on the supermoduli space,” [arXiv:1901.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='02176 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [13] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Erbin, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Maccaferri, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Schnabl and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Vošmera, “Classical algebraic structures in string theory effective actions,” JHEP 11 (2020), 123 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1007/JHEP11(2020)123 [arXiv:2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='16270 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [14] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Crainic;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' “On the perturbation lemma, and deformations,” [arXiv:math/0403266] [15] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Konopka, “The S-Matrix of superstring field theory,” JHEP 11 (2015), 187 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1007/JHEP11(2015)187 [arXiv:1507.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='08250 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [16] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Loday, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Vallette;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' “Algebraic operads,” Springer Berlin, Heidelberg, 2012 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1007/978-3-642- 30362-3 [17] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Li and K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Zeng, “Homotopy Algebras in Higher Spin Theory,” Adv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Theor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 24 (2020) no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='3, 757-819 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='4310/ATMP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='v24.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='n3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='a5 [arXiv:1807.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='06037 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [18] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Hata and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Zwiebach, “Developing the covariant Batalin-Vilkovisky approach to string theory,” Annals Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 229 (1994), 177-216 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1006/aphy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1994.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1006 [arXiv:hep-th/9301097 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [19] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Mukherji and A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Sen, “Some all order classical solutions in nonpolynomial closed string field theory,” Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' B 363 (1991), 639-664 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1016/0550-3213(91)80037-M [20] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Scheinpflug and M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Schnabl, “Conformal perturbation theory from open string field theory,” [arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='05216 [hep-th]].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' [21] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Okawa, “Analytic methods in open string field theory,” Prog.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Theor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content=' 128 (2012), 1001-1060 doi:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1143/PTP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='128.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
+page_content='1001 25' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/jtFPT4oBgHgl3EQf1TVB/content/2301.13182v1.pdf'}
diff --git a/kNFJT4oBgHgl3EQfYSw1/content/tmp_files/2301.11525v1.pdf.txt b/kNFJT4oBgHgl3EQfYSw1/content/tmp_files/2301.11525v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..4a3662d766c6b3d811a479d67e816db01fbf0375
--- /dev/null
+++ b/kNFJT4oBgHgl3EQfYSw1/content/tmp_files/2301.11525v1.pdf.txt
@@ -0,0 +1,1612 @@
+Mixed Attention Network for Hyperspectral Image Denoising
+Zeqiang Lai Ying Fu
+Beijing Institute of Technology
+{laizeqiang, fuying}@bit.edu.cn
+Abstract
+Hyperspectral image denoising is unique for the highly
+similar and correlated spectral information that should be
+properly considered. However, existing methods show lim-
+itations in exploring the spectral correlations across differ-
+ent bands and feature interactions within each band. Be-
+sides, the low- and high-level features usually exhibit differ-
+ent importance for different spatial-spectral regions, which
+is not fully explored for current algorithms as well. In this
+paper, we present a Mixed Attention Network (MAN) that
+simultaneously considers the inter- and intra-spectral cor-
+relations as well as the interactions between low- and high-
+level spatial-spectral meaningful features. Specifically, we
+introduce a multi-head recurrent spectral attention that ef-
+ficiently integrates the inter-spectral features across all the
+spectral bands. These features are further enhanced with
+a progressive spectral channel attention by exploring the
+intra-spectral relationships. Moreover, we propose an at-
+tentive skip-connection that adaptively controls the propor-
+tion of the low- and high-level spatial-spectral features from
+the encoder and decoder to better enhance the aggregated
+features. Extensive experiments show that our MAN out-
+performs existing state-of-the-art methods on simulated and
+real noise settings while maintaining a low cost of param-
+eters and running time.
+Code is available at https:
+//github.com/Zeqiang-Lai/MAN.
+1. Introduction
+Hyperspectral image (HSI) is made up of numerous
+bands across a wide range of the spectrum. Different from
+common color images, HSIs divide the spectrum into much
+more bands than three and their wavelengths can be ex-
+tended beyond the visible. Such properties make HSI es-
+pecially attractive and useful for the applications of remote
+sensing [3, 29, 41], face recognition [43, 45], classification
+[2, 6, 46], etc. However, due to the limitation of existing
+hyperspectral imaging techniques, the captured HSIs often
+suffer from severe corruption, which makes the develop-
+ment of robust denoising algorithms an urgent need.
+100
+101
+102
+103
+Running Time (sec)
+38.0
+39.0
+40.0
+41.0
+42.0
+42.8
+43.5
+PSNR (dB)
+BM4D
+TDL
+ITSReg
+WLRTR
+KBR
+NGmeet
+HSID
+QRNN3D
+T3SC
+GRUNet
+TRQ3D
+MAN-S
+MAN-M
+MAN-L
+Performance Comparsion of Different Methods
+Optimization-based
+Deep-learning-based
+Ours
+0.2M
+0.5M
+1.0M
+5.0M
+Figure 1. Speed and performance comparison. Our method out-
+performs state-of-the-art methods with low-cost parameters and
+runtime. Comparisons are performed on the ICVL dataset with
+blind Gaussian noise.
+Traditionally, optimization algorithms are often adopted
+to solve HSI denoising with different hand-crafted priors
+exploring the domain knowledge of the HSI, i.e., global
+correlation along the spectrum and spatial-spectral corre-
+lation. Typical priors that are extensively studied includes,
+total variation [31, 36], wavelet [23], low-rank [28, 33, 44],
+and etc. By considering the spectral and spatial redundancy,
+non-local patch-similarity [22] is also widely used in con-
+jugation with variable splitting algorithms [15] and tensor-
+based dictionary learning [26]. These methods generally
+have no requirement for a large amount of training data or
+even be training-free, but most of them are time-consuming
+(see Figure 1) and their performance is strongly correlated
+with the matching degree of handcrafted priors with the un-
+derlying characteristics of HSIs, which weakens their per-
+formance for real-world denoising with complex noise.
+Recent works on HSI denoising shift their attention to
+the learning-based approaches to model complex intrin-
+sic characteristics of HSI, e.g., inter-spectral correlations
+with sliding-window convolution [37], and local spatial-
+1
+arXiv:2301.11525v1 [cs.CV] 27 Jan 2023
+
+spectral correlations with 3D convolution [34]. Augmented
+with various commonly used techniques, e.g., residual con-
+nection [37], skip connection [13], and recurrent network
+[34], these methods have obtained much better results than
+optimization-based ones. Nevertheless, their performance
+and efficiency are still limited and the thorough considera-
+tion of inter- and intra-spectral correlations is still lacking.
+Besides, though most HSI restoration models [13, 34] em-
+ploy skip connections to prevent the loss of low-level in-
+formation, the development of such modules mostly stays
+at the original additive or concat versions. However, these
+vanilla ones either neglect the different importance of the
+features from encoder and decoder, or mix the tasks of fea-
+tures weighting and processing, which makes them less ef-
+fective for handling diverse information from different fea-
+ture channels and spectral bands. Therefore, it is foresee-
+able that whether a model can comprehensively take infor-
+mation from different aspects into account would be the key
+to further boosting the HSI denoising performance.
+In this paper, we propose a Mixed Attention Network
+(MAN) for hyperspectral image denoising, which simulta-
+neously considers the inter- and intra-spectral correlations
+as well as the interactions between low- and high-level
+spatial-spectral meaningful features. Specifically, we intro-
+duce a Multi-Head Recurrent Spectral Attention (MHRSA)
+block in place of vanilla spectral fusion methods [34,37]. It
+applies recurrent attention across the spectral dimension to
+dynamically mix the inter-spectral features from all spec-
+tral bands. MHRSA adopts two simple MLPs rather than
+dense convolutions to directly transform the input features
+into candidate values and attention maps, which makes it
+not only computationally efficient but also lightweight in
+terms of model parameters. Another important character-
+istic of our MHRSA is the efficient bi-directional context
+aggregation through a multi-head feature partition. This is
+achieved by splitting rather than repeating the features into
+two heads and performing parallel recurrent attention in re-
+verse directions. Apart from inter-spectral correlations, we
+propose a Progressive Spectral Channel Attention (PSCA)
+that sequentially applies the dynamic channel mixing and
+the static channel mixing to progressively mix the intra-
+spectral features for each band. The MHRSA and PSCA
+cooperate with each other and enable the informative fea-
+tures to propagate and interact along both inter-spectral and
+intra-spectral dimensions, which results in exceedingly dis-
+criminative features for subsequent reconstruction.
+Furthermore, we reexamine the skip connections of the
+U-shaped models [13], which are used to ensure the preser-
+vation of fine-grained details. By analyzing the underlying
+operations of the most commonly used additive and con-
+cat versions, we propose a novel Attentive Skip Connection
+(ASC) that adaptively combines the low- and high-level fea-
+tures from encoder and decoder. The ASC assigns element-
+wise fusion weight for each spatial-spectral location, which
+allows subsequent layers to selectively focus on more infor-
+mative regions, thus leading to better denoising results. We
+conduct comprehensive experiments and demonstrate the
+state-of-the-art performance of our MAN on various HSI
+denoising datasets under simulated and real-world noise.
+Our contributions are summarized as follows:
+• We propose a mixed attention network for hyperspec-
+tral image denoising, which simultaneously considers the
+inter- and intra-spectral as well as low- and high-level
+spatial-spectral feature correlations.
+• We introduce a multi-head spectral recurrent attention
+block to dynamically aggregate the inter-spectral features,
+and a progressive spectral channel attention block for in-
+tegrating the intra-spectral features.
+• We present an attentive skip connection to adaptively
+strengthen the important spatial-spectral meaningful fea-
+tures that flow forward from low- to high-level.
+2. Related Works
+The methods for HSI denoising could generally be
+divided into optimization-based methods, deep-learning
+methods, as well as hybrid methods. In this section, we
+provide an overview of their recent major approaches.
+2.1. Optimization-based Methods
+Traditional methods solve the HSI denoising by treat-
+ing it as an optimization problem, where they attempt to
+find unknown clean HSI by minimizing an optimization
+objective that incorporates the properties of spectrum and
+images. Such incorporation is generally achieved by de-
+signing different hand-crafted priors, e.g., total variation
+priors [31, 36], wavelet priors [23], and low-rank priors
+[28,31,33,44]. By considering the non-local self-similarity
+in the spectral and spatial dimensions, many works such as
+block-matching and 4-D filtering (BM4D) [22] and the ten-
+sor dictionary learning [26] are also proposed.
+These optimization-based HSI denoising methods are
+flexible to remove different types of noise [10, 16] and can
+be even extended to tasks beyond denoising [9, 14, 15, 25].
+However, their performance is significantly restricted by the
+matching degree of handcrafted prior and the underlying
+properties of HSI, which is difficult to ensure for complex
+real scenes. To address this problem, plug-and-play meth-
+ods [7,12,47] integrate the optimization-based method with
+a learning-based prior, i.e., a plug-and-play Gaussian de-
+noiser [11,40], to tackle complex noises that cannot be eas-
+ily modeled by handcrafted prior. Specifically, Liu et al [20]
+propose a fibered rank constrained tensor restoration frame-
+work and adopt BM3D [11] as an extra plug-and-play regu-
+larization. In [21], FFDNet [40] is used for local regularity,
+alongside the Kronecker-basis-representation-based tensor
+low-rankness for global structures regularity.
+2
+
+2.2. Deep-Learning-based Methods
+Deep-learning-based methods have gained superior pop-
+ularity in recent years. Inspired by 2D image denoising net-
+work DnCNN [39], Chang et al. [8] propose HSI-DeNet
+that learns multi-channel 2-D filters to model spectral cor-
+relation. Yuan et al. [37] introduce residual network struc-
+ture with a sliding window strategy for remote sensed HSI.
+To further exploit the spatial-spectral correlation, Dong et
+al. [13] designed a 3D U-net architecture. These methods
+have been successfully applied to different HSIs, but most
+of them are limited at exploring the inter-spectral correla-
+tions, which is significantly important for HSI denoising.
+To address such issue, Wei et al. [34] proposed a 3D convo-
+lutional quasi-recurrent neural network (QRNN3D), which
+combines the 3D convolution to extract spatial-spectral cor-
+related features, and a quasi-recurrent network to integrate
+information from different bands in a global perspective.
+Based on it, Lai et al. [19] propose GRUNet that further
+improves the performance with residual blocks.
+Recent
+works also explore some hybrid approaches by borrowing
+key techniques from different types of methods. For exam-
+ple, T3SC [4] is introduced as a hybrid method that com-
+bines sparse coding principles and deep neural networks.
+TRQ3DNet [24] augments the QRNN3D with a Uformer
+[32] block to enhance the capabilities for capturing long-
+range spatial dependency. Compared with these methods,
+our method considers not only the inter-spectral correlation
+with a more powerful multi-head recurrent spectral atten-
+tion but also the intra-spectral interaction with a progres-
+sive spectral channel attention block, which results in more
+discriminative features for the denoising. Moreover, we re-
+examine the low- and high-level feature fusion and propose
+an attentive skip connection for better feature aggregation,
+which is ignored in most existing works.
+3. Mixed Attention Network
+In this section, we first describe the overall architecture
+of our Mixed Attention Network (MAN). Then we pro-
+vide detailed illustrations for each attention block, including
+multi-head recurrent spectral attention, progressive spectral
+channel attention, and attentive skip connection.
+Overall Architecture.
+The overall architecture of the
+proposed MAN is shown in Figure 2. The input noisy image
+X ∈ RH×W ×S is sequentially processed by a multi-level
+encoder to obtain multi-scale image features, which is then
+decoded by a symmetrically designed decoder to obtain the
+reconstructed clean image. We build the entire network by
+stacking Mixed Attention Blocks (MAB) that consist of a
+3D convolution, a multi-head recurrent spectral attention,
+and a progressive spectral channel attention to explore the
+inter- and intra-spectral correlations. Each MAB takes in-
+put features F ∈ RH×W ×S×C and generates output fea-
+Progressive Spectral
+Channel Attention
+Multi-Head Recurrent
+Spectral Attention
+Conv3D
+Encoder
+Decoder
+Input
+Output
+Bottleneck
+Plain MAB
+Up/Down MAB
+Attentive Skip Connection
+(a) Mixed Attention Network
+(b) Mixed Attention Block
+!F
+F
+Figure 2. Overall architectures of the proposed mixed attention
+network and mixed attention block.
+tures ˆF ∈ R ˆ
+H× ˆ
+W ×S× ˆ
+C with the same spectral resolution
+S. Different from other encoder-decoder-based networks,
+MAN replaces the vanilla skip connection with an attentive
+one that adaptively controls which part of low-level features
+should flow forward from the encoder to the decoder.
+3.1. Multi-Head Recurrent Spectral Attention
+Different from RGB images, HSIs have a wide range of
+the spectrum, and the images across different bands share
+generally identical spatial content. Moreover, the pixel val-
+ues with the same spatial location across different bands are
+correlated with specific spectral patterns, which are deter-
+mined by the material of the object they belong to. This
+illustrates one important characteristic of HSI, i.e., inter-
+spectral correlation/similarity. Similar to non-local spatial
+correlation/similarity, it is straightforward to utilize inter-
+spectral correlations to perform denoising with methods like
+non-local means (NLM) [5]. However, unlike spatial simi-
+lar pixels, pixels along the spectrum have a different range
+of values, which makes naive NLM unsuitable as its averag-
+ing operation could break the spectral relationship of each
+pixel across the spectrum.
+To address the above issue, we propose a Multi-Head
+Recurrent Spectral Attention (MHRSA) block that dynam-
+ically computes the weights for averaging the pixels along
+the spectrum for each band. Each band is assigned a dif-
+ferent weight for averaging information from other bands,
+which avoids the destruction of spectral dependency. An il-
+lustration of the MHRSA block is given in Figure 3. Given
+a feature map F ∈ RH×W ×S×C extracted from the previ-
+ous 3D convolution, it first uses two MLPs followed by two
+different activation functions to transform the input features
+into the candidate features Z and merging weights W, as
+Z = tanh(MLP1(F))
+W = sigmoid(MLP2(F))
+MLP(X) = W1 · (tanh(W2 · X))
+(1)
+where W1, W2 ∈ RC×C. These processes can be equiv-
+alently considered as the query, key, and value projections
+in self-attention [30]. The one difference is that we directly
+3
+
+*
++
+1-
+*
+Conv
+Conv
+𝑀!
+F"#
+F$
+#
++
+F%#
+𝑋
+𝑌
+𝑊& ∈ ℝ'×&'
+𝑊) ∈ ℝ&'×'
+Dynamic Channel Mixing
+Static Channel Mixing
+Static
+Attention
+Static
+Attention
+Linear
+GELU
+Backward Head
+Forward Head
+Split
+Combine
+Split
+Combine
+Attention
+Attention
+𝐹 ∈ ℝ!×#×$×%
+H×W
+C
+𝑆
+𝑂 ∈ ℝ!×#×&×%
+H×W
+C
+𝑆
+𝑊 ∈ ℝ!×#×&×%
+'
+𝑊 ∈ ℝ!×#×&×%
+'
+MLP
+MLP
+⃐𝑍 ∈ ℝ!×#×&×%
+'
+⃑𝑍 ∈ ℝ!×#×(×%
+'
+Tanh
+Sigmoid
+𝑂 ∈ ℝ!×#×&×%
+'
+𝑂 ∈ ℝ!×#×&×%
+'
+Figure 3. Illustration of the multi-head recurrent spectral attention. MLP stands for two-layer linear projections with Tanh activation. C,
+S, HW denote the feature, spectral, and spatial dimensions, respectively.
+compute the attention weight instead of obtaining the atten-
+tion map through the covariance of key and query. The other
+is that we perform the attention operation through a recur-
+rent merging step that only requires linear memory and time
+complexity other than the quadratic ones, which makes our
+method more suitable for high dimensional HSI data.
+Specifically, the recurrent merging step for spectral mix-
+ing is performed by accumulating the candidate features Z
+for each band based on merging weight W as
+Oi = (1 − Wi) ⊙ Zi + Wi ⊙ Oi−1
+(2)
+where Oi, Zi, Wi are the output features, candidate features,
+and merging weights of ith band, respectively. It can be ob-
+served that such a merging step fuse the features from all the
+previous bands Zi, i < j for jth band. Thus, it correlates the
+inter-spectral features and can potentially utilize the infor-
+mation from cleaner bands for denoising noisier bands.
+Furthermore, though the above merging step progres-
+sively aggregates the inter-spectral features, each band can
+only see the features in one direction. It can lead to in-
+complete spectral context. To address the issue, we pro-
+pose a multi-head attention in which we equally split the
+input features into two parts and perform parallel merging
+for each part with different directions. With such a strategy,
+we could achieve feature aggregation with global spectral
+context without any extra computation and parameters.
+3.2. Progressive Spectral Channel Attention
+The MHRSA provides stronger capabilities for explor-
+ing the inter-spectral correlations, which makes our model
+powerful at borrowing the features from more informative
+bands for less informative ones. However, these advantages
+would be weakened if the features of each band itself are
+not discriminative enough. Therefore, we propose to fur-
+ther strengthen the features of each band by exploring the
+intra-spectral correlations with a Progressive Spectral Chan-
+nel Attention (PSCA) module, as shown in Figure 4.
+Our
+PSCA
+generalizes
+the
+band-wise
+Squeeze-
+Excitation (SE) [17] like channel attention to a pixel-wise
+operation with a progressive attention pipeline, which
+contains a Dynamic Channel Mixing (DCM) and a Static
+Channel Mixing (SCM) processes, as
+Y = SCM(GELU(DCM(X)))
+(3)
+where X, Y are the input and output feature maps.
+Static Channel Mixing.
+Different from SE-like chan-
+nel attention that performs channel-wise gating to con-
+trol which channel should pass more information to subse-
+quent layers, our SCM mix the information of all channels
+through a static attention map W ∈ RCin×Cout as,
+SCM(X) = X · W.
+(4)
+The attention map is jointly learned with the network train-
+ing and it encodes the dataset level prior for selecting more
+diverse and important information across different features.
+Dynamic Channel Mixing.
+To further strengthen the ca-
+pability at identifying the essential features for each input
+HSI. We propose DCM that adopts a pixel-wise dynamic
+weight for channel mixing. Specifically, DCM first com-
+putes a scaling weight S for each input through a linear pro-
+jection, W2 ∈ RC×C. Then, we rescale the static attention
+map W1 with the scaling weight through element-wise mul-
+tiplication for each pixel location. Finally, we perform the
+channel mixing with the new attention map as
+DCM(X) = W1 · S ⊙ X,
+(5)
+S = X · W2.
+(6)
+With these two channel mixing steps, we could obtain sub-
+stantially better representations for each spectral band, thus
+could lead to better performance.
+3.3. Attentive Skip Connection
+Skip connection is the key part that distinguishes the U-
+Net [13] from other network architectures. It provides a di-
+rect but effective way to recover the low-level information
+that is lost during the downsampling operations of the com-
+monly used encoder-decoder convolutional neural network.
+The most commonly used additive skip connection is im-
+plemented as an addition between encoder and decoder fea-
+tures at the same depth. Supposing features of encoder and
+4
+
+𝑋
+𝑌
+𝑊! ∈ ℝ"×!"
+𝑊$ ∈ ℝ!"×"
+Dynamic Channel Mixing
+Static Channel Mixing
+Static
+Attention
+Static
+Attention
+Linear
+GELU
+Figure 4. Illustration of progressive spectral channel attention. ⊙,
+⊗ denote element-wise- and original matrix multiplication.
+decoder at the ith depth level are Fi
+e and Fi
+d, respectively,
+the addition skip connection can be precisely described as
+Fi+1
+d
+= Fi
+d + Fi
+e.
+(7)
+The major problem of it lies in the same weight on the shal-
+low features from the encoder and highly processed features
+from the decoder. This could be problematic due to the im-
+balanced information density of different spectral bands and
+feature channels. For example, it would be more helpful to
+preserve more low-level features for cleaner spectral bands
+to retain the details while we might want more high-level
+features for noisy bands to properly remove all the noise.
+To address the issue, we propose an Attentive Skip Con-
+nection (ASC) that explicitly weights the features from dif-
+ferent sources using the attention weight computed from
+two convolution layers. By denoting the weight of ith depth
+level as Mi, the detailed formulation is
+F = LeakyReLU(Conv1x1([Fi
+d, Fi
+e]))
+Mi = σ(Conv3x3(F)),
+Fi+1
+d
+= (1 − Mi) ⊙ Fi
+d + Mi ⊙ Fi
+e,
+(8)
+where σ denotes the sigmoid function. A visualization of
+our attentive skip connection is shown in Figure 5. Our
+ASC adaptively strengthens the features of more important
+spatial locations and spectral bands through element-wise
+gating, thus leading to better reconstruction quality.
+Analysis and in-depth comparison.
+We provide an in-
+depth analysis of the difference between different types of
+skip connections. Typically, the output of the skip connec-
+tion module would be processed by the subsequent convo-
+lution layer. For additive skip connection, the computation
+can be formulated as,
+Fo = W × (Fe + Fd) = W × Fe + W × Fd.
+(9)
+where Fo is the output features from the subsequent convo-
+lution layer, W denotes the convolution kernel, × denotes
+convolution, and Fe and Fd are the features from the en-
+coder and the decoder. By using distributive law, we could
+know that we actually treat the features from the encoder
+and decoder as features in the same manifold, which are
+processed by the same convolution kernel.
+For concat skip connection, the formulation would be,
+Fo = W × [Fe, Fd] = W1 × Fe + W2 × Fd.
+(10)
+*
++
+1-
+*
+Conv
+Conv
+𝑀!
+F"#
+F$
+#
++
+F%#
+𝑋
+𝑌
+𝑊& ∈ ℝ'×&'
+𝑊) ∈ ℝ&'×'
+Dynamic Channel Mixing
+Static Channel Mixing
+Static
+Attention
+Static
+Attention
+Linear
+GELU
+Backward Head
+Forward Head
+Split
+Combine
+Split
+Combine
+Attention
+Attention
+𝐹 ∈ ℝ!×#×$×%
+H×W
+C
+𝑆
+𝑂 ∈ ℝ!×#×&×%
+H×W
+C
+𝑆
+𝑊 ∈ ℝ!×#×&×%
+'
+𝑊 ∈ ℝ!×#×&×%
+'
+MLP
+MLP
+⃐𝑍 ∈ ℝ!×#×&×%
+'
+⃑𝑍 ∈ ℝ!×#×(×%
+'
+Tanh
+Sigmoid
+𝑂 ∈ ℝ!×#×&×%
+'
+𝑂 ∈ ℝ!×#×&×%
+'
+Figure 5. Illustration of the attentive skip connection block.
+where W is a convolution that transforms features with 2C
+channels to C channels. According to the properties of con-
+volution, we could divide W into two parts, W1 and W2,
+which indicates that concat skip connection assigns differ-
+ent kernels for features from the encoder and decoder.
+From the above analysis, we know that additive skip
+connection averages the features from the encoder and de-
+coder extracted by the same convolution kernel. This might
+weaken the discriminative features due to the feature aver-
+age. On the other hand, the concat skip connection extracts
+feature using two different kernels and then performs fea-
+ture average. Thus, it is possible for it to generate features
+that avoid the problem of feature weakening in additive skip
+connection. However, its success requires two convolution
+kernels to simultaneously consider the feature extraction
+and reweighting, which is difficult to achieve. This not only
+includes the training difficulties but also the limitation of
+convolution itself for pixel-wise reweighting.
+To address the aforementioned issues, our ASC separates
+the feature reweighting and extraction into two steps as,
+Fo = W × (M ⊙ Fe + (1 − M) ⊙ Fd
+�
+��
+�
+feature reweighting
+)
+(11)
+where M is the pixel-wise weight for reweighting. Such de-
+composition has two advantages. First, it avoids the need
+to learn two very discriminative convolutions as the con-
+cat skip connection, which makes the network concentrate
+more on finding more useful features. Second, it explicitly
+reweights the features from encoder and decoder, which dis-
+tinguishes the more informative features to pass through the
+subsequent layers.
+4. Experiments
+In this section, we provide the experiments for simulated
+and real-world HSI denoising. An ablation study and dis-
+cussion of each component are also included.
+4.1. Datasets
+We conducted experiments of simulated HSI denoising
+with ICVL [1] and real-worlding HSI denoising with Urban
+and HSIDwRD [42]. ICVL contains 201 clean HSI in the
+resolution of 1392 × 1300 that divides the spectrum into
+31 spectral bands. Urban and HSIDwRD contain 1 HSI
+and 59 HSIs with 210 and 34 bands. Following [34], we
+split the ICVL into three parts, i.e., 100 images for train-
+ing, 51 for validation, and 50 for testing. For training, we
+5
+
+Sigma Metric
+Noisy
+ITSReg
+[35]
+WLRTR
+[9]
+KBR
+[27]
+NGmeet
+[15]
+HSID
+[37]
+QRNN3D
+[34]
+T3SC
+[4]
+GRUNet
+[19]
+TRQ3D
+[24]
+MAN-S
+(ours)
+MAN-M
+(ours)
+MAN-L
+(ours)
+30
+PSNR
+18.59
+41.48
+42.62
+41.48
+42.99
+41.72
+42.22
+42.36
+42.84
+43.25
+43.83
+44.07
+44.17
+SSIM
+0.110
+0.961
+0.988
+0.984
+0.989
+0.987
+0.988
+0.986
+0.989
+0.990
+0.991
+0.991
+0.991
+SAM
+0.807
+0.088
+0.056
+0.088
+0.050
+0.067
+0.062
+0.079
+0.052
+0.046
+0.043
+0.042
+0.041
+50
+PSNR
+14.15
+38.88
+39.72
+39.16
+40.26
+39.39
+40.15
+40.47
+40.75
+41.30
+41.60
+41.84
+41.94
+SSIM
+0.046
+0.941
+0.978
+0.974
+0.980
+0.980
+0.982
+0.980
+0.983
+0.985
+0.985
+0.986
+0.986
+SAM
+0.991
+0.098
+0.073
+0.100
+0.059
+0.083
+0.074
+0.087
+0.062
+0.053
+0.052
+0.050
+0.049
+70
+PSNR
+11.23
+36.71
+37.52
+36.71
+38.66
+37.77
+38.30
+39.05
+39.02
+39.86
+40.05
+40.32
+40.40
+SSIM
+0.025
+0.923
+0.967
+0.961
+0.974
+0.972
+0.974
+0.974
+0.977
+0.980
+0.980
+0.981
+0.981
+SAM
+1.105
+0.112
+0.095
+0.113
+0.067
+0.096
+0.094
+0.096
+0.080
+0.061
+0.060
+0.058
+0.057
+Blind
+PSNR
+17.34
+40.62
+41.66
+40.68
+42.23
+40.95
+41.37
+41.52
+42.03
+42.47
+43.08
+43.35
+43.44
+SSIM
+0.114
+0.953
+0.983
+0.979
+0.985
+0.984
+0.985
+0.983
+0.987
+0.988
+0.988
+0.989
+0.989
+SAM
+0.859
+0.087
+0.064
+0.080
+0.053
+0.072
+0.068
+0.085
+0.057
+0.054
+0.046
+0.045
+0.044
+(a) GT
+(b) WLRTR
+(c) KBR
+(d) NGmeet
+(e) HSID
+(f) QRNN3D
+(g) T3SC
+(h) GRUNet
+(i) TRQ3D
+(j) MAN-M
+Table 1. Simulated Gaussian noise removal results under several noise levels on ICVL. Blind suggests each image is corrupted by Gaussian
+noise with random sigma (ranged from 30 to 70). Visual results show the 20th band under noise level 50.
+process the original HSIs into multiple overlapped smaller
+data cubes via even-stride cropping. The spatial resolution
+of each cube is 64 × 64 and the spectral resolution remains
+unchanged. Random rotation and scaling are also employed
+to further augment the dataset. For testing, the main region
+in the size of 512 × 512 × 31 is used for ICVL, and 15
+images in HSIDwRD are randomly selected. We use pre-
+trained models of ICVL for Urban and HSIDwRD.
+4.2. Implementation Details
+Adam [18] optimizer is adopted to minimize the mean
+square error. We follow the training strategy as [34], with
+slight modifications on the setup of the learning rate. The
+strategy is briefly described here and we refer interested
+readers to [34] for more details. In short, the network is first
+trained on Gaussian noise with a fixed level of 50 for 20
+epochs and then finetuned with a random noise level from a
+given set for 40 epochs to produce the first-stage Gaussian
+denoising model. In the second stage, the complex denois-
+ing model is obtained via another 30 epochs of fine-tuning
+using the trained Gaussian denoising model. The learning
+rate is set to 1 × 10−3 at first, but decayed by 0.1 every 5 or
+10 epochs, and finally reduced to 1 × 10−5 for each stage.
+The batch size is set to 16.
+4.3. Results on Gaussian Noise
+For the experiments with Gaussian noise, we evaluate
+our model with different noise strengths, including 30, 50,
+70, and random strengths ranging from 30 to 70. The simu-
+lated noisy input is generated by adding zero-mean additive
+white Gaussian noise with a given variance. We use a single
+model to tackle different noise levels.
+For systematical evaluation, we compare our method
+with four traditional HSI denoising methods, i.e., ITSReg
+[35], KBR [27], WLRTR [9], and NGmeet [15], and five
+deep-learning ones, including HSID [37], QRNN3D [34],
+T3SC [4], GRUNet [19], and TRQ3D [24]. In the spirit
+of fairness, the hyperparameters in traditional methods are
+carefully tuned, and the deep-learning-based models are re-
+trained if needed.
+The quantitative and visual results are shown in Table
+1. The comparisons of runtime and model size are pro-
+vided in Table 5. It can be easily observed that our method
+achieves the best performance in all different settings with a
+large margin over the competing methods. Specifically, our
+MAN-L achieves over 0.6 dB PSNR improvement on aver-
+age on four different settings while maintaining compara-
+ble parameters and runtime. our MAN-S achieves over 0.3
+dB improvement against the best competing method, i.e.,
+TRQ3D, with even fewer parameters. Moreover, our model
+is better at preserving spectral fidelity than previous meth-
+ods, which is reflected by the improvement in SAM.
+4.4. Results on Complex Noise
+Unlike the simulated noisy HSI with Gaussian noise, the
+real HSI is usually corrupted by different types of complex
+noise, e.g., strip noise, impulse noise, and deadline noise.
+To evaluate the robustness of our method in real scenes,
+6
+
+Type
+Metric
+Noisy
+LRMR
+[38]
+LRTV
+[16]
+NMoG
+[10]
+TDTV
+[31]
+HSID
+[37]
+QRNN3D
+[34]
+T3SC
+[4]
+GRUNet
+[19]
+TRQ3D
+[24]
+MAN-S
+(ours)
+MAN-M
+(ours)
+G+Stripe
+PSNR
+17.80
+32.62
+33.49
+33.87
+37.67
+37.77
+42.35
+40.85
+42.39
+43.05
+44.24
+44.60
+SSIM
+0.159
+0.717
+0.905
+0.799
+0.940
+0.942
+0.976
+0.986
+0.991
+0.992
+0.993
+0.994
+SAM
+0.910
+0.187
+0.078
+0.265
+0.081
+0.104
+0.055
+0.072
+0.050
+0.043
+0.039
+0.038
+G+Deadline
+PSNR
+17.61
+31.83
+32.37
+32.87
+36.15
+37.65
+42.23
+39.54
+42.11
+42.95
+44.13
+44.51
+SSIM
+0.155
+0.709
+0.895
+0.797
+0.930
+0.940
+0.976
+0.983
+0.991
+0.992
+0.993
+0.994
+SAM
+0.917
+0.227
+0.115
+0.276
+0.099
+0.102
+0.056
+0.096
+0.050
+0.044
+0.040
+0.038
+G+Impulse
+PSNR
+14.80
+29.70
+31.56
+28.60
+36.67
+35.00
+39.23
+36.06
+40.70
+41.27
+41.88
+41.97
+SSIM
+0.114
+0.623
+0.871
+0.652
+0.935
+0.899
+0.945
+0.952
+0.985
+0.983
+0.982
+0.979
+SAM
+0.926
+0.311
+0.242
+0.486
+0.094
+0.174
+0.109
+0.203
+0.067
+0.075
+0.090
+0.095
+G+Mixture
+PSNR
+14.08
+28.68
+30.47
+27.31
+34.77
+34.05
+38.25
+34.48
+38.51
+40.27
+41.03
+41.18
+SSIM
+0.099
+0.608
+0.858
+0.632
+0.919
+0.888
+0.938
+0.946
+0.981
+0.983
+0.981
+0.979
+SAM
+0.944
+0.353
+0.287
+0.513
+0.113
+0.181
+0.107
+0.228
+0.081
+0.075
+0.092
+0.097
+(a) GT
+(b) LRTV
+(c) NMoG
+(d) TDTV
+(e) HSID
+(f) QRNN3D
+(g) T3SC
+(h) GRUNet
+(i) TRQ3D
+(j) MAN-M
+Table 2. Simulated complex noise removal results under several noise types on ICVL. Visual results show the 20th band under stripe noise.
+Metrics NMoG
+[10]
+TDTV
+[31]
+QRNN3D
+[34]
+HSID
+[37]
+HSIDwRD
+[42]
+MAN-M
+(ours)
+PSNR
+30.90
+31.14
+31.13
+31.05
+31.23
+31.36
+SSIM
+0.907
+0.881
+0.940
+0.934
+0.939
+0.940
+SAM
+1.761
+1.853
+0.094
+0.096
+0.092
+0.092
+Table 3. Quantitative results on HSIDwRD [42].
+we also conduct experiments with different combinations
+of complex noise, following the settings in [34].
+We compare our method with eight recently developed
+methods including five deep-learning-based methods, i.e.,
+HSID [37], QRNN3D [34], GRUNet [19], T3SC [4], and
+TRQ3D [24], as well as four optimization-based ones, i.e.,
+LRMR [38], LRTV [16], NMoG [10] and TDTV [31]. We
+choose a different set of optimization-based methods from
+the ones used for Gaussian noise because these methods
+only perform well for the noise settings they can solve.
+Table 2 gives the quantitative and visual comparison be-
+tween our method against the competing ones. It can be
+seen that our method obtain substantial improvement on
+various metrics, especially on PSNR, which is over 1.5 dB
+on average, and ranged from 1.03 ∼ 1.7 dB.
+4.5. Results on Real-world Noise
+Urban.
+Figure 6 provide the visual result of our model on
+real-world noisy remotely sensed HSI, i.e., Urban. It can
+be seen that our model could properly remove the heavy
+noise while retaining the details. QRNN3D removes the
+most noise but loses the details such as stripes on the roof.
+MHRSA
+ASC
+PSCA
+Params
+Runtime
+PSNR
+SAM
+-
+-
+-
+0.43
+0.25
+39.98
+0.072
+✓
+-
+-
+0.50
+0.46
+41.13
+0.058
+✓
+✓
+-
+0.81
+0.55
+41.48
+0.053
+✓
+✓
+✓
+0.89
+0.65
+41.84
+0.050
+Table 4. Ablation study of the proposed network modules.
+HSID tends to produce blurry results than Ours.
+HSIDwRD.
+We conduct experiments on the real natural
+HSI dataset [42] with the first two leading optimization-
+based and CNN-based methods from complex denoising
+and the method proposed in [42]. The quantitative results
+are shown in Table 3. It can be observed that our results
+outperform the other ones in terms of all metrics, which
+demonstrates the effectiveness of our methods.
+4.6. Discussions
+Ablation Study.
+To verify the effectiveness of each pro-
+posed component.
+We evaluate the PSNR improvement
+each component brings by adding them one by one, start-
+ing from a baseline model, i.e., UNet with 3D convolution
+and activation only. All the models are trained for the Gaus-
+sian denoising task and evaluated on 50 noise strength. The
+experimental results are shown in Table 4. As we can ob-
+serve, the introduction of MHRSA brings the most signif-
+icant improvement. This indicates the importance of ex-
+ploring inter-spectral correlations. The attentive skip con-
+nection contributes 0.35 dB improvement by taking into
+7
+
+(a) Noisy
+(b) NMoG [10]
+(c) TDTV [31]
+(d) HSID [37]
+(e) QRNN3D [34]
+(f) GRUNet [19]
+(g) MAN(Ours)
+Figure 6. Real world noise removal results at 207th band of Urban Dataset.
+Method
+ITSReg
+[35]
+WLRTR
+[9]
+KBR
+[27]
+NGmeet
+[15]
+HSID
+[37]
+QRNN3D
+[34]
+T3SC
+[4]
+GRUNet
+[19]
+TRQ3D
+[24]
+MAN-S
+(ours)
+MAN-M
+(ours)
+MAN-L
+(ours)
+PSNR
+38.88
+39.16
+38.70
+40.26
+39.39
+40.15
+40.47
+40.75
+41.30
+41.60
+41.84
+41.94
+Params (M)
+-
+-
+-
+-
+0.40
+0.86
+0.83
+14.2
+0.68
+0.50
+0.89
+1.39
+Runtime (s)
+907
+1600
+1755
+166
+0.48
+0.44
+0.95
+0.87
+0.47
+0.49
+0.65
+0.97
+Table 5. Performance-Params-Runtime comparisons with the SOTA methods. Evaluated with Gaussian denoising (sigma=50) on ICVL.
+Ground Truth
+Band 7
+Band 12
+Band 21
+Noisy Input
+Channel 4
+Channel 18
+Channel 27
+Channel 38
+Band 31
+Figure 7. Visualization of the attention map of our attentive skip
+connection across different bands and channels.
+0.20
+0.15
+0.10
+0.05
+0.00
+0
+10
+20
+30
+40
+Pixel j
+Band i
+30
+25
+20
+15
+10
+0.04
+0.03
+0.02
+0.01
+5
+0
+20
+15
+10
+Band y
+30
+25
+20
+15
+10
+Band x
+Figure 8. Visualization of the spectral attention map from our
+MHRSA. (a) The left part is the average attention map between
+different bands. (b) The right part is the attention distribution of
+15th band across different pixels.
+account low- and high-level features fusion.
+Our PSCA
+strengthens the features from MHRSA by considering intra-
+spectral relationships, which further provides 0.36 dB im-
+provement.
+Visualization of Attentive Skip Connection.
+We visual-
+ize the attention map of the first attentive skip connection
+block for one sample. As is shown in Figure 7, the attention
+map varies across different bands and channels, which indi-
+cates that the equal attention of vanilla additive one might
+be less effective. In particular, it can be seen that the net-
+work learns to pay attention to different regions across dif-
+ferent bands and channels, i.e., band12 and channel 21 for
+edges, and different sources, e.g., channel 18 for encoder,
+channel 38 for decoder.
+Visualization of Spectral Attention.
+Figure 8 shows the
+spectral attention map of one particular example HSI at the
+first MHRSA layer. It can be seen that our MHRSA pays at-
+tention to all bands with different weights, which is helpful
+for adaptively aggregating the useful features from different
+bands to assist the denoising. In general, the overall pattern
+is that each band pays more attention to the bands around
+it, but the attention distributions may vary with respect to
+pixels at different locations.
+5. Conclusion
+In this paper, we propose a mixed attention network
+for hyperspectral image denoising. Our method introduces
+several key components to properly explore the inter- and
+intra-spectral correlations as well as the low- and high-level
+spatial-spectral feature interactions. Specifically, these are
+achieved with a multi-head recurrent spectral attention that
+recurrently merges the features across different bands, a
+progressive channel attention that progressively mixes the
+different features within each band, and an attentive skip
+connection that aggregates the features from encoder and
+decoder with different importance weight. We perform ex-
+tensive experiments on simulated and real-world noise, and
+it shows that our method outperforms the existing state-of-
+the-art methods with significant improvement while main-
+taining a smaller model size and running time.
+Ethical considerations and future work.
+Our work has
+no ethical issues. In this work, we explore the proposed
+MHRSA with multi-head in two directions, but it is also
+possible to extend it to multi-axis in which we perform at-
+tention in different dimensions, e.g., channel and spectral.
+8
+
+References
+[1] Boaz Arad and Ohad Ben-Shahar. Sparse recovery of hyper-
+spectral signal from natural rgb images. In Proceedings of
+the European Conference on Computer Vision, pages 19–34.
+Springer, 2016. 5
+[2] Saeideh Ghanbari Azar, Saeed Meshgini, Tohid Yousefi
+Rezaii, and Soosan Beheshti. Hyperspectral image classi-
+fication based on sparse modeling of spectral blocks. Neuro-
+computing, 407:12–23, 2020. 1
+[3] George Alan Blackburn. Hyperspectral remote sensing of
+plant pigments. Journal of Experimental Botany, 58(4):855–
+867, 2007. 1
+[4] Th´eo Bodrito, Alexandre Zouaoui, Jocelyn Chanussot, and
+Julien Mairal.
+A trainable spectral-spatial sparse coding
+model for hyperspectral image restoration. 34:5430–5442,
+2021. 3, 6, 7, 8
+[5] Antoni Buades, Bartomeu Coll, and Jean-Michel Morel.
+Non-local means denoising.
+Image Processing On Line,
+1:208–212, 2011. 3
+[6] Xianghai Cao, Yiming Ge, Renjie Li, Jing Zhao, and
+Licheng Jiao. Hyperspectral imagery classification with deep
+metric learning. Neurocomputing, 356:217–227, 2019. 1
+[7] Stanley H Chan, Xiran Wang, and Omar A Elgendy. Plug-
+and-play admm for image restoration: Fixed-point conver-
+gence and applications.
+IEEE Transactions on Computa-
+tional Imaging, 3(1):84–98, 2016. 2
+[8] Yi Chang, Luxin Yan, Houzhang Fang, Sheng Zhong, and
+Wenshan Liao. Hsi-denet: Hyperspectral image restoration
+via convolutional neural network.
+IEEE Transactions on
+Geoscience and Remote Sensing, 57(2):667–682, 2018. 3
+[9] Yi Chang, Luxin Yan, Xi-Le Zhao, Houzhang Fang, Zhijun
+Zhang, and Sheng Zhong. Weighted low-rank tensor recov-
+ery for hyperspectral image restoration. IEEE Transactions
+on Cybernetics, 50(11):4558–4572, 2020. 2, 6, 8
+[10] Yang Chen, Xiangyong Cao, Qian Zhao, Deyu Meng, and
+Zongben Xu.
+Denoising hyperspectral image with non-
+iid noise structure.
+IEEE Transactions on Cybernetics,
+48(3):1054–1066, 2017. 2, 7, 8
+[11] Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and
+Karen Egiazarian. Image denoising by sparse 3-d transform-
+domain collaborative filtering. IEEE Transactions on Image
+Processing, 16(8):2080–2095, 2007. 2
+[12] Aram Danielyan, Vladimir Katkovnik, and Karen Egiazar-
+ian. Image deblurring by augmented lagrangian with bm3d
+frame prior. In Proceedings of the Workshop on Information
+Theoretic Methods in Science and Engineering, volume 1,
+2010. 2
+[13] Weisheng Dong, Huan Wang, Fangfang Wu, Guangming
+Shi, and Xin Li. Deep spatial–spectral representation learn-
+ing for hyperspectral image denoising. IEEE Transactions
+on Computational Imaging, 5(4):635–648, 2019. 2, 3, 4
+[14] Ying Fu, Antony Lam, Imari Sato, and Yoichi Sato. Adap-
+tive spatial-spectral dictionary learning for hyperspectral im-
+age restoration. International Journal of Computer Vision,
+122(2):228–245, 2017. 2
+[15] Wei He, Quanming Yao, Chao Li, Naoto Yokoya, and Qibin
+Zhao. Non-local meets global: An integrated paradigm for
+hyperspectral denoising. In Proceedings of the IEEE Con-
+ference on Computer Vision and Pattern Recognition, pages
+6868–6877, 2019. 1, 2, 6, 8
+[16] Wei He, Hongyan Zhang, Liangpei Zhang, and Huanfeng
+Shen. Total-variation-regularized low-rank matrix factoriza-
+tion for hyperspectral image restoration. IEEE Transactions
+on Geoscience and Remote Sensing, 54(1):178–188, 2015.
+2, 7
+[17] Jie Hu, Li Shen, and Gang Sun. Squeeze-and-excitation net-
+works. In Proceedings of the IEEE conference on computer
+vision and pattern recognition, pages 7132–7141, 2018. 4
+[18] Diederik P Kingma and Jimmy Ba. Adam: A method for
+stochastic optimization.
+arXiv preprint arXiv:1412.6980,
+2014. 6
+[19] Zeqiang Lai, Kaixuan Wei, and Ying Fu. Deep plug-and-play
+prior for hyperspectral image restoration. Neurocomputing,
+481:281–293, 2022. 3, 6, 7, 8
+[20] YunYang Liu, XiLe Zhao, YuBang Zheng, TianHui Ma, and
+Hongyan Zhang. Hyperspectral image restoration by tensor
+fibered rank constrained optimization and plug-and-play reg-
+ularization. IEEE Transactions on Geoscience and Remote
+Sensing, pages 1–17, 2021. 2
+[21] TianHui Ma, Zongben Xu, Deyu Meng, and XiLe Zhao.
+Hyperspectral image restoration combining intrinsic image
+characterization with robust noise modeling. IEEE Journal
+of Selected Topics in Applied Earth Observations and Re-
+mote Sensing, 14:1628–1644, 2020. 2
+[22] Matteo Maggioni, Vladimir Katkovnik, Karen Egiazarian,
+and Alessandro Foi. Nonlocal transform-domain filter for
+volumetric data denoising and reconstruction. IEEE Trans-
+actions on Image Processing, 22(1):119–133, 2012. 1, 2
+[23] Hisham Othman and Shen-En Qian. Noise reduction of hy-
+perspectral imagery using hybrid spatial-spectral derivative-
+domain wavelet shrinkage.
+IEEE Transactions on Geo-
+science and Remote Sensing, 44(2):397–408, 2006. 1, 2
+[24] Li Pang, Weizhen Gu, and Xiangyong Cao. Trq3dnet: A
+3d quasi-recurrent and transformer based network for hyper-
+spectral image denoising. 14(18):4598, 2022. 3, 6, 7, 8
+[25] J. Peng, Q. Xie, Q. Zhao, Y. Wang, L. Yee, and D. Meng.
+Enhanced 3dtv regularization and its applications on hsi de-
+noising and compressed sensing. IEEE Transactions on Im-
+age Processing, 29:7889–7903, 2020. 2
+[26] Yi Peng, Deyu Meng, Zongben Xu, Chenqiang Gao, Yi
+Yang, and Biao Zhang. Decomposable nonlocal tensor dic-
+tionary learning for multispectral image denoising. In Pro-
+ceedings of the IEEE Conference on Computer Vision and
+Pattern Recognition, pages 2949–2956, 2014. 1, 2
+[27] Xie Qi,
+Zhao Qian,
+Deyu Meng,
+and Zongben Xu.
+Kronecker-basis-representation based tensor sparsity and its
+applications to tensor recovery. IEEE Transactions on Pat-
+tern Analysis and Machine Intelligence, PP(99):1–1, 2017.
+6, 8
+[28] Le Sun, Byeungwoo Jeon, Yuhui Zheng, and Zebin Wu. Hy-
+perspectral image restoration using low-rank representation
+on spectral difference image. IEEE Geoscience and Remote
+Sensing Letters, 14(7):1151–1155, 2017. 1, 2
+[29] Prasad S Thenkabail and John G Lyon. Hyperspectral remote
+sensing of vegetation. CRC Press, 2016. 1
+9
+
+[30] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko-
+reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia
+Polosukhin. Attention is all you need. Advances in neural
+information processing systems, 30, 2017. 3
+[31] Yao Wang, Jiangjun Peng, Qian Zhao, Yee Leung, Xi-Le
+Zhao, and Deyu Meng. Hyperspectral image restoration via
+total variation regularized low-rank tensor decomposition.
+IEEE Journal of Selected Topics in Applied Earth Observa-
+tions and Remote Sensing, 11(4):1227–1243, 2017. 1, 2, 7,
+8
+[32] Zhendong Wang, Xiaodong Cun, Jianmin Bao, Wengang
+Zhou, Jianzhuang Liu, and Houqiang Li. Uformer: A general
+u-shaped transformer for image restoration. In Proceedings
+of the IEEE/CVF Conference on Computer Vision and Pat-
+tern Recognition, pages 17683–17693, 2022. 3
+[33] Kaixuan Wei and Ying Fu. Low-rank bayesian tensor factor-
+ization for hyperspectral image denoising. Neurocomputing,
+331:412–423, 2019. 1, 2
+[34] Kaixuan Wei, Ying Fu, and Hua Huang. 3-d quasi-recurrent
+neural network for hyperspectral image denoising.
+IEEE
+Transactions on Neural Networks and Learning Systems,
+32(1):363–375, 2021. 2, 3, 5, 6, 7, 8
+[35] Qi Xie, Qian Zhao, Deyu Meng, Zongben Xu, Shuhang Gu,
+Wangmeng Zuo, and Lei Zhang. Multispectral images de-
+noising by intrinsic tensor sparsity regularization. In Pro-
+ceedings of the IEEE Conference on Computer Vision and
+Pattern Recognition, pages 1692–1700, 2016. 6, 8
+[36] Qiangqiang Yuan, Liangpei Zhang, and Huanfeng Shen. Hy-
+perspectral image denoising employing a spectral–spatial
+adaptive total variation model. IEEE Transactions on Geo-
+science and Remote Sensing, 50(10):3660–3677, 2012. 1,
+2
+[37] Qiangqiang Yuan, Qiang Zhang, Jie Li, Huanfeng Shen, and
+Liangpei Zhang.
+Hyperspectral image denoising employ-
+ing a spatial–spectral deep residual convolutional neural net-
+work. IEEE Transactions on Geoscience and Remote Sens-
+ing, 57(2):1205–1218, 2018. 1, 2, 3, 6, 7, 8
+[38] Hongyan Zhang, Wei He, Liangpei Zhang, Huanfeng Shen,
+and Qiangqiang Yuan. Hyperspectral image restoration using
+low-rank matrix recovery. IEEE Transactions on Geoscience
+and Remote Sensing, 52(8):4729–4743, 2013. 7
+[39] Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and
+Lei Zhang. Beyond a gaussian denoiser: Residual learning of
+deep cnn for image denoising. IEEE Transactions on Image
+Processing, 26(7):3142–3155, 2017. 3
+[40] Kai Zhang, Wangmeng Zuo, and Lei Zhang. Ffdnet: Toward
+a fast and flexible solution for cnn-based image denoising.
+IEEE Transactions on Image Processing, 27(9):4608–4622,
+2018. 2
+[41] Liangpei Zhang and Xin Huang. Object-oriented subspace
+analysis for airborne hyperspectral remote sensing imagery.
+Neurocomputing, 73(4-6):927–936, 2010. 1
+[42] Tao Zhang, Ying Fu, and Cheng Li. Hyperspectral image de-
+noising with realistic data. In Proceedings of the IEEE/CVF
+International Conference on Computer Vision, pages 2248–
+2257, 2021. 5, 7
+[43] Haitao Zhao and Shaoyuan Sun.
+Sparse tensor embed-
+ding based multispectral face recognition. Neurocomputing,
+133:427–436, 2014. 1
+[44] XiLe Zhao, Hao Zhang, TaiXiang Jiang, Michael K Ng, and
+Xiong-Jun Zhang. Fast algorithm with theoretical guarantees
+for constrained low-tubal-rank tensor recovery in hyperspec-
+tral images denoising. Neurocomputing, 413:397–409, 2020.
+1, 2
+[45] Zhihong Pan, G. Healey, M. Prasad, and B. Tromberg. Face
+recognition in hyperspectral images. IEEE Transactions on
+Pattern Analysis and Machine Intelligence, 25(12):1552–
+1560, 2003. 1
+[46] Feng Zhou, Renlong Hang, Qingshan Liu, and Xiaotong
+Yuan.
+Hyperspectral image classification using spectral-
+spatial lstms. Neurocomputing, 328:39–47, 2019. 1
+[47] Daniel Zoran and Yair Weiss. From learning models of natu-
+ral image patches to whole image restoration. In Proceedings
+of the IEEE International Conference on Computer Vision,
+pages 479–486. IEEE, 2011. 2
+10
+
diff --git a/kNFJT4oBgHgl3EQfYSw1/content/tmp_files/load_file.txt b/kNFJT4oBgHgl3EQfYSw1/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e625bc1f5a0145c4cecb6631bf694df7ba59fc6d
--- /dev/null
+++ b/kNFJT4oBgHgl3EQfYSw1/content/tmp_files/load_file.txt
@@ -0,0 +1,874 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf,len=873
+page_content='Mixed Attention Network for Hyperspectral Image Denoising Zeqiang Lai Ying Fu Beijing Institute of Technology {laizeqiang, fuying}@bit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='cn Abstract Hyperspectral image denoising is unique for the highly similar and correlated spectral information that should be properly considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, existing methods show lim- itations in exploring the spectral correlations across differ- ent bands and feature interactions within each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Be- sides, the low- and high-level features usually exhibit differ- ent importance for different spatial-spectral regions, which is not fully explored for current algorithms as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In this paper, we present a Mixed Attention Network (MAN) that simultaneously considers the inter- and intra-spectral cor- relations as well as the interactions between low- and high- level spatial-spectral meaningful features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, we introduce a multi-head recurrent spectral attention that ef- ficiently integrates the inter-spectral features across all the spectral bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' These features are further enhanced with a progressive spectral channel attention by exploring the intra-spectral relationships.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Moreover, we propose an at- tentive skip-connection that adaptively controls the propor- tion of the low- and high-level spatial-spectral features from the encoder and decoder to better enhance the aggregated features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Extensive experiments show that our MAN out- performs existing state-of-the-art methods on simulated and real noise settings while maintaining a low cost of param- eters and running time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Code is available at https: //github.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='com/Zeqiang-Lai/MAN.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Introduction Hyperspectral image (HSI) is made up of numerous bands across a wide range of the spectrum.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Different from common color images, HSIs divide the spectrum into much more bands than three and their wavelengths can be ex- tended beyond the visible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Such properties make HSI es- pecially attractive and useful for the applications of remote sensing [3, 29, 41], face recognition [43, 45], classification [2, 6, 46], etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, due to the limitation of existing hyperspectral imaging techniques, the captured HSIs often suffer from severe corruption, which makes the develop- ment of robust denoising algorithms an urgent need.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 100 101 102 103 Running Time (sec) 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='8 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='5 PSNR (dB) BM4D TDL ITSReg WLRTR KBR NGmeet HSID QRNN3D T3SC GRUNet TRQ3D MAN-S MAN-M MAN-L Performance Comparsion of Different Methods Optimization-based Deep-learning-based Ours 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='2M 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='5M 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0M 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='0M Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Speed and performance comparison.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our method out- performs state-of-the-art methods with low-cost parameters and runtime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Comparisons are performed on the ICVL dataset with blind Gaussian noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Traditionally, optimization algorithms are often adopted to solve HSI denoising with different hand-crafted priors exploring the domain knowledge of the HSI, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', global correlation along the spectrum and spatial-spectral corre- lation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Typical priors that are extensively studied includes, total variation [31, 36], wavelet [23], low-rank [28, 33, 44], and etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' By considering the spectral and spatial redundancy, non-local patch-similarity [22] is also widely used in con- jugation with variable splitting algorithms [15] and tensor- based dictionary learning [26].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' These methods generally have no requirement for a large amount of training data or even be training-free, but most of them are time-consuming (see Figure 1) and their performance is strongly correlated with the matching degree of handcrafted priors with the un- derlying characteristics of HSIs, which weakens their per- formance for real-world denoising with complex noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Recent works on HSI denoising shift their attention to the learning-based approaches to model complex intrin- sic characteristics of HSI, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', inter-spectral correlations with sliding-window convolution [37], and local spatial- 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='11525v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='CV] 27 Jan 2023 spectral correlations with 3D convolution [34].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Augmented with various commonly used techniques, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', residual con- nection [37], skip connection [13], and recurrent network [34], these methods have obtained much better results than optimization-based ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Nevertheless, their performance and efficiency are still limited and the thorough considera- tion of inter- and intra-spectral correlations is still lacking.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Besides, though most HSI restoration models [13, 34] em- ploy skip connections to prevent the loss of low-level in- formation, the development of such modules mostly stays at the original additive or concat versions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, these vanilla ones either neglect the different importance of the features from encoder and decoder, or mix the tasks of fea- tures weighting and processing, which makes them less ef- fective for handling diverse information from different fea- ture channels and spectral bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Therefore, it is foresee- able that whether a model can comprehensively take infor- mation from different aspects into account would be the key to further boosting the HSI denoising performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In this paper, we propose a Mixed Attention Network (MAN) for hyperspectral image denoising, which simulta- neously considers the inter- and intra-spectral correlations as well as the interactions between low- and high-level spatial-spectral meaningful features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, we intro- duce a Multi-Head Recurrent Spectral Attention (MHRSA) block in place of vanilla spectral fusion methods [34,37].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It applies recurrent attention across the spectral dimension to dynamically mix the inter-spectral features from all spec- tral bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' MHRSA adopts two simple MLPs rather than dense convolutions to directly transform the input features into candidate values and attention maps, which makes it not only computationally efficient but also lightweight in terms of model parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Another important character- istic of our MHRSA is the efficient bi-directional context aggregation through a multi-head feature partition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' This is achieved by splitting rather than repeating the features into two heads and performing parallel recurrent attention in re- verse directions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Apart from inter-spectral correlations, we propose a Progressive Spectral Channel Attention (PSCA) that sequentially applies the dynamic channel mixing and the static channel mixing to progressively mix the intra- spectral features for each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The MHRSA and PSCA cooperate with each other and enable the informative fea- tures to propagate and interact along both inter-spectral and intra-spectral dimensions, which results in exceedingly dis- criminative features for subsequent reconstruction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Furthermore, we reexamine the skip connections of the U-shaped models [13], which are used to ensure the preser- vation of fine-grained details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' By analyzing the underlying operations of the most commonly used additive and con- cat versions, we propose a novel Attentive Skip Connection (ASC) that adaptively combines the low- and high-level fea- tures from encoder and decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The ASC assigns element- wise fusion weight for each spatial-spectral location, which allows subsequent layers to selectively focus on more infor- mative regions, thus leading to better denoising results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We conduct comprehensive experiments and demonstrate the state-of-the-art performance of our MAN on various HSI denoising datasets under simulated and real-world noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our contributions are summarized as follows: We propose a mixed attention network for hyperspec- tral image denoising, which simultaneously considers the inter- and intra-spectral as well as low- and high-level spatial-spectral feature correlations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We introduce a multi-head spectral recurrent attention block to dynamically aggregate the inter-spectral features, and a progressive spectral channel attention block for in- tegrating the intra-spectral features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We present an attentive skip connection to adaptively strengthen the important spatial-spectral meaningful fea- tures that flow forward from low- to high-level.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Related Works The methods for HSI denoising could generally be divided into optimization-based methods, deep-learning methods, as well as hybrid methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In this section, we provide an overview of their recent major approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Optimization-based Methods Traditional methods solve the HSI denoising by treat- ing it as an optimization problem, where they attempt to find unknown clean HSI by minimizing an optimization objective that incorporates the properties of spectrum and images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Such incorporation is generally achieved by de- signing different hand-crafted priors, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', total variation priors [31, 36], wavelet priors [23], and low-rank priors [28,31,33,44].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' By considering the non-local self-similarity in the spectral and spatial dimensions, many works such as block-matching and 4-D filtering (BM4D) [22] and the ten- sor dictionary learning [26] are also proposed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' These optimization-based HSI denoising methods are flexible to remove different types of noise [10, 16] and can be even extended to tasks beyond denoising [9, 14, 15, 25].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, their performance is significantly restricted by the matching degree of handcrafted prior and the underlying properties of HSI, which is difficult to ensure for complex real scenes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To address this problem, plug-and-play meth- ods [7,12,47] integrate the optimization-based method with a learning-based prior, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', a plug-and-play Gaussian de- noiser [11,40], to tackle complex noises that cannot be eas- ily modeled by handcrafted prior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, Liu et al [20] propose a fibered rank constrained tensor restoration frame- work and adopt BM3D [11] as an extra plug-and-play regu- larization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In [21], FFDNet [40] is used for local regularity, alongside the Kronecker-basis-representation-based tensor low-rankness for global structures regularity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Deep-Learning-based Methods Deep-learning-based methods have gained superior pop- ularity in recent years.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Inspired by 2D image denoising net- work DnCNN [39], Chang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' [8] propose HSI-DeNet that learns multi-channel 2-D filters to model spectral cor- relation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Yuan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' [37] introduce residual network struc- ture with a sliding window strategy for remote sensed HSI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To further exploit the spatial-spectral correlation, Dong et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' [13] designed a 3D U-net architecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' These methods have been successfully applied to different HSIs, but most of them are limited at exploring the inter-spectral correla- tions, which is significantly important for HSI denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To address such issue, Wei et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' [34] proposed a 3D convo- lutional quasi-recurrent neural network (QRNN3D), which combines the 3D convolution to extract spatial-spectral cor- related features, and a quasi-recurrent network to integrate information from different bands in a global perspective.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Based on it, Lai et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' [19] propose GRUNet that further improves the performance with residual blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Recent works also explore some hybrid approaches by borrowing key techniques from different types of methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For exam- ple, T3SC [4] is introduced as a hybrid method that com- bines sparse coding principles and deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' TRQ3DNet [24] augments the QRNN3D with a Uformer [32] block to enhance the capabilities for capturing long- range spatial dependency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Compared with these methods, our method considers not only the inter-spectral correlation with a more powerful multi-head recurrent spectral atten- tion but also the intra-spectral interaction with a progres- sive spectral channel attention block, which results in more discriminative features for the denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Moreover, we re- examine the low- and high-level feature fusion and propose an attentive skip connection for better feature aggregation, which is ignored in most existing works.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Mixed Attention Network In this section, we first describe the overall architecture of our Mixed Attention Network (MAN).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Then we pro- vide detailed illustrations for each attention block, including multi-head recurrent spectral attention, progressive spectral channel attention, and attentive skip connection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Overall Architecture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The overall architecture of the proposed MAN is shown in Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The input noisy image X ∈ RH×W ×S is sequentially processed by a multi-level encoder to obtain multi-scale image features, which is then decoded by a symmetrically designed decoder to obtain the reconstructed clean image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We build the entire network by stacking Mixed Attention Blocks (MAB) that consist of a 3D convolution, a multi-head recurrent spectral attention, and a progressive spectral channel attention to explore the inter- and intra-spectral correlations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Each MAB takes in- put features F ∈ RH×W ×S×C and generates output fea- Progressive Spectral Channel Attention Multi-Head Recurrent Spectral Attention Conv3D Encoder Decoder Input Output Bottleneck Plain MAB Up/Down MAB Attentive Skip Connection (a) Mixed Attention Network (b) Mixed Attention Block !' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='F F Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Overall architectures of the proposed mixed attention network and mixed attention block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' tures ˆF ∈ R ˆ H× ˆ W ×S× ˆ C with the same spectral resolution S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Different from other encoder-decoder-based networks, MAN replaces the vanilla skip connection with an attentive one that adaptively controls which part of low-level features should flow forward from the encoder to the decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Multi-Head Recurrent Spectral Attention Different from RGB images, HSIs have a wide range of the spectrum, and the images across different bands share generally identical spatial content.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Moreover, the pixel val- ues with the same spatial location across different bands are correlated with specific spectral patterns, which are deter- mined by the material of the object they belong to.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' This illustrates one important characteristic of HSI, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', inter- spectral correlation/similarity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Similar to non-local spatial correlation/similarity, it is straightforward to utilize inter- spectral correlations to perform denoising with methods like non-local means (NLM) [5].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, unlike spatial simi- lar pixels, pixels along the spectrum have a different range of values, which makes naive NLM unsuitable as its averag- ing operation could break the spectral relationship of each pixel across the spectrum.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To address the above issue, we propose a Multi-Head Recurrent Spectral Attention (MHRSA) block that dynam- ically computes the weights for averaging the pixels along the spectrum for each band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Each band is assigned a dif- ferent weight for averaging information from other bands, which avoids the destruction of spectral dependency.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' An il- lustration of the MHRSA block is given in Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Given a feature map F ∈ RH×W ×S×C extracted from the previ- ous 3D convolution, it first uses two MLPs followed by two different activation functions to transform the input features into the candidate features Z and merging weights W, as Z = tanh(MLP1(F)) W = sigmoid(MLP2(F)) MLP(X) = W1 · (tanh(W2 · X)) (1) where W1, W2 ∈ RC×C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' These processes can be equiv- alently considered as the query, key, and value projections in self-attention [30].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The one difference is that we directly 3 + 1- Conv Conv 𝑀!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' F"# F$ # + F%# 𝑋 𝑌 𝑊& ∈ ℝ\'×&\' 𝑊) ∈ ℝ&\'×\' Dynamic Channel Mixing Static Channel Mixing Static Attention Static Attention Linear GELU Backward Head Forward Head Split Combine Split Combine Attention Attention 𝐹 ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='×#×$×% H×W C 𝑆 𝑂 ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='×#×&×% H×W C 𝑆 𝑊 ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' 𝑊 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' MLP MLP ⃐𝑍 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' ⃑𝑍 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×(×% ' Tanh Sigmoid 𝑂 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' 𝑂 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' Figure 3." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Illustration of the multi-head recurrent spectral attention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' MLP stands for two-layer linear projections with Tanh activation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' C, S, HW denote the feature, spectral, and spatial dimensions, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' compute the attention weight instead of obtaining the atten- tion map through the covariance of key and query.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The other is that we perform the attention operation through a recur- rent merging step that only requires linear memory and time complexity other than the quadratic ones, which makes our method more suitable for high dimensional HSI data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, the recurrent merging step for spectral mix- ing is performed by accumulating the candidate features Z for each band based on merging weight W as Oi = (1 − Wi) ⊙ Zi + Wi ⊙ Oi−1 (2) where Oi, Zi, Wi are the output features, candidate features, and merging weights of ith band, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can be ob- served that such a merging step fuse the features from all the previous bands Zi, i < j for jth band.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Thus, it correlates the inter-spectral features and can potentially utilize the infor- mation from cleaner bands for denoising noisier bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Furthermore, though the above merging step progres- sively aggregates the inter-spectral features, each band can only see the features in one direction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can lead to in- complete spectral context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To address the issue, we pro- pose a multi-head attention in which we equally split the input features into two parts and perform parallel merging for each part with different directions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' With such a strategy, we could achieve feature aggregation with global spectral context without any extra computation and parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Progressive Spectral Channel Attention The MHRSA provides stronger capabilities for explor- ing the inter-spectral correlations, which makes our model powerful at borrowing the features from more informative bands for less informative ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, these advantages would be weakened if the features of each band itself are not discriminative enough.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Therefore, we propose to fur- ther strengthen the features of each band by exploring the intra-spectral correlations with a Progressive Spectral Chan- nel Attention (PSCA) module, as shown in Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our PSCA generalizes the band-wise Squeeze- Excitation (SE) [17] like channel attention to a pixel-wise operation with a progressive attention pipeline, which contains a Dynamic Channel Mixing (DCM) and a Static Channel Mixing (SCM) processes, as Y = SCM(GELU(DCM(X))) (3) where X, Y are the input and output feature maps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Static Channel Mixing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Different from SE-like chan- nel attention that performs channel-wise gating to con- trol which channel should pass more information to subse- quent layers, our SCM mix the information of all channels through a static attention map W ∈ RCin×Cout as, SCM(X) = X · W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (4) The attention map is jointly learned with the network train- ing and it encodes the dataset level prior for selecting more diverse and important information across different features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Dynamic Channel Mixing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To further strengthen the ca- pability at identifying the essential features for each input HSI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We propose DCM that adopts a pixel-wise dynamic weight for channel mixing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, DCM first com- putes a scaling weight S for each input through a linear pro- jection, W2 ∈ RC×C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Then, we rescale the static attention map W1 with the scaling weight through element-wise mul- tiplication for each pixel location.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Finally, we perform the channel mixing with the new attention map as DCM(X) = W1 · S ⊙ X, (5) S = X · W2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (6) With these two channel mixing steps, we could obtain sub- stantially better representations for each spectral band, thus could lead to better performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Attentive Skip Connection Skip connection is the key part that distinguishes the U- Net [13] from other network architectures.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It provides a di- rect but effective way to recover the low-level information that is lost during the downsampling operations of the com- monly used encoder-decoder convolutional neural network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The most commonly used additive skip connection is im- plemented as an addition between encoder and decoder fea- tures at the same depth.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Supposing features of encoder and 4 𝑋 𝑌 𝑊!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' ∈ ℝ"×!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='" 𝑊$ ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' "×" Dynamic Channel Mixing Static Channel Mixing Static Attention Static Attention Linear GELU Figure 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Illustration of progressive spectral channel attention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' ⊙, ⊗ denote element-wise- and original matrix multiplication.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' decoder at the ith depth level are Fi e and Fi d, respectively, the addition skip connection can be precisely described as Fi+1 d = Fi d + Fi e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (7) The major problem of it lies in the same weight on the shal- low features from the encoder and highly processed features from the decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' This could be problematic due to the im- balanced information density of different spectral bands and feature channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For example, it would be more helpful to preserve more low-level features for cleaner spectral bands to retain the details while we might want more high-level features for noisy bands to properly remove all the noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To address the issue, we propose an Attentive Skip Con- nection (ASC) that explicitly weights the features from dif- ferent sources using the attention weight computed from two convolution layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' By denoting the weight of ith depth level as Mi, the detailed formulation is F = LeakyReLU(Conv1x1([Fi d, Fi e])) Mi = σ(Conv3x3(F)), Fi+1 d = (1 − Mi) ⊙ Fi d + Mi ⊙ Fi e, (8) where σ denotes the sigmoid function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' A visualization of our attentive skip connection is shown in Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our ASC adaptively strengthens the features of more important spatial locations and spectral bands through element-wise gating, thus leading to better reconstruction quality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Analysis and in-depth comparison.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We provide an in- depth analysis of the difference between different types of skip connections.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Typically, the output of the skip connec- tion module would be processed by the subsequent convo- lution layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For additive skip connection, the computation can be formulated as, Fo = W × (Fe + Fd) = W × Fe + W × Fd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (9) where Fo is the output features from the subsequent convo- lution layer, W denotes the convolution kernel, × denotes convolution, and Fe and Fd are the features from the en- coder and the decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' By using distributive law, we could know that we actually treat the features from the encoder and decoder as features in the same manifold, which are processed by the same convolution kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For concat skip connection, the formulation would be, Fo = W × [Fe, Fd] = W1 × Fe + W2 × Fd.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (10) + 1- Conv Conv 𝑀!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' F"# F$ # + F%# 𝑋 𝑌 𝑊& ∈ ℝ\'×&\' 𝑊) ∈ ℝ&\'×\' Dynamic Channel Mixing Static Channel Mixing Static Attention Static Attention Linear GELU Backward Head Forward Head Split Combine Split Combine Attention Attention 𝐹 ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='×#×$×% H×W C 𝑆 𝑂 ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='×#×&×% H×W C 𝑆 𝑊 ∈ ℝ!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' 𝑊 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' MLP MLP ⃐𝑍 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' ⃑𝑍 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×(×% ' Tanh Sigmoid 𝑂 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' 𝑂 ∈ ℝ!" metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content="×#×&×% ' Figure 5." metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Illustration of the attentive skip connection block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' where W is a convolution that transforms features with 2C channels to C channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' According to the properties of con- volution, we could divide W into two parts, W1 and W2, which indicates that concat skip connection assigns differ- ent kernels for features from the encoder and decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' From the above analysis, we know that additive skip connection averages the features from the encoder and de- coder extracted by the same convolution kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' This might weaken the discriminative features due to the feature aver- age.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' On the other hand, the concat skip connection extracts feature using two different kernels and then performs fea- ture average.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Thus, it is possible for it to generate features that avoid the problem of feature weakening in additive skip connection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' However, its success requires two convolution kernels to simultaneously consider the feature extraction and reweighting, which is difficult to achieve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' This not only includes the training difficulties but also the limitation of convolution itself for pixel-wise reweighting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To address the aforementioned issues, our ASC separates the feature reweighting and extraction into two steps as, Fo = W × (M ⊙ Fe + (1 − M) ⊙ Fd � �� � feature reweighting ) (11) where M is the pixel-wise weight for reweighting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Such de- composition has two advantages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' First, it avoids the need to learn two very discriminative convolutions as the con- cat skip connection, which makes the network concentrate more on finding more useful features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Second, it explicitly reweights the features from encoder and decoder, which dis- tinguishes the more informative features to pass through the subsequent layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Experiments In this section, we provide the experiments for simulated and real-world HSI denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' An ablation study and dis- cussion of each component are also included.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Datasets We conducted experiments of simulated HSI denoising with ICVL [1] and real-worlding HSI denoising with Urban and HSIDwRD [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' ICVL contains 201 clean HSI in the resolution of 1392 × 1300 that divides the spectrum into 31 spectral bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Urban and HSIDwRD contain 1 HSI and 59 HSIs with 210 and 34 bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Following [34], we split the ICVL into three parts, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', 100 images for train- ing, 51 for validation, and 50 for testing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For training, we 5 Sigma Metric Noisy ITSReg [35] WLRTR [9] KBR [27] NGmeet [15] HSID [37] QRNN3D [34] T3SC [4] GRUNet [19] TRQ3D [24] MAN-S (ours) MAN-M (ours) MAN-L (ours) 30 PSNR 18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='59 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='48 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='62 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='48 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='99 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='72 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='22 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='36 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='84 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='25 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='83 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='07 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='17 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='110 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='961 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='988 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='984 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='989 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='987 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='988 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='986 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='989 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='990 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='991 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='991 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='991 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='807 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='088 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='056 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='088 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='050 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='067 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='062 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='079 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='052 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='046 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='043 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='042 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='041 50 PSNR 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='15 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='88 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='72 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='16 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='26 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='39 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='15 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='47 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='75 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='30 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='60 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='84 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='94 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='046 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='941 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='978 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='974 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='980 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='980 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='982 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='980 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='983 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='985 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='985 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='986 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='986 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='991 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='098 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='073 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='100 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='059 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='083 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='074 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='087 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='062 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='053 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='052 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='050 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='049 70 PSNR 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='23 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='71 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='52 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='71 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='66 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='77 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='30 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='05 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='02 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='86 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='05 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='32 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='40 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='025 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='923 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='967 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='961 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='974 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='972 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='974 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='974 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='977 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='980 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='980 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='981 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='981 SAM 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='105 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='112 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='095 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='113 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='067 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='096 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='094 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='096 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='080 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='061 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='060 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='058 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='057 Blind PSNR 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='34 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='62 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='66 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='68 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='23 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='95 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='37 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='52 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='03 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='47 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='08 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='35 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='44 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='114 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='953 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='983 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='979 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='985 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='984 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='985 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='983 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='987 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='988 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='988 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='989 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='989 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='859 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='087 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='064 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='080 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='053 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='072 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='068 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='085 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='057 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='054 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='046 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='045 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='044 (a) GT (b) WLRTR (c) KBR (d) NGmeet (e) HSID (f) QRNN3D (g) T3SC (h) GRUNet (i) TRQ3D (j) MAN-M Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Simulated Gaussian noise removal results under several noise levels on ICVL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Blind suggests each image is corrupted by Gaussian noise with random sigma (ranged from 30 to 70).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Visual results show the 20th band under noise level 50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' process the original HSIs into multiple overlapped smaller data cubes via even-stride cropping.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The spatial resolution of each cube is 64 × 64 and the spectral resolution remains unchanged.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Random rotation and scaling are also employed to further augment the dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For testing, the main region in the size of 512 × 512 × 31 is used for ICVL, and 15 images in HSIDwRD are randomly selected.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We use pre- trained models of ICVL for Urban and HSIDwRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Implementation Details Adam [18] optimizer is adopted to minimize the mean square error.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We follow the training strategy as [34], with slight modifications on the setup of the learning rate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The strategy is briefly described here and we refer interested readers to [34] for more details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In short, the network is first trained on Gaussian noise with a fixed level of 50 for 20 epochs and then finetuned with a random noise level from a given set for 40 epochs to produce the first-stage Gaussian denoising model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In the second stage, the complex denois- ing model is obtained via another 30 epochs of fine-tuning using the trained Gaussian denoising model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The learning rate is set to 1 × 10−3 at first, but decayed by 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='1 every 5 or 10 epochs, and finally reduced to 1 × 10−5 for each stage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The batch size is set to 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Results on Gaussian Noise For the experiments with Gaussian noise, we evaluate our model with different noise strengths, including 30, 50, 70, and random strengths ranging from 30 to 70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The simu- lated noisy input is generated by adding zero-mean additive white Gaussian noise with a given variance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We use a single model to tackle different noise levels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' For systematical evaluation, we compare our method with four traditional HSI denoising methods, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', ITSReg [35], KBR [27], WLRTR [9], and NGmeet [15], and five deep-learning ones, including HSID [37], QRNN3D [34], T3SC [4], GRUNet [19], and TRQ3D [24].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In the spirit of fairness, the hyperparameters in traditional methods are carefully tuned, and the deep-learning-based models are re- trained if needed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The quantitative and visual results are shown in Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The comparisons of runtime and model size are pro- vided in Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can be easily observed that our method achieves the best performance in all different settings with a large margin over the competing methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, our MAN-L achieves over 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='6 dB PSNR improvement on aver- age on four different settings while maintaining compara- ble parameters and runtime.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' our MAN-S achieves over 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='3 dB improvement against the best competing method, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', TRQ3D, with even fewer parameters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Moreover, our model is better at preserving spectral fidelity than previous meth- ods, which is reflected by the improvement in SAM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Results on Complex Noise Unlike the simulated noisy HSI with Gaussian noise, the real HSI is usually corrupted by different types of complex noise, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', strip noise, impulse noise, and deadline noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To evaluate the robustness of our method in real scenes, 6 Type Metric Noisy LRMR [38] LRTV [16] NMoG [10] TDTV [31] HSID [37] QRNN3D [34] T3SC [4] GRUNet [19] TRQ3D [24] MAN-S (ours) MAN-M (ours) G+Stripe PSNR 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='80 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='62 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='49 33.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='87 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='67 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='77 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='35 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='85 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='39 43.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='05 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='24 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='60 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='159 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='717 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='905 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='799 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='940 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='942 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='976 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='986 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='991 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='992 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='993 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='994 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='910 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='187 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='078 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='265 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='081 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='104 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='055 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='072 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='050 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='043 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='039 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='038 G+Deadline PSNR 17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='61 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='83 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='37 32.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='87 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='15 37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='65 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='23 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='54 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='11 42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='95 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='13 44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='51 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='155 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='709 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='895 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='797 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='930 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='940 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='976 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='983 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='991 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='992 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='993 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='994 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='917 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='227 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='115 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='276 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='099 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='102 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='056 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='096 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='050 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='044 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='040 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='038 G+Impulse PSNR 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='80 29.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='70 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='56 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='60 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='67 35.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='00 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='23 36.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='06 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='70 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='27 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='88 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='97 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='114 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='623 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='871 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='652 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='935 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='899 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='945 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='952 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='985 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='983 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='982 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='979 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='926 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='311 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='242 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='486 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='094 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='174 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='109 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='203 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='067 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='075 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='090 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='095 G+Mixture PSNR 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='08 28.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='68 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='47 27.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='31 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='77 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='05 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='25 34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='48 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='51 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='27 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='03 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='18 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='099 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='608 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='858 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='632 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='919 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='888 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='938 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='946 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='981 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='983 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='981 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='979 SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='944 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='353 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='287 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='513 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='113 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='181 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='107 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='228 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='081 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='075 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='092 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='097 (a) GT (b) LRTV (c) NMoG (d) TDTV (e) HSID (f) QRNN3D (g) T3SC (h) GRUNet (i) TRQ3D (j) MAN-M Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Simulated complex noise removal results under several noise types on ICVL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Visual results show the 20th band under stripe noise.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Metrics NMoG [10] TDTV [31] QRNN3D [34] HSID [37] HSIDwRD [42] MAN-M (ours) PSNR 30.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='90 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='14 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='13 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='05 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='23 31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='36 SSIM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='907 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='881 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='940 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='934 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='939 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='940 SAM 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='761 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='853 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='094 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='096 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='092 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='092 Table 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Quantitative results on HSIDwRD [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' we also conduct experiments with different combinations of complex noise, following the settings in [34].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We compare our method with eight recently developed methods including five deep-learning-based methods, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', HSID [37], QRNN3D [34], GRUNet [19], T3SC [4], and TRQ3D [24], as well as four optimization-based ones, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', LRMR [38], LRTV [16], NMoG [10] and TDTV [31].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We choose a different set of optimization-based methods from the ones used for Gaussian noise because these methods only perform well for the noise settings they can solve.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Table 2 gives the quantitative and visual comparison be- tween our method against the competing ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can be seen that our method obtain substantial improvement on various metrics, especially on PSNR, which is over 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='5 dB on average, and ranged from 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='03 ∼ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='7 dB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Results on Real-world Noise Urban.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Figure 6 provide the visual result of our model on real-world noisy remotely sensed HSI, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', Urban.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can be seen that our model could properly remove the heavy noise while retaining the details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' QRNN3D removes the most noise but loses the details such as stripes on the roof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' MHRSA ASC PSCA Params Runtime PSNR SAM 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='43 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='25 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='98 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='072 ✓ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='46 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='13 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='058 ✓ ✓ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='81 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='55 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='053 ✓ ✓ ✓ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='89 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='65 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='84 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='050 Table 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Ablation study of the proposed network modules.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' HSID tends to produce blurry results than Ours.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' HSIDwRD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We conduct experiments on the real natural HSI dataset [42] with the first two leading optimization- based and CNN-based methods from complex denoising and the method proposed in [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The quantitative results are shown in Table 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can be observed that our results outperform the other ones in terms of all metrics, which demonstrates the effectiveness of our methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Discussions Ablation Study.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' To verify the effectiveness of each pro- posed component.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We evaluate the PSNR improvement each component brings by adding them one by one, start- ing from a baseline model, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', UNet with 3D convolution and activation only.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' All the models are trained for the Gaus- sian denoising task and evaluated on 50 noise strength.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The experimental results are shown in Table 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' As we can ob- serve, the introduction of MHRSA brings the most signif- icant improvement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' This indicates the importance of ex- ploring inter-spectral correlations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' The attentive skip con- nection contributes 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='35 dB improvement by taking into 7 (a) Noisy (b) NMoG [10] (c) TDTV [31] (d) HSID [37] (e) QRNN3D [34] (f) GRUNet [19] (g) MAN(Ours) Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Real world noise removal results at 207th band of Urban Dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Method ITSReg [35] WLRTR [9] KBR [27] NGmeet [15] HSID [37] QRNN3D [34] T3SC [4] GRUNet [19] TRQ3D [24] MAN-S (ours) MAN-M (ours) MAN-L (ours) PSNR 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='88 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='16 38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='70 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='26 39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='39 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='15 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='47 40.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='75 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='30 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='60 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='84 41.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='94 Params (M) 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='40 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='86 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='83 14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='68 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='50 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='89 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='39 Runtime (s) 907 1600 1755 166 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='48 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='44 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='95 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='87 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='47 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='49 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='65 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='97 Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Performance-Params-Runtime comparisons with the SOTA methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Evaluated with Gaussian denoising (sigma=50) on ICVL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Ground Truth Band 7 Band 12 Band 21 Noisy Input Channel 4 Channel 18 Channel 27 Channel 38 Band 31 Figure 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Visualization of the attention map of our attentive skip connection across different bands and channels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='20 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='15 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='05 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='00 0 10 20 30 40 Pixel j Band i 30 25 20 15 10 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='04 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='03 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='02 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='01 5 0 20 15 10 Band y 30 25 20 15 10 Band x Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Visualization of the spectral attention map from our MHRSA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (a) The left part is the average attention map between different bands.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' (b) The right part is the attention distribution of 15th band across different pixels.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' account low- and high-level features fusion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our PSCA strengthens the features from MHRSA by considering intra- spectral relationships, which further provides 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='36 dB im- provement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Visualization of Attentive Skip Connection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We visual- ize the attention map of the first attentive skip connection block for one sample.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' As is shown in Figure 7, the attention map varies across different bands and channels, which indi- cates that the equal attention of vanilla additive one might be less effective.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In particular, it can be seen that the net- work learns to pay attention to different regions across dif- ferent bands and channels, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', band12 and channel 21 for edges, and different sources, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', channel 18 for encoder, channel 38 for decoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Visualization of Spectral Attention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Figure 8 shows the spectral attention map of one particular example HSI at the first MHRSA layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' It can be seen that our MHRSA pays at- tention to all bands with different weights, which is helpful for adaptively aggregating the useful features from different bands to assist the denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In general, the overall pattern is that each band pays more attention to the bands around it, but the attention distributions may vary with respect to pixels at different locations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Conclusion In this paper, we propose a mixed attention network for hyperspectral image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our method introduces several key components to properly explore the inter- and intra-spectral correlations as well as the low- and high-level spatial-spectral feature interactions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Specifically, these are achieved with a multi-head recurrent spectral attention that recurrently merges the features across different bands, a progressive channel attention that progressively mixes the different features within each band, and an attentive skip connection that aggregates the features from encoder and decoder with different importance weight.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' We perform ex- tensive experiments on simulated and real-world noise, and it shows that our method outperforms the existing state-of- the-art methods with significant improvement while main- taining a smaller model size and running time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Ethical considerations and future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Our work has no ethical issues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In this work, we explore the proposed MHRSA with multi-head in two directions, but it is also possible to extend it to multi-axis in which we perform at- tention in different dimensions, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=', channel and spectral.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 8 References [1] Boaz Arad and Ohad Ben-Shahar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Sparse recovery of hyper- spectral signal from natural rgb images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the European Conference on Computer Vision, pages 19–34.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Springer, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 5 [2] Saeideh Ghanbari Azar, Saeed Meshgini, Tohid Yousefi Rezaii, and Soosan Beheshti.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image classi- fication based on sparse modeling of spectral blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neuro- computing, 407:12–23, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [3] George Alan Blackburn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral remote sensing of plant pigments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Journal of Experimental Botany, 58(4):855– 867, 2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [4] Th´eo Bodrito, Alexandre Zouaoui, Jocelyn Chanussot, and Julien Mairal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' A trainable spectral-spatial sparse coding model for hyperspectral image restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 34:5430–5442, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3, 6, 7, 8 [5] Antoni Buades, Bartomeu Coll, and Jean-Michel Morel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Non-local means denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Image Processing On Line, 1:208–212, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3 [6] Xianghai Cao, Yiming Ge, Renjie Li, Jing Zhao, and Licheng Jiao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral imagery classification with deep metric learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 356:217–227, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [7] Stanley H Chan, Xiran Wang, and Omar A Elgendy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Plug- and-play admm for image restoration: Fixed-point conver- gence and applications.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Computa- tional Imaging, 3(1):84–98, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [8] Yi Chang, Luxin Yan, Houzhang Fang, Sheng Zhong, and Wenshan Liao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hsi-denet: Hyperspectral image restoration via convolutional neural network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geoscience and Remote Sensing, 57(2):667–682, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3 [9] Yi Chang, Luxin Yan, Xi-Le Zhao, Houzhang Fang, Zhijun Zhang, and Sheng Zhong.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Weighted low-rank tensor recov- ery for hyperspectral image restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Cybernetics, 50(11):4558–4572, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2, 6, 8 [10] Yang Chen, Xiangyong Cao, Qian Zhao, Deyu Meng, and Zongben Xu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Denoising hyperspectral image with non- iid noise structure.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Cybernetics, 48(3):1054–1066, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2, 7, 8 [11] Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Image denoising by sparse 3-d transform- domain collaborative filtering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Image Processing, 16(8):2080–2095, 2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [12] Aram Danielyan, Vladimir Katkovnik, and Karen Egiazar- ian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Image deblurring by augmented lagrangian with bm3d frame prior.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the Workshop on Information Theoretic Methods in Science and Engineering, volume 1, 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [13] Weisheng Dong, Huan Wang, Fangfang Wu, Guangming Shi, and Xin Li.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Deep spatial–spectral representation learn- ing for hyperspectral image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Computational Imaging, 5(4):635–648, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2, 3, 4 [14] Ying Fu, Antony Lam, Imari Sato, and Yoichi Sato.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Adap- tive spatial-spectral dictionary learning for hyperspectral im- age restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' International Journal of Computer Vision, 122(2):228–245, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [15] Wei He, Quanming Yao, Chao Li, Naoto Yokoya, and Qibin Zhao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Non-local meets global: An integrated paradigm for hyperspectral denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition, pages 6868–6877, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2, 6, 8 [16] Wei He, Hongyan Zhang, Liangpei Zhang, and Huanfeng Shen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Total-variation-regularized low-rank matrix factoriza- tion for hyperspectral image restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geoscience and Remote Sensing, 54(1):178–188, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2, 7 [17] Jie Hu, Li Shen, and Gang Sun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Squeeze-and-excitation net- works.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7132–7141, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 4 [18] Diederik P Kingma and Jimmy Ba.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Adam: A method for stochastic optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' arXiv preprint arXiv:1412.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content='6980, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 6 [19] Zeqiang Lai, Kaixuan Wei, and Ying Fu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Deep plug-and-play prior for hyperspectral image restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 481:281–293, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3, 6, 7, 8 [20] YunYang Liu, XiLe Zhao, YuBang Zheng, TianHui Ma, and Hongyan Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image restoration by tensor fibered rank constrained optimization and plug-and-play reg- ularization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geoscience and Remote Sensing, pages 1–17, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [21] TianHui Ma, Zongben Xu, Deyu Meng, and XiLe Zhao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image restoration combining intrinsic image characterization with robust noise modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Journal of Selected Topics in Applied Earth Observations and Re- mote Sensing, 14:1628–1644, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [22] Matteo Maggioni, Vladimir Katkovnik, Karen Egiazarian, and Alessandro Foi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Nonlocal transform-domain filter for volumetric data denoising and reconstruction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Trans- actions on Image Processing, 22(1):119–133, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [23] Hisham Othman and Shen-En Qian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Noise reduction of hy- perspectral imagery using hybrid spatial-spectral derivative- domain wavelet shrinkage.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geo- science and Remote Sensing, 44(2):397–408, 2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [24] Li Pang, Weizhen Gu, and Xiangyong Cao.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Trq3dnet: A 3d quasi-recurrent and transformer based network for hyper- spectral image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 14(18):4598, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3, 6, 7, 8 [25] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Peng, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Xie, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Zhao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Yee, and D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Meng.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Enhanced 3dtv regularization and its applications on hsi de- noising and compressed sensing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Im- age Processing, 29:7889–7903, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [26] Yi Peng, Deyu Meng, Zongben Xu, Chenqiang Gao, Yi Yang, and Biao Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Decomposable nonlocal tensor dic- tionary learning for multispectral image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Pro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2949–2956, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [27] Xie Qi, Zhao Qian, Deyu Meng, and Zongben Xu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Kronecker-basis-representation based tensor sparsity and its applications to tensor recovery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Pat- tern Analysis and Machine Intelligence, PP(99):1–1, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 6, 8 [28] Le Sun, Byeungwoo Jeon, Yuhui Zheng, and Zebin Wu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hy- perspectral image restoration using low-rank representation on spectral difference image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Geoscience and Remote Sensing Letters, 14(7):1151–1155, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [29] Prasad S Thenkabail and John G Lyon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral remote sensing of vegetation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' CRC Press, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 9 [30] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Attention is all you need.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Advances in neural information processing systems, 30, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3 [31] Yao Wang, Jiangjun Peng, Qian Zhao, Yee Leung, Xi-Le Zhao, and Deyu Meng.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image restoration via total variation regularized low-rank tensor decomposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Journal of Selected Topics in Applied Earth Observa- tions and Remote Sensing, 11(4):1227–1243, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2, 7, 8 [32] Zhendong Wang, Xiaodong Cun, Jianmin Bao, Wengang Zhou, Jianzhuang Liu, and Houqiang Li.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Uformer: A general u-shaped transformer for image restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 17683–17693, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3 [33] Kaixuan Wei and Ying Fu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Low-rank bayesian tensor factor- ization for hyperspectral image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 331:412–423, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [34] Kaixuan Wei, Ying Fu, and Hua Huang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3-d quasi-recurrent neural network for hyperspectral image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Neural Networks and Learning Systems, 32(1):363–375, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2, 3, 5, 6, 7, 8 [35] Qi Xie, Qian Zhao, Deyu Meng, Zongben Xu, Shuhang Gu, Wangmeng Zuo, and Lei Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Multispectral images de- noising by intrinsic tensor sparsity regularization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Pro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1692–1700, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 6, 8 [36] Qiangqiang Yuan, Liangpei Zhang, and Huanfeng Shen.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hy- perspectral image denoising employing a spectral–spatial adaptive total variation model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geo- science and Remote Sensing, 50(10):3660–3677, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [37] Qiangqiang Yuan, Qiang Zhang, Jie Li, Huanfeng Shen, and Liangpei Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image denoising employ- ing a spatial–spectral deep residual convolutional neural net- work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geoscience and Remote Sens- ing, 57(2):1205–1218, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2, 3, 6, 7, 8 [38] Hongyan Zhang, Wei He, Liangpei Zhang, Huanfeng Shen, and Qiangqiang Yuan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image restoration using low-rank matrix recovery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Geoscience and Remote Sensing, 52(8):4729–4743, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 7 [39] Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Image Processing, 26(7):3142–3155, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 3 [40] Kai Zhang, Wangmeng Zuo, and Lei Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Ffdnet: Toward a fast and flexible solution for cnn-based image denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Image Processing, 27(9):4608–4622, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 [41] Liangpei Zhang and Xin Huang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Object-oriented subspace analysis for airborne hyperspectral remote sensing imagery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 73(4-6):927–936, 2010.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [42] Tao Zhang, Ying Fu, and Cheng Li.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image de- noising with realistic data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2248– 2257, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 5, 7 [43] Haitao Zhao and Shaoyuan Sun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Sparse tensor embed- ding based multispectral face recognition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 133:427–436, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [44] XiLe Zhao, Hao Zhang, TaiXiang Jiang, Michael K Ng, and Xiong-Jun Zhang.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Fast algorithm with theoretical guarantees for constrained low-tubal-rank tensor recovery in hyperspec- tral images denoising.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 413:397–409, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1, 2 [45] Zhihong Pan, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Healey, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Prasad, and B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Tromberg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Face recognition in hyperspectral images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(12):1552– 1560, 2003.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [46] Feng Zhou, Renlong Hang, Qingshan Liu, and Xiaotong Yuan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Hyperspectral image classification using spectral- spatial lstms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' Neurocomputing, 328:39–47, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 1 [47] Daniel Zoran and Yair Weiss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' From learning models of natu- ral image patches to whole image restoration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' In Proceedings of the IEEE International Conference on Computer Vision, pages 479–486.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' IEEE, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
+page_content=' 2 10' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/kNFJT4oBgHgl3EQfYSw1/content/2301.11525v1.pdf'}
diff --git a/kdE2T4oBgHgl3EQfIgZc/vector_store/index.pkl b/kdE2T4oBgHgl3EQfIgZc/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..d4108051f170f368f3fbdd7951c1b8835efc5ee5
--- /dev/null
+++ b/kdE2T4oBgHgl3EQfIgZc/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c685dc4cbce51e07d4b5f385b4283014620895a3b0ff0e24af24fbe392ff0f7
+size 233909
diff --git a/kdFLT4oBgHgl3EQfcy-f/content/2301.12084v1.pdf b/kdFLT4oBgHgl3EQfcy-f/content/2301.12084v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..9d3ba420d52b2f11d1a6f0ef3a1726280a0badde
--- /dev/null
+++ b/kdFLT4oBgHgl3EQfcy-f/content/2301.12084v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7eb4adb414865b3f0514b4831ad846b760354bb753731720ff8d3b8e599cff0
+size 613181
diff --git a/kdFLT4oBgHgl3EQfcy-f/vector_store/index.faiss b/kdFLT4oBgHgl3EQfcy-f/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..4dc7aeed20d9f92615b9978722e348a7f0d7416f
--- /dev/null
+++ b/kdFLT4oBgHgl3EQfcy-f/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc5ec8a9b190208b36fa952986446ab67a0105270aab859be07268ba08e005fc
+size 1310765
diff --git a/kdFLT4oBgHgl3EQfcy-f/vector_store/index.pkl b/kdFLT4oBgHgl3EQfcy-f/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..0c4ece05b94eff30cd84f059283748f53768d5f3
--- /dev/null
+++ b/kdFLT4oBgHgl3EQfcy-f/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f58cc47bcaaa3d26bf256a2b802ba9d5c55f756f08bac39e50a65a526281b0e
+size 52001
diff --git a/ktE0T4oBgHgl3EQfYwAL/vector_store/index.pkl b/ktE0T4oBgHgl3EQfYwAL/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..de57253acb1a451afa3e9d63e48c6b12d145ca8b
--- /dev/null
+++ b/ktE0T4oBgHgl3EQfYwAL/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e31336c8157190d92844859cbb84c74165028415e799d026ee794570265d514
+size 92016
diff --git a/mdE2T4oBgHgl3EQfeQfB/content/2301.03915v1.pdf b/mdE2T4oBgHgl3EQfeQfB/content/2301.03915v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..d347a4bd161d846786c61ea5f9a40df6c8c02329
--- /dev/null
+++ b/mdE2T4oBgHgl3EQfeQfB/content/2301.03915v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a24cf7626b4bf0daf5b3be93abac736e4e2e76b36bc62b472b63c62d2ad4fa50
+size 1092392
diff --git a/mdE2T4oBgHgl3EQfeQfB/vector_store/index.pkl b/mdE2T4oBgHgl3EQfeQfB/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..f7040124228ee94f1929c34125f9d06bb0068bce
--- /dev/null
+++ b/mdE2T4oBgHgl3EQfeQfB/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ccd797f6c369df8757fd722c98cd38fedb25fc90a9329dc209d71a0468d87b3
+size 158400
diff --git a/mdFAT4oBgHgl3EQfcR2T/vector_store/index.faiss b/mdFAT4oBgHgl3EQfcR2T/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..4880af85c6d858c04639bcb734595c3be371e2ba
--- /dev/null
+++ b/mdFAT4oBgHgl3EQfcR2T/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13aeb6e485f746498262403d924e7799b49517635d1fdd52acf040274456c037
+size 6357037
diff --git a/nNA0T4oBgHgl3EQfJv9r/content/tmp_files/2301.02094v1.pdf.txt b/nNA0T4oBgHgl3EQfJv9r/content/tmp_files/2301.02094v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d25772734a963bb040d9ba675a55481d9711a138
--- /dev/null
+++ b/nNA0T4oBgHgl3EQfJv9r/content/tmp_files/2301.02094v1.pdf.txt
@@ -0,0 +1,1810 @@
+Strong decays of P N
+ψ (4312)+ → J/ψ(ηc)p within the
+Bethe-Salpeter framework
+Qiang Li1 Chao-Hsi Chang2,3,4 Tianhong Wang5 Guo-Li Wang5,6
+1School of Physical Science and Technology, Northwestern Polytechnical University, Xi’an 710072,
+China
+2CAS Key Laboratory of Theoretical Physics, Institute of Theoretical Physics, Chinese Academy
+of Sciences, Beijing 100190, China
+3School of Physical Sciences, University of Chinese Academy of Sciences, Beijing 100049, China
+4School of Physical Science and Technology, Lanzhou University, Lanzhou 730000, China
+5School of Physics, Harbin Institute of Technology, Harbin 150001, China
+6Department of Physics, Hebei University, Baoding 071002, China
+E-mail: liruo@nwpu.edu.cn, zhangzx@itp.ac.cn, thwang@hit.edu.cn,
+gl wang@hit.edu.cn
+Abstract: Based on the effective Lagrangian in the heavy quark limit, we calculate the
+one-boson-exchange interaction kernel of ¯DΣc in the isospin-1
+2 state. We present the Bethe-
+Salpeter equation and wave function for the constituent particles to be a (pseudo)scalar
+meson and a 1
+2 baryon. By using the Bethe-Salpeter equation, we can obtain P N
+ψ (4312)+
+as the ¯DΣc molecular state with JP = ( 1
+2)−. Combining the effective Lagrangian and the
+obtained BS wave function, the partial decay widths of P N
+ψ (4312)+ to J/ψp and ηcp are
+calculated to be 0.11−0.04
++0.06 MeV and 5.6−1.9
++2.6 × 10−2 MeV, respectively, which are consistent
+with the LHCb experimental measurements.
+Our results indicate the fraction of J/ψp
+channel amounts to ∼ 1% of P N
+ψ (4312)+, and can reach to ∼ 30% when the interaction
+kernel is reduced by half. Our results favor the interpretation of P N
+ψ (4312)+ as the ¯DΣc
+molecular state with JP = ( 1
+2)− and isospin I = 1
+2.
+arXiv:2301.02094v1 [hep-ph] 5 Jan 2023
+
+Contents
+1
+Introduction
+1
+2
+P N
+ψ (4312)+ as the ¯DΣc molecular state
+2
+2.1
+Bethe-Salpeter equation of a JP = 0− meson and a 1
+2
++ baryon
+3
+2.2
+Interaction kernel from the one-boson exchange
+5
+2.3
+Salpeter wave function for the JP = 1
+2
+− pentaquark states
+7
+3
+Strong decay of P N
+ψ (4312)+ → J/ψ(ηc)p within the BS wave function
+7
+3.1
+Amplitude for P N
+ψ (4312)+ → J/ψp
+9
+3.1.1
+Amplitude with D exchange
+9
+3.1.2
+Amplitude with D∗ exchange
+11
+3.2
+Amplitude for P N
+ψ (4312)+ → ηcp
+12
+3.3
+Partial decay width
+13
+4
+Numerical results and discussions
+14
+A Expressions of the decay form factors
+17
+1
+Introduction
+In 2019, a narrow pentaquark state Pc(4312)+ is first observed in the J/ψp invariant mass
+spectrum [1] by the LHCb collaboration, which indicates this state at least to contain five
+valence quarks, namely, [c¯cuud] quark contents. This pentaquark state will be labeled as
+P N
+ψ (4312)+ in this work following the new naming scheme proposed by the LHCb collabora-
+tion [2], where the superscript N denotes the isotopic spin I = 1
+2 and the subscript ψ denotes
+the hidden charm flavor. The measured mass and total width are MP N
+ψ (4312)+ = 4312 MeV
+and ΓP N
+ψ (4312)+ = 9.8 ± 2.7+3.7
+−4.5 MeV [1] respectively. The proximity to the ¯DΣc threshold
+of the observed narrow peak suggests that they play an important role in the dynamics of
+P N
+ψ (4312)+ state, and makes the ¯DΣc molecular state picture a natural interpretation to
+this exotic particle.
+The hidden charm molecular pentaquark states have been proposed before the exper-
+imental confirmation [3–9]. After the LHCb discoveries, lots of literature explored these
+newly observed pentaquark states from different aspects within different approaches, such
+as Refs. [10–27]. Although the properties of the P N
+ψ (4312)+ are most likely to be the S-
+wave combination of ¯DΣc with I(JP ) = 1/2(1/2−) [10–16, 26, 27], the contrary view [28],
+or the possibilities of the compact pentaquark state [29, 30] or kinematical effects [31, 32]
+still exist. Though suggested by the LHCb to be labeled as P N
+ψ (4312)+, the essence of this
+pentaquark state is still an open question.
+– 1 –
+
+Besides the spectrum or electromagnetic properties [33–35], the strong decay properties
+play important roles in determining the nature of the pentaquark states. The decay to
+J/ψp is the discovery channel and also the only detected decay mode of P N
+ψ (4312)+ so far,
+and hence this decay channel should be payed more attention to explore the property of
+P N
+ψ (4312)+. Several approaches are used to study the decay properties of these pentaquark
+states [10, 15, 30, 36–41], including the effective Lagrangian methods [10, 15], the flavor-spin
+and heavy quark spin symmetry [30, 36], the chiral constituent quark model [37], QCD sum
+rules [40, 41], etc. Most of the previous studies are based on the nonrelativistic Schrodinger
+or Lippmann-Schwenger equation and the results are dependent on several introduced free
+parameters, especially the cutoff value in the form factors. These undetermined parameters
+weaken the prediction power of the theories and bring ambiguity in interpreting the nature
+of P N
+ψ (4312)+. Some researches also suggest the ηcp channel can be an important decay
+mode of P N
+ψ (4312)+ [37, 38], especially, the methods by using the heavy quark symmetry
+predict that the decay ratio of P N
+ψ (4312)+ to ηcp over J/ψp can reach about three [20, 36].
+However, no experimental evidence is reported in a recent search for pentaquark state
+P N
+ψ (4312)+ in Λ0
+b → ηcpK− decay channel [42]. More studies on the decay behaviors of
+P N
+ψ (4312)+ can be important and helpful to explore its inner structure and dynamics.
+In this work, we will calculate the partial decay widths of P N
+ψ (4312)+ to J/ψp and ηcp
+by combing the Bethe-Salpeter (BS) framework with the effective Lagrangian. The Bethe-
+Salpeter equation(BSE) is a relativistic two-body bound state equation. Another advantage
+is that the constructed BS wave functions only depend on the good quantum number spin-
+parity and Lorentz covariance. The BS methods have already been successfully used to cope
+with mass spectra of the doubly heavy baryons [43, 44], producing the recently observed
+molecular pentaquarks [26] and the fully heavy tetraquark TQQ ¯Q ¯Q states [45], and also the
+hadronic transitions and decays[46–52].
+The theoretical calculations from BS methods
+achieve satisfactory consistences with the experimental measurements.
+This paper is organized as follows. After the introduction, we start with the Bethe-
+Salpeter equation for P N
+ψ (4312)+ as the molecular state of a (pseudo)scalar meson and a
+baryon, including the interaction kernel and the relevant Salpeter wave function (Sect. 2),
+then we calculate the strong decay widths of P N
+ψ (4312)+ → J/ψ(ηc)p (Sect. 3). We finally
+present the numerical results, discussion and summaries in Sect. 4.
+2
+P N
+ψ (4312)+ as the ¯DΣc molecular state
+In this part, we will first briefly review Bethe-Salpeter equation of a scalar meson and
+a baryon under the instantaneous approximation.
+Then we introduce the pentaquark
+interaction kernel based on the one-boson exchange. The relativistic BS wave functions of
+the JP = ( 1
+2)− P N
+ψ state will be introduced and solved numerically to prepare for the next
+decay calculations.
+– 2 –
+
+P N
+ψ , P
+P N
+ψ , P
+¯D, p1
+Σc, p2
+¯D, k1
+Σc, k2
+p1
+p2
+i
+i′′
+i′
+i
+=
+K
+Fig. 1:
+Bethe-Salpeter equation of the molecular state consisting of the constituent
+(pseudo)scalar meson and J = 1
+2 baryon. The orange letters denote the Dirac indices.
+The blue symbols P, p1(k1), p2(k2) denote the momenta of the pentaquark, constituent
+meson, and the constituent baryon respectively.
+2.1
+Bethe-Salpeter equation of a JP = 0− meson and a 1
+2
++ baryon
+Fig. 1 schematically depicts the Bethe-Salpeter equation for the bound state consisting of
+a constituent meson and a constituent baryon, which can be expressed as
+Γ(P, q, r) =
+�
+d4k
+(2π)4 (−i)K(k, q)[S(k2)Γ(P, k, r)D(k1)],
+(2.1)
+where Γ(P, q, r) denotes the vertex of the pentaquark, constituent meson and baryon; we
+used P, q, and r to represent the pentaquark total momentum, inner relative momentum,
+and spin state respectively; the inner relative momentum q and k are defined as q ≡
+α2p1 − α1p2, k ≡ α2k1 − α1k2, with α1(2) ≡
+m1(2)
+m1+m2 , k1(2) denoting the momentum of the
+constituent meson (baryon), and m1(2) is the corresponding mass; S(k2) =
+i
+/k2−m2 is the
+free Dirac propagator of the baryon; D(k1) =
+i
+k2
+1−m2
+1 denotes the usual scalar propagator.
+The iϵ should be implied in all the propagators. Since both the two constituent particles,
+namely ¯D and Σc, contain a heavy charm quark, the relative velocity would be small. Then
+the interaction kernel K(k, q) is assumed to be instantaneous and is not dependent on the
+time component of the exchanged momentum (k − q), namely, K(k, q) ∼ K(k⊥ − q⊥),
+where k⊥ = k − kP ˆP with kP ≡ k · ˆP and ˆP = P
+M , and M is the pentaquark mass. The
+spacelike momentum q⊥ is defined similarly. Throughout this work, this instantaneous
+approximation is assumed for the pentaquark kernel.
+The four-dimensional Bethe-Salpeter wave function is defined as
+ψ(q) = S(p2)Γ(q)D(p1),
+(2.2)
+where the dependence on P and r omitted for simplicity.
+Since the interaction kernel
+K(k⊥ − q⊥) is instantaneous, the integral over the time component of q can be absorbed
+into the wave function and it is useful to define the three-dimensional Salpeter wave function
+as
+ϕ(q⊥) ≡ −i
+� dqP
+2π ψ(q).
+(2.3)
+– 3 –
+
+where the factor (−i) is just a convention for later convenience.
+Performing the contour integral over qP on both sides of Eq. (2.2), we can obtain the
+Salpeter equation (SE) for meson-baryon bound state [26],
+ϕ(q⊥) =
+1
+2w1
+�
+Λ+(p2⊥)
+M − w1 − w2
++
+Λ−(p2⊥)
+M + w1 + w2
+�
+Γ(q⊥),
+(2.4)
+where wi =
+�
+m2
+i − p2
+i⊥
+�1/2 (i = 1, 2) denotes the kinetic energy of the constituent meson
+and baryon respectively. The projector operators Λ±(p2⊥) are defined as
+Λ±(p2⊥) = 1
+2 [1 ± H2(p2⊥)] γ0,
+(2.5)
+H2(p2⊥) = 1
+w2
+(/p2⊥ + m2) γ0.
+(2.6)
+Notice that H2(p2⊥) is just the corresponding Dirac Hamiltonian divided by the kinetic
+energy w2. Using the projector operator, we can further define the positive and negative
+energy wave functions as ϕ± ≡ Λ±γ0ϕ, and we also have ϕ = ϕ+ + ϕ−. The SE above can
+be further rewritten as the following type
+Mϕ = (w1 + w2)H2(p2⊥)ϕ +
+1
+2w1
+γ0Γ(q⊥).
+(2.7)
+where the vertex Γ(q⊥) is now expressed as the integral of the Salpeter wave function,
+Γ(q⊥) =
+�
+d3k⊥
+(2π)3 K(k⊥ − q⊥)ϕ(k⊥).
+(2.8)
+The Salpeter Eq. (2.7) is in fact an eigenvalue equation about the Salpeter wave function
+ϕα(q⊥), where the pentaquark mass M behaves as the eigenvalue. The three-dimensional
+BSE, namely, Eq. (2.7), indicates that the mass of the pentaquark state consists of two
+parts, the kinetic energy and the potential energy.
+The normalization condition of the BS wave function is generally expressed as,
+− i
+� �
+d4q
+(2π)4
+d4k
+(2π)4 ¯ψ(P, q, ¯r) ∂
+∂P 0 I(P, k, q)ψ(P, k, r) = 2Mδr¯r,
+where ¯ψ = ψ†γ0 and δr¯r is the Kronecker symbol; the integral kernel I in the normalization
+condition reads,
+I(P, k, q) = (2π)2δ4(k − q)S−1(p2)D−1(p1) + iK(P, k, q).
+Notice in this work, under the instantaneous approximation, the interaction kernel has no
+dependence on P 0, which indicates the normalization would only involve the inverses of
+the two propagators. Performing the contour integral, the normalization condition can be
+– 4 –
+
+further expressed by the Salpeter wave function as
+�
+d3q⊥
+(2π)3 (2w1) ¯ϕ(q⊥, ¯r)γ0ϕ(q⊥, r) = 2Mδr¯r,
+(2.9)
+where ¯ϕ = ϕ†γ0 and the symbol ¯r just denotes the spin state; also notice both the BS wave
+function ψ(q) and the Salpeter wave function ϕ(q⊥) are four-component spinor.
+2.2
+Interaction kernel from the one-boson exchange
+The P N
+ψ (4312)+ are consistent with the ¯DΣc molecular state with isospin I =
+1
+2 and
+I3 = + 1
+2, which can be expressed in the uncoupled representation as
+| 1
+2, 1
+2⟩ =
+√
+2
+√
+3 |1, +1⟩ | 1
+2, − 1
+2⟩ −
+1
+√
+3 |1, 0⟩ | 1
+2, 1
+2⟩ =
+√
+2
+√
+3 |Σ++
+c
+⟩ |D−⟩ −
+1
+√
+3 |Σ+
+c ⟩ | ¯D0⟩ .
+(2.10)
+In the molecular state scenario of P N
+ψ (4312)+, the interaction kernel between the two
+constituents Σc and ¯D can be realized by the one-boson exchange. Notice the usual one-
+pion exchange is not possible in the ¯DΣc bound state for the parity. We only need to
+consider the light scalar and vector meson exchange.
+Considering the heavy quark spin-flavor symmetry, hidden local symmetry and the
+light quark chiral symmetry, the involved Lagrangian describing the charmed anti-heavy-
+light meson and a light scalar and vector meson reads [6, 53]
+LHHV = −ρV1 ⟨ ¯H¯cvαVαH¯c⟩ − ρT1 ⟨ ¯H¯cσαβ(∂αVβ − ∂βVα)H¯c⟩ + σ1 ⟨ ¯H¯cσH¯c⟩ .
+(2.11)
+Here ⟨ ⟩ denotes taking the Dirac trace, and σ denotes field of the light scalar meson. ρV 1,
+ρT1, and σ1 denote the corresponding coupling constants. H¯c represents the field of the
+( ¯D, ¯D∗) doublet in the heavy quark limit,
+H¯c =
+� ¯D∗µγµ + i ¯Dγ5
+� 1 − /v
+2
+,
+(2.12)
+where ¯D = ( ¯D0, D−, D−
+s ) denotes anti-charmed heavy-light meson fields in flavor triplet,
+and ¯D∗µ is the corresponding vector state; ¯H¯c = γ0H†
+¯cγ0 is the usual conjunction in Dirac
+space; and v denotes four-velocity of the heavy-light meson. The symbol V denotes the
+3 × 3 matrix consisting of the 9 light vector meson fields [6, 53]
+V =
+�
+���
+(ρ0+ω)
+√
+2
+ρ+
+K∗+
+ρ−
+− (ρ0−ω)
+√
+2
+K∗0
+K∗−
+¯K∗0
+φ
+�
+��� .
+(2.13)
+Considering the heavy quark symmetry, hidden local symmetry and chiral symmetry,
+the effective Lagrangian of the heavy-light baryon and light mesons reads [6, 54–56]
+LB6B6V = ρV2 ⟨¯SµvαVαSµ⟩ + iρT2 ⟨¯Sµ(∂µVν − ∂νVµ)Sν⟩ + σ2 ⟨¯SµσSµ⟩ .
+(2.14)
+– 5 –
+
+Here ⟨ ⟩ denotes taking trace in the 3 × 3 flavor space.
+The baryon spin doublet are
+incorporated in field
+Sµ = − 1
+√
+3(γµ + vµ)γ5B6 + B∗
+6µ,
+(2.15)
+where the systematic baryon sextet B6 in 3 × 3 matrix reads
+B6 =
+�
+���
+Σ++
+c
+1
+√
+2Σ+
+c
+1
+√
+2Ξ′+
+c
+1
+√
+2Σ+
+c
+Σ0
+c
+1
+√
+2Ξ′0
+c
+1
+√
+2Ξ′+
+c
+1
+√
+2Ξ′0
+c
+Ω0
+c
+�
+��� .
+(2.16)
+The conjugation defines as usual for the spinor field ¯Smn
+µ
+= (Smn
+µ )†γ0. An asterisk on the
+symbol denotes the corresponding spin-3
+2 baryon, which is not involved in this work.
+Using above relevant Lagrangian and based on the one-boson exchange, we calculate
+the interaction kernel of ¯DΣc in isospin-1
+2 as
+K(s⊥) = F 2(s2
+⊥)
+�
+V1 + V2
+/s⊥
+|⃗s |
+�
+,
+(2.17)
+where F(s2
+⊥) denotes the regulator in the heavy hadron ( ¯D or Σc here) vertex; and the
+potential V1 and V2 is specifically expressed as,
+V1 = −2σ1σ2MD
+1
+E2σ
++ ρV1ρV2MD
+� 1
+E2ρ
+−
+1
+2E2ω
+�
+,
+V2 = −1
+3ρV1ρT2MD|s|
+� 2
+E2ρ
+− 1
+E2ω
+�
+,
+(2.18)
+where Eρ = (s2 + m2
+ρ)1/2 denotes the energy of the inter-mediator ρ meson, and similar for
+Eσ and Eω. The influence of the potential strength on the decay widths will be discussed
+later.
+There is no general method to choose the regulator functions. In this work, we use the
+following propagator-type form factor, namely,
+F(s2) =
+m2
+Λ
+s2 + m2
+Λ
+,
+(2.19)
+where mΛ is the introduced cutoff parameter to characterize the regulator function. Notice
+mΛ is the only free parameter in this analysis and can be determined by fitting bound
+state mass to the experimental data, which is found to be mΛ = 1.25 GeV for P N
+ψ (4312)+
+and close to the mass scale of the exchanged particle. In the limit s2 → 0, the heavy
+hadron is seen by the inter-mediator mesons as a point-like particle, and hence the form
+factor is normalized to 1. The cutoff value mΛ is usually believed to be much larger than
+the typical energy scale √2µϵ ∼ 0.1 GeV for P N
+ψ (4312)+ [15, 57], where µ =
+m1m2
+m1+m2 is the
+reduced mass of the two-hadron system and ϵ = (m1 +m2 −M) denotes the bound energy.
+– 6 –
+
+Our determined cutoff value is consistent with this universal estimation. The obtained V1
+and V2 for isospin-1
+2 are displayed graphically in Fig. 3(a).
+2.3
+Salpeter wave function for the JP = 1
+2
+− pentaquark states
+According to the spin-parity properties, and also considering the proper Lorentz structures,
+the Salpeter wave function of JP = 1
+2
+− pentaquarks consisting of a 0− meson and 1
+2
++ baryon
+can be generally constructed as
+ϕ(P, q⊥, r) =
+�
+f1 + f2
+/q⊥
+q
+�
+γ5u(P, r),
+(2.20)
+where the radial wave function f1(2)(|⃗q |) only explicitly depend on |⃗q |; u(P, r) denotes
+Dirac spinor with spin state r. In terms of the spherical harmonics Y m
+l , the wave function
+can be rewritten as
+ϕ(P, q⊥, r) = 2√π
+�
+f1Y 0
+0 + 1
+√
+3f2
+�
+Y 1
+1 γ− + Y −1
+1
+γ+ − Y 0
+1 γ3��
+γ5u(P, r),
+(2.21)
+where γ± = ∓ 1
+√
+2(γ1 ± iγ2). Then it is obvious to see that f1 and f2 represent the S- and
+P-wave components, respectively. Inserting the wave function into Eq. (2.22), we obtain
+the normalization satisfied by the radial wave functions as
+�
+d3q⊥
+(2π)3 2w1
+�
+f2
+1 + f2
+2
+�
+= 1.
+(2.22)
+Inserting the Salpeter wave function Eq. (2.20) into the Salpeter equation (2.7), elim-
+inating the spinor, calculating the trace, we can obtain two coupled eigenvalue equations
+with the pentaquark mass M as the eigenvalue and f1(2) as the eigen wave functions (see
+Ref. [26, 43, 44] for details). Solving the eigenvalue equations numerically, we can obtain
+the corresponding mass spectra and numerical wave functions, which are also graphically
+displayed in Fig. 3(b).
+3
+Strong decay of P N
+ψ (4312)+ → J/ψ(ηc)p within the BS wave function
+In this section, we first present the relevant effective Lagrangian; then we give the decay
+amplitude by using the BS wave function combining with the effective Lagrangian; finally,
+the expressions of the partial decay widths are presented in terms of the relevant form
+factors.
+For P N
+ψ (4312)+ → J/ψ(ηc)p, the involved interactions are J/ψDD(∗), ηcDD∗, and
+ΣcND(∗), which involve the Lagrangian of the doubly heavy meson and the heavy-light
+meson. The heavy-light charmed mesons in S-wave can be represented by [53, 58, 59]
+Hc = 1 + /v
+2
+(D∗µγµ + iDγ5),
+(3.1)
+where D∗µ and D denote the corresponding vector and pseudoscalar charmed D mesons
+respectively. The anti-heavy-light meson doublet H¯c has been presented in Eq. (2.12).
+– 7 –
+
+For doubly heavy mesons, the heavy quark flavor symmetry does not hold any longer,
+while the heavy quark spin symmetry still holds. In the ground states, the charmonium
+forms a doublet consisting of a pseudoscalar ηc and a vector state J/ψ, which can be
+represented by [60]
+R = 1 + /v
+2
+(ψµγµ + iηcγ5)1 − /v
+2
+,
+(3.2)
+where ψµ and ηc denotes the fields of the corresponding mesons. Here all the hadron fields
+in above equations contain a factor of √MH with MH the corresponding meson mass.
+By assuming the invariance under independent rotations of the heavy quark spins, it
+is possible to write down the effective coupling between the S-wave charmonia and the
+heavy-light mesons as [61]
+L2 = g2Tr
+�
+R ¯H¯c
+←→
+/∂ ¯Hc
+�
++ H.c.,
+(3.3)
+which is invariant under independent heavy quark spin symmetry; and the notation A←→
+∂ B ≡
+A∂B −∂AB is used. Consequently, we obtain the following effective Lagrangian describing
+J/ψ and ηc coupling to the DD∗,
+L2 = + gψDDψ†µ ¯D←→
+∂µD
+− igψDD∗ 1
+Mψ
+ϵµναβ∂µψ†
+ν( ¯D←→
+∂αD∗
+β + ¯D∗
+α
+←→
+∂β D)
++ gψD∗D∗ψ†µ( ¯D∗ν←→
+∂ν D∗
+µ + ¯D∗
+µ
+←→
+∂ν D∗ν − ¯D∗ν←→
+∂µD∗
+ν)
++ gDD∗ηcη†
+c(∂µ ¯DD∗µ − ¯D∗µ∂µD)
++ igD∗D∗ηc
+1
+Mηc
+ϵµναβ∂µη†
+c ¯D∗
+ν
+←→
+∂αD∗
+β + H.c.,
+(3.4)
+where we have divide a meson mass in the second and the last Lagrangians to keep all
+the coupling constants dimensionless. The symbol ϵµναβ denotes the totally antisymmetric
+Levi-Civita tensor with ϵµναβ = −ϵµναβ and convention ϵ0123 = 1. All these coupling con-
+stants are related to a single coupling g2, which is determined to be g2 = �Mψ/(2MDfψ)
+with fψ denoting the J/ψ decay constant [61]. Then all other coupling constants can also
+be expressed in terms of the gψDD as
+gψDD = Mψ
+fψ ,
+gψDD∗ =
+�
+MD∗
+MD
+�1/2
+gψDD,
+gψD∗D∗ =
+�
+MD∗
+MD
+�
+gψDD,
+gDD∗ηc =
+�
+MηcMD∗
+MψMD
+�1/2
+gψDD,
+gD∗D∗ηc =
+�
+Mηc
+Mψ
+�1/2 MD∗
+MD gψDD.
+(3.5)
+– 8 –
+
+In next section, we will also discuss the effects of these coupling constants on the final
+decay widths.
+3.1
+Amplitude for P N
+ψ (4312)+ → J/ψp
+P N
+ψ (4312)+ as the ¯DΣc molecular state can decay to J/ψp by exchanging either a D or a
+D∗ virtual meson, and the total amplitude is the sum of the two.
+Pc, P
+¯D, k1
+Σc, k2
+J/ψ, p1
+p, p2
+D, k3
+Pc, P
+¯D, k1
+Σc, k2
+J/ψ, p1
+p, p2
+µ
+ν
+D∗, k3
+Fig. 2: Strong decay of P N
+ψ (4312)+ to the J/ψp by exchanging a virtual mediator D (left
+panel) and D∗ (right panel). P, k1, k2, P1, P2 denote the momenta of P N
+ψ , constituent
+meson, constituent baryon, the final J/ψ, and the final p respectively.
+3.1.1
+Amplitude with D exchange
+The left panel of Fig. 2 shows the Feynman diagram of P N
+ψ (4312)+ → J/ψp by exchanging
+the D meson. Besides the pentaquark vertex, we also need two other effective Lagrangian
+to obtain the decay width. From above results, the effective Lagrangian describing the
+DDJ/ψ interaction read
+LψDD = gψDDψ†µ( ¯D∂µD − ∂µ ¯DD),
+(3.6)
+where D = (D0, D+, D+
+s )T represents the charmed meson fields in flavor triplet. Whereas
+the effective Lagrangian for NDΣc interaction behaves as [10, 62]
+LNDΣc = −igNDΣc ¯Nγ5Σc ¯D† + H.c.,
+(3.7)
+where N stands for nucleon field doublet; Σc = σ · Σc with σ denoting the Pauli matrix
+and Σc denoting the Σc baryon isospin triplet.
+The invariant amplitude for P N
+ψ (4312)+ → J/ψp by exchanging a D can then be
+expressed by the Bethe-Salpeter vertex as
+A1 =
+�
+d4k
+(2π)4 ¯u2(−igNDΣc)γ5 [S(k2)Γ(P, k, r)D(k1)] (gψDD)D(k3)(i)e∗
+1 · (k3 + k1),
+(3.8)
+where u2 is short for u(r2)(P2) with r2 representing the proton spin state; e1 is short for
+e(r1)(P1) representing the polarization vector of the final J/ψ with P1 denoting the J/ψ
+– 9 –
+
+momentum and r1 = 0, ±1 representing the 3 possible polarization states. The polarization
+vector e1 fulfills the Lorentz condition
+eα
+1 P1α = 0.
+(3.9)
+The momentum of the exchanged virtual charmed meson is denoted as k3 = (k1 − P1). We
+will use M1 and M2 to denote the masses of the final J/ψ and proton respectively.
+Also notice k1 = (α1P + k) and k3 are involved in the four-dimensional integration
+over k. To simplify this amplitude, first we strip off the triangle amplitude involved the
+integral over k as
+A1T;αu(P, r) = γ5
+�
+d4k
+(2π)4 [S(k2)Γ(P, k, r)D(k1)] D(k3)(2k1α),
+(3.10)
+where the Lorentz condition of the vector meson is utilized, and we also strip off the spinor
+u(P, r) for later convenience. The decay amplitude A can then be simplified as
+A1 = (gNDΣcgψDD)e∗α
+1 ¯u2A1T;αu(P, r).
+(3.11)
+Then we perform the contour integral over kP on Eq. (3.10), and obtain
+Aα
+1T u(P, r) = γ5
+�
+dk3
+⊥
+(2π)3
+1
+w3
+(aα
+1 ϕ+ + aα
+2 ϕ−),
+(3.12)
+where we used the expression of the positive(negative) energy wave functions ϕ± = Λ±γ0ϕ;
+the two coefficients a1 and a2 behaves as
+aα
+1 = c1xα
+1 + c2xα
+2 + c3xα
+3 ,
+(3.13)
+aα
+2 = c4xα
+4 + c5xα
+5 + c6xα
+6 ,
+(3.14)
+where xi = k1(kP = kPi) with (i = 1, · · · 6), and kPis are defined as
+kP1 = ζ+
+1 , kP2 = ζ+
+2 , kP3 = ζ+
+3 , kP4 = ζ−
+1 , kP5 = ζ−
+2 , kP6 = ζ−
+3 ,
+(3.15)
+where the abbreviations ζ±
+1 ≡ −(α1M ∓w1), ζ±
+2 ≡ (α2M ∓w2), and ζ±
+3 ≡ (E1 −α1M ±w3)
+are used. The coefficients cis (i = 1, · · · 6) are defined as
+c1(4) =
+1
+w1 + w3 ∓ E1
+,
+c2(5) =
+(−1)
+w2 + w3 ∓ E2
+,
+c3(6) =
+(w1 + w2 ∓ M)
+(w1 + w3 ± E1)(w2 + w3 ∓ E2).
+(3.16)
+Now the amplitude A1T;α has been expressed by the three-dimensional Salpeter wave func-
+– 10 –
+
+tion ϕ(k⊥), and can be further simplified as the two form factors
+A1T;α = (s11γα + s12 ˆPα),
+(3.17)
+Inserting the obtained P N
+ψ (4312)+ wave function, namely, Eq. (2.20), into Eq. (3.12) and
+then calculating the three-dimensional integral numerically, we can obtain the amplitude
+Aα
+1T in terms of s11 and s12. In the appendix A we collect the specific expressions of the
+two form factors in terms of the Salpeter wave functions f1 and f2. The amplitude A1 then
+behaves as
+A1 = (gNDΣcgψDD)e∗α
+1 ¯u2(s11γα + s12 ˆPα)u(P, r).
+(3.18)
+3.1.2
+Amplitude with D∗ exchange
+For P N
+ψ (4312)+ decaying by exchanging D∗ in the lowest level, the relevant Feynman
+diagram is displayed in the right panel of Fig. 2, and the two involved interaction vertexes
+are DD∗J/ψ and ΣcD∗p. The DD∗J/ψ interaction is represented by the following effective
+Lagrangian
+LψDD∗ = (−i)gψDD∗ϵµναβ 1
+Mψ
+∂µψ†
+ν( ¯D∂αD∗
+β − ∂α ¯DD∗
+β),
+(3.19)
+Notice here the coupling constant gψDD∗ is defined to have the same dimension with the
+pseudoscalar coupling constant gψDD. The effective Lagrangian of ND∗Σc reads [62]
+LND∗Σc = gND∗Σc ¯NγαΣcD∗†
+α + H.c..
+(3.20)
+All the coupling constants in these effective Lagrangian will be specified in next section.
+The invariant amplitude for P N
+ψ (4312)+ → J/ψp by exchanging a D∗ can be expressed
+by the Bethe-Salpeter vertex as
+A2 = (gΣcND∗gψDD∗)¯u2γν
+�
+d4k
+(2π)4 [S(k2)Γ(P, k, r)D(k1)] Dµν(k3) 1
+M1
+ϵP1αβµe∗
+1α(k3 + k1)β,
+(3.21)
+where the propagator of the exchanged D∗ meson behaves as
+Dµν(k3) = i−gµν + k3µk3ν/m2
+3
+k2
+3 − m2
+3 + iϵ
+,
+(3.22)
+Here the propagator mass m3 is MD∗. Notice the contraction with Levi-Civita tensor forces
+the momentum part in numerator of Dµν(k3) to be zero. The amplitude can be further
+simplified as
+A2 = (gΣcND∗gψDD∗)e∗α
+1 ¯u2γµ 1
+M1
+ϵP1αβµAβ
+2T u(P, r),
+(3.23)
+– 11 –
+
+where we have stripped off the amplitude involved the integral over k as before
+Aβ
+2T u(P, r) = −
+�
+d4k
+(2π)4 [S(k2)Γ(P, k, r)D(k1)] D(k3)(2kβ
+1 ).
+(3.24)
+In order to express the amplitude by the three-dimensional Salpeter wave function, we
+perform the contour integration over kP on Eq. (3.24) as usual and obtain
+Aβ
+2T u(P, r) = −
+�
+d3k⊥
+(2π)3
+1
+w3
+�
+aβ
+1ϕ+ + aβ
+2ϕ−�
+.
+(3.25)
+Combining Aβ
+2T u(P, r) with γµ 1
+M1 ϵP1αβµ, and using the following identity of the Levi-Civita
+symbol,
+iγµϵµαβν = γ5(γαγβγν − γαgβν + γβgαν − γνgαβ),
+(3.26)
+we can express the decay amplitude A2 for D∗ exchange by the following form factors
+A2 = (gψDD∗gΣcND∗)e∗α
+1 ¯u2(s21γα + s22 ˆPα)u(P, r).
+(3.27)
+Namely, the amplitude A2 can be expressed by the same form as A1, which is just what it
+should be. In above equations, the specific expressions of s21 and s22 can be obtained by in-
+serting the Salpeter wave functions into Eq. (3.25) and performing the integral numerically.
+The specific expressions are presented in the appendix A.
+3.2
+Amplitude for P N
+ψ (4312)+ → ηcp
+Since ηc with JP = 0−, the decay to ηcp can only happen by exchanging a D∗ while the
+mode of exchanging a D is forbidden. From Eq. (3.4), the effective Lagrangian responsible
+for DD∗ηc interaction reads
+LDD∗ηc = gDD∗ηcη†
+c∂µ ¯DD∗µ.
+(3.28)
+The effective Lagrangian describing ND∗Σc interaction has been presented in Eq. (3.20).
+The corresponding Feynman diagram is similar with that for the decay to J/ψp with D∗
+exchange. The decay amplitude for P N
+ψ (4312)+ → ηcp behaves as
+A = gND∗Σc ¯u2γα
+�
+d4k
+(2π)4 [S(k2)Γ(k, r)D(k1)]Dαβ(k3)gD∗Dηc(−iP β
+1 ).
+(3.29)
+As usual, it is convenient to strip off the part involved the integral over k as
+AT u(P, r) =
+�
+d4k
+(2π)4 (γαP β
+1 )[S(k2)Γ(k, r)D(k1)]Dαβ(k3),
+– 12 –
+
+Performing the contour integral over kP , we can express AT by the three-dimensional
+Salpeter wave function
+AT u(P, r) =
+�
+d3k⊥
+(2π)3
+1
+2w3
+(γαP β
+1 )
+3
+�
+i=1
+�
+cidαβ(yi)ϕ+ + ci+3dαβ(yi+3)ϕ−�
+,
+(3.30)
+where the positive (negative) energy wave function is related to the Salpeter wave function
+by ϕ± = Λ±γ0ϕ; and we define dαβ and yi as
+dαβ(yi) = −gαβ + yiαyiβ
+m2
+3
+,
+yi = k3(kP = kPi) = xiP ˆP + k⊥ − P1.
+(3.31)
+Inserting the Salpeter wave function Eq. (2.20) of P N
+ψ (4312)+, we obtain AT expressed by
+one form factor,
+AT = s0γ5.
+(3.32)
+Finally, we obtain the amplitude for decay to ηcp by form factor s0 with a simple form
+A = −i (gND∗ΣcgηcDD∗) s0[¯u2γ5u(P, r)].
+(3.33)
+The expression of s0 is also listed in appendix A as the integral over Salpeter wave functions.
+3.3
+Partial decay width
+Combing the two amplitudes from D and D∗ mediators together, we obtain the full invari-
+ant amplitude for P N
+ψ (4312)+ → J/ψp decay by two form factors,
+A = A1 + A2 = e∗α
+1 ¯u2
+�
+s1γα + s2 ˆPα
+�
+u(P, r).
+(3.34)
+where s1 and s2 are related to the coupling constants and are expressed as
+s1 = gψDDgNDΣcs11 + gψDD∗gND∗Σcs21,
+s2 = gψDDgNDΣcs12 + gψDD∗gND∗Σcs22.
+(3.35)
+Squaring the amplitude, summing all the polarization states, we obtain
+�
+r1,r2,r
+|A|2 =
+�
+−gαβ + P α
+1 P β
+1
+M2
+1
+�
+Tr
+�/P2 + M2
+�
+AT;α
+�/P + M
+� ¯
+AT;β,
+(3.36)
+where
+AT;α =
+�
+s1γα + s2 ˆPα
+�
+,
+(3.37)
+– 13 –
+
+and ¯
+AT;β = γ0A†
+T;βγ0 is defined as the usual conjugation variable; we also used the rela-
+tionship of the summation over the vector polarization states r1,
+�
+r1
+eα
+(r1)eβ
+(r1) = −gαβ + P α
+1 P β
+1
+M2
+1
+;
+(3.38)
+and the summation over the polarization states of the spinors
+�
+r2
+u(r2)(P2)¯u(r2)(P2) = (/P2 + M2),
+(3.39)
+�
+r
+u(P, r)¯u(P, r) = (/P + M).
+(3.40)
+For P N
+ψ (4312)+ → ηcp decay, the squared amplitude behaves as
+�
+r2,r
+|A|2 = Tr
+�/P2 + M2
+�
+A
+�/P + M
+� ¯
+A = 4(gND∗ΣcgηcD∗D)2s2
+0M(E2 − M2).
+(3.41)
+Namely, the squared amplitude is proportional to the kinetic energy (E2 − M2) of the
+proton.
+Finally, the partial decay width of P N
+ψ (4312)+ → J/ψ(ηc)p is expressed as
+Γ[P N
+ψ (4312)+ →J/ψ(ηc)p] =
+|P1|
+8πM2
+1
+2
+�
+r,r1,r2
+|A|2,
+(3.42)
+where the three momentum of J/ψ(ηc) is given by
+|P1| =
+1
+2M
+��
+M2 − (M1 + M2)2��
+M2 − (M1 − M2)2�� 1
+2 .
+(3.43)
+4
+Numerical results and discussions
+Before giving the decay widths, we first specify the effective interaction coefficients we used
+in above effective Lagrangian. The interaction coefficients between the heavy hadron and
+the light bosons read [6, 9, 12, 14, 26]: ρV1 = βgV
+√
+2 = 3.75, ρT1 = λgV
+√
+2 = 2.34 GeV−1, and
+σ1 = 0.76; ρV2 = βSgV
+√
+2
+= 7.26, ρT2 = λSgV
+√
+2
+= 13.81 GeV−1, and σ2 = 6.2. In the heavy
+quark limit, the coupling constants between the heavy hadrons read [61] gDDψ = Mψ
+fψ and
+gDD∗ψ = ( MD∗
+MD )1/2gDDψ with the J/ψ decay constant fψ = 0.416 GeV estimated from the
+dilepton decay width [63]; the DD∗ηc coupling constant reads gDD∗ηc = ( MηcMD∗
+MψMD )1/2gDDψ.
+Combined with the total amplitude Eq. (3.34), it can be found that the partial decay width
+is proportional to
+1
+f2
+ψ . The coupling constants related to the baryons used are gNDΣc = 2.69
+and gND∗Σc = 3.0 [10, 62]. These values are the standard parameters used in this work, and
+we will also vary the standard parameters to explore their influence on the wave functions
+and the final decay widths.
+– 14 –
+
+ GeV
+s
+0
+0.5
+1
+1.5
+2
+2.5
+3
+-1
+ GeV
+2
+F
+iV
+25
+−
+20
+−
+15
+−
+10
+−
+5
+−
+0
+1
+ V
+2
+ V
+(a)
+ GeV
+q
+0
+0.2
+0.4
+0.6
+0.8
+1
+1.2
+1.4
+-2
+BS radial wave function GeV
+0
+10
+20
+30
+40
+1f
+
+2f
+
+(b)
+ GeV
+q
+0
+0.5
+1
+1.5
+2
+2.5
+3
+-2
+BS radial wave function GeV
+0
+1
+2
+3
+4
+)
+n
+V
+(0.5
+1f
+
+)
+n
+V
+(0.5
+2f
+
+(c)
+ GeV
+q
+0
+0.2
+0.4
+0.6
+0.8
+1
+1.2
+1.4
+-2
+BS radial wave function GeV
+0
+10
+20
+30
+40
+)
+n
+V
+(1.5
+1f
+
+)
+n
+V
+(1.5
+2f
+
+(d)
+Fig. 3: The figure (a) shows the isospin-1
+2 potentials VnF 2 (n = 1, 2). Figure (b) is the
+Bethe-Salpeter radial wave function f1 and f2 for P N
+ψ (4312)+ as the ¯DΣc molecular state
+based on the one-boson exchange; while (c) and (d) are the radial wave functions when the
+interaction potential Vn in Eq. (2.18) is reduced and increased by 50% based, respectively,
+where the corresponding regulator values are mΛ = 1.288 GeV and 0.73 GeV respectively.
+The only free parameter in this work is the regulator mΛ in the form factor F(s2) in
+Eq. (2.19). All the other parameters have been determined by the previous experimental
+data. By solving the relevant BS eigenvalue equation, we find proper cutoff values of mΛ can
+produce bound state of ¯DΣc based on the one-boson exchange kernel in isospin-1
+2. Then by
+fitting the bound state mass M to the experimental measurement MP N
+ψ (4312)+ = 4.312 GeV,
+we fix mΛ in Eq. (2.19) to be 0.87 GeV. Then the obtained V1 and V2 in the interaction
+kernel are graphically shown in Fig. 3(a).
+In Fig. 3(b) we show the obtained BS wave functions f1 and f2 for P N
+ψ (4312)+. On
+the other hand, the obtained radial wave functions depend on the obtained potential V1(2),
+which is directly related to the coupling constants σ1(2), ρV1(2) and ρT2. To reflect the
+influence of these parameters on the wave functions and decay widths, we vary the nu-
+merical values of V1(2) under standard parameters by ∓50%. Under these variations, the
+– 15 –
+
+obtained regulator values are then mΛ = 1.288 GeV and 0.73 GeV respectively, and the
+corresponding wave functions obtained are displayed in Fig. 3(c) and Fig. 3(d). As V1(2)
+decreases, the fitted regulator parameter mΛ increases, and also the role of wave function
+f1 becomes more important.
+Our results of the mass spectra for I(JP ) = 1/2(1/2)− ¯DΣc molecule indicate that
+there only exists one bound state, namely, P N
+ψ (4312)+ as the ground state of ¯DΣc molecule.
+Our results do not support any radially excited states. This conclusion is robust even under
+the ±50% change of the interaction kernel V1(2).
+The obtained numerical values of the form factors in amplitudes are
+s0 = 5.1 × 10−3
+for decay to ηcp channel in Eq. (3.33); and form factors for decay to J/ψp channel in
+Eq. (3.34) are
+s11 = 2.3 × 10−3, s12 = 3.7 × 10−3, s21 = −8.2 × 10−3, s22 = 1.5 × 10−2.
+Inserting above form factors into the decay width expressions, we obtain the partial decay
+widths as Γ(P N
+ψ (4312)+ →J/ψp) = 0.11 MeV and Γ(P N
+ψ (4312)+ →ηcp) = 0.056 MeV. The
+obtained partial width for decay to J/ψp is ∼ 1% of the total width of Γ = 9.8 ± 2.7+3.7
+−4.5
+MeV [1] reported by the LHCb collaboration. The J/ψp channel is also the only observed
+decay mode of P N
+ψ (4312)+ currently. While the decay fraction of P N
+ψ (4312)+ to ηcp is about
+50% smaller than the J/ψp channel. There is still no evident signal in recent experimental
+search of P N
+ψ (4312)+ in ηcp channel [42]. Notice the obtained results are totally predictive
+and there are no any free adjustable parameters since the regulator mΛ has been fixed by
+the mass of P N
+ψ (4312)+. When the interaction kernel V1(2) varies by ±50% based on the
+standard parameters, the obtained decay widths are 2.89 MeV and 0.025 MeV for J/ψp
+channel, respectively; while for ηcp channel, the results are 0.012 MeV and 0.0024 MeV,
+respectively. Namely, our results indicate that the fraction of J/ψp decay channel can
+amount to ∼ 30% when the strength of V1(2) is reduced by half.
+Tab. I: Comparison of partial decay width of P N
+ψ (4312)+ → J/ψ(ηc)p with other works
+in units of MeV, where our theoretical uncertainties are induced by varying the relevant
+coupling constants by ±10% in the effective Lagrangian. Our results also indicate that
+the decay width of J/ψp channel can amount to 2.89 MeV when the strength of V1(2) is
+reduced by half.
+Channel
+This
+[38]
+[40]
+[15]
+[37]
+[10]
+[24]
+J/ψp
+0.11−0.04
++0.06
+0.32 ± 0.08
+1.67+0.92
+−0.56
+10−3 ∼ 0.1
+0.033
+(3 ∼ 8)
+9.3+19.5
+−9.3
+ηcp
+0.056−0.019
++0.026
+0.98 ± 0.25
+5.54+0.75
+−0.50
+10−2 ∼ 0.4
+0.066
+−
+0.26−0.24
++0.55
+A comparison of our results with other works is listed in Tab. I. Our obtained partial
+decay widths are roughly consistent with those in Refs. [15, 37, 38]. Notice the theoretical
+results for decay widths of P N
+ψ (4312)→J/ψ(ηc)p are quite different from each other for the
+– 16 –
+
+complication of this problem. More researches are needed. Since the obtained partial decay
+widths are also directly dependent on the coupling constants gψDD, gψDD∗, gNDΣc, gND∗Σc,
+and gDD∗ηc in the relevant effective Lagrangian. To see the sensitivity of the our partial
+decay widths on these parameters, we calculate the theoretical uncertainties by varying
+the every coupling constant by ±10%, and then searching the parameter space to find the
+maximum deviation. The obtained theoretical errors are also listed in above Tab. I, where
+the width uncertainties induced from the coupling constants amount to about ∼ 0.05 MeV
+for channel J/ψp, and ∼ 0.02 MeV for channel ηcp.
+Finally, we give a brief summary. In this work, firstly, based on the effective Lagrangian
+in the the heavy quark limit, we calculate the one-boson-exchange interaction kernel of
+¯DΣc in the isospin-1
+2 state. Then by using the Bethe-Salpeter equation, we obtain the
+mass spectrum and wave functions of the experimental P N
+ψ (4312)+ as the ¯DΣc molecular
+state with JP = ( 1
+2)−. Finally combining the effective Lagrangian and the obtained BS
+wave function, we calculate the partial decay width to be 0.11−0.04
++0.06 MeV for P N
+ψ (4312)+ →
+J/ψ, and 5.6−1.9
++2.6 × 10−2 MeV for ηcp channel.
+The obtained results indicate that the
+fraction of P N
+ψ (4312)+ → J/ψp amounts to ∼ 1%, and can even reach to ∼ 30% when
+the interaction kernel is reduced by half. Our results are roughly consistent with some
+other calculations and the LHCb experimental measurements. However, more theoretical
+analysis and experimental measurements are necessary to determine the properties of the
+pentaquark state P N
+ψ (4312)+. The interpretation of P N
+ψ (4312)+ as the ¯DΣc molecular state
+with JP = ( 1
+2)− and isospin I = 1
+2 is favored by this work.
+A
+Expressions of the decay form factors
+For completeness, we list the specific expressions of the relevant form factors here, which
+are all represented by the integral over the radial Salpeter wave functions f1 and f2. Parts
+of following expressions are calculated with the help of the FeynCalc package [64–66]. The
+four form factors for decay P N
+ψ (4312)+ →J/ψp in Eq. (3.35) are
+s11 = −
+�
+d3k⊥
+(2π)3
+1
+w2
+c22k [kv0f1 + (w2u0 + m2v0) f2] ,
+s12 =
+�
+d3k⊥
+(2π)3
+1
+w2P 2
+1
+(Y1f1 + Y2f2) ,
+s21 = −
+�
+d3k⊥
+(2π)3
+1
+w2P 2
+1
+(Z1f1 + Z2f2) ,
+s22 = −
+�
+d3k⊥
+(2π)3
+1
+w2P 2
+1
+(Z3f1 + Z4f2) ,
+(A.1)
+where Y1, Y2, and Z1 ∼ Z4 read
+Y1 =P 2
+1 w2u1 + c1E1kP1v1 + c1kM2P1v1 − c1kMP1v1 − m2P 2
+1 v1 − c1E1kP1w2u0
+− c21E2
+1k2v0 − c21E1k2M2v0 + c21E1k2Mv0 + c1E1km2P1v0 + c22k2P 2
+1 v0,
+(A.2)
+– 17 –
+
+Y2 =c1E1P1w2u1 − c1MP1w2u1 + c1M2P1w2u1 − c1E1m2P1v1 − 2c1E2m2P1v1
++ c1m2M2P1v1 + c1m2MP1v1 + kP 2
+1 v1 − c21E2
+1kw2u0 + c21E1kMw2u0
+− c21E1kM2w2u0 + c22kP 2
+1 w2u0 + c21E2
+1km2v0 + 2c21E1E2km2v0
+− c1E1k2P1v0 − c21E1km2M2v0 − c21E1km2Mv0 + c22km2P 2
+1 v0,
+(A.3)
+Z1 = − E1P 2
+1 w2u1 + MP 2
+1 w2u1 + M2P 2
+1 w2u1 − c1E2
+1kP1v1 + c1kM2
+1 P1v1 + E1m2P 2
+1 v1
+− m2MP 2
+1 v1 − m2M2P 2
+1 v1 + c1E2
+1kP1w2u0 − c1E1kMP1w2u0 − c1E1kM2P1w2u0
++ c21E3
+1k2v0 − c1E2
+1km2P1v0 − c22E1k2P 2
+1 v0 − c21E1k2M2
+1 v0
++ c1E1km2MP1v0 + c1E1km2M2P1v0 − c22k2MP 2
+1 v0 − c22k2M2P 2
+1 v0,
+(A.4)
+Z2 =c1M2
+1 P1w2u1 + c1E2
+1m2P1v1 + 2c1E1E2m2P1v1 − 2c1E1m2MP1v1
++ c1m2M2
+1 P1v1 + kMP 2
+1 v1 + kM2P 2
+1 v1 + c21E3
+1kw2u0 + c1E2
+1k2P1v0
+− c21E1kM2
+1 w2u0 − c22E1kP 2
+1 w2u0 − c22kMP 2
+1 w2u0 − c22kM2P 2
+1 w2u0
++ 2c21E2
+1km2Mv0 − c1E1k2MP1v0 − c1E1k2M2P1v0 + c22E1km2P 2
+1 v0
+− c21E1km2M2
+1 v0 + 2c22E2km2P 2
+1 v0 − 3c22km2MP 2
+1 v0 − c22km2M2P 2
+1 v0
+− 2c21E2
+1E2km2v0 − c21E3
+1km2v0 − E1kP 2
+1 v1 − c1E2
+1P1w2u1,
+(A.5)
+Z3 = − MP 2
+1 w2u1 − M2P 2
+1 w2u1 + c1E1kMP1v1 − c1E1kM2P1v1 − c1kM2
+1 P1v1
++ m2M2P 2
+1 v1 + c1E1kMP1w2u0 + c1E1kM2P1w2u0 − c21E2
+1k2Mv0
++ c21E2
+1k2M2v0 + c21E1k2M2
+1 v0 − c1E1km2MP1v0 − c1E1km2M2P1v0
++ m2MP 2
+1 v1 − c22k2M2P 2
+1 v0 + 3c22k2MP 2
+1 v0,
+(A.6)
+Z4 =c1E1MP1w2u1 − c1E1M2P1w2u1 − c1M2
+1 P1w2u1 + c1E1m2MP1v1
+− c1m2M2
+1 P1v1 − kMP 2
+1 v1 − kM2P 2
+1 v1 − c21E2
+1kMw2u0 + c21E2
+1kM2w2u0
++ c21E1kM2
+1 w2u0 + 3c22kMP 2
+1 w2u0 − c22kM2P 2
+1 w2u0 − c21E2
+1km2Mv0
++ c21E2
+1km2M2v0 + c1E1k2MP1v0 + c1E1k2M2P1v0 + c21E1km2M2
+1 v0
++ 3c22km2MP 2
+1 v0 − c22km2M2P 2
+1 v0 − c1E1m2M2P1v1.
+(A.7)
+In above expressions, P1 = |P1|, and
+c = cos θ,
+c21 = 1
+2(3 cos2 θ − 1),
+c22 = 1
+2(cos2 θ − 1),
+(A.8)
+where θ denotes the angle between k and P1. We also define un and vn (n = 0, 1, 2) for
+later convenience
+un = (c1xn
+1P + c2xn
+2P + c3xn
+3P ) + (c4xn
+4P + c5xn
+5P + c6xn
+6P ),
+vn = (c1xn
+1P + c2xn
+2P + c3xn
+3P ) − (c4xn
+4P + c5xn
+5P + c6xn
+6P ).
+(A.9)
+The expressions of ci are listed in Eq. (3.16).
+– 18 –
+
+The form factor s0 in Eq. (3.33) for P N
+ψ (4312)+ → ηcp decay behaves
+s0 =
+�
+d3k⊥
+(2π)3
+1
+4m2
+3w2w3P 2
+1
+�
+(P 2
+1 X1 + kc1X3)f1 + (kP 2
+1 X2 + c1X4)f2
+�
+,
+(A.10)
+where X1 ∼ X4 read
+X1 = − ckMP1w2u0 − ckM2P1w2u0 + m2
+3Mw2u0 + m2
+3M2w2u0 − MM2
+1 w2u0
+− M2
+1 M2w2u0 + ckm3P1w2u1 + E1m3Mw2u1 + E1m3M2w2u1 + m3M2
+1 w2u1
+− E1m2
+3w2u2 + ck3P1v0 + k2M2
+1 v0 + ckm2MP1v0 + ckm2M2P1v0
+− ckm2m3P1v1 − m2m2
+3Mv0 + m2MM2
+1 v0 − m2m2
+3M2v0 + m2M2
+1 M2v0
+− E1k2m3v1 − E1m2m3Mv1 − E1m2m3M2v1 + E1m2m2
+3v2 − m2m3M2
+1 v1,
+(A.11)
+X2 =ckP1w2u0 + M2
+1 w2u0 − E1m3w2u1 + ckm2P1v0 + ckm3P1v1 + m3M2
+1 v1
+− ckMP1v0 − ckM2P1v0 + m2
+3Mv0 + m2
+3M2v0 + m2M2
+1 v0 − MM2
+1 v0
+− M2
+1 M2v0 + E1m3Mv1 + E1m3M2v1 − E1m2m3v1 − E1m2
+3v2,
+(A.12)
+X3 = − cE1kP1w2u0 − E1M2
+1 w2u0 + ckMP1w2u0 + ckM2P1w2u0 + MM2
+1 w2u0
++ M2M2
+1 w2u0 + E2
+1m3w2u1 − E1m3Mw2u1 − E1m3M2w2u1 + cE1km2P1v0
+− cE1km3P1v1 − cE1kMP1v0 − cE1kM2P1v0 + E1m2M2
+1 v0 + E1m2
+3Mv0
++ E1m2
+3M2v0 − E1MM2
+1 v0 − E1M2M2
+1 v0 + ckm3MP1v1 + ckm3M2P1v1
+− ckm2MP1v0 − ckm2M2P1v0 + ckM2
+1 P1v0 − m2
+3M2
+1 v0 − m2MM2
+1 v0
+− m2M2M2
+1 v0 + M4
+1 v0 + E2
+1m3Mv1 + E2
+1m3M2v1 − E2
+1m2m3v1 + E2
+1m2
+3v2
+− 2E1m3M2
+1 v1 + E1m2m3Mv1 + E1m2m3M2v1 − E1m2
+3Mv2 − E1m2
+3M2v2
++ m3MM2
+1 v1 + m3M2M2
+1 v1,
+(A.13)
+X4 =v0m2M4
+1 + u0w2M4
+1 − v0m2m2
+3M2
+1 + E1k2v0M2
+1 + 2E2k2v0M2
+1 + m2m3M2v1M2
+1
+− k2Mv0M2
+1 − E1Mv0m2M2
+1 + k2v0M2M2
+1 − E1v0m2M2M2
+1 + ckv0m2P1M2
+1
+− u0m2
+3w2M2
+1 − E1Mu0w2M2
+1 − E1u0M2w2M2
+1 + cku0P1w2M2
+1 + 2E2m3w2u1M2
+1
+− Mm3w2u1M2
+1 + m3M2w2u1M2
+1 − 2E1m2m3v1M2
+1 + Mm2m3v1M2
+1
++ E1Mv0m2m2
+3 + E1v0m2m2
+3M2 + cE1k3v0P1 + 2cE2k3v0P1 − ck3Mv0P1
+− cE1kMv0m2P1 + ck3v0M2P1 − cE1kv0m2M2P1 + E1Mu0m2
+3w2 + E1u0m2
+3M2w2
+− cE1kMu0P1w2 − cE1ku0M2P1w2 + E2
+1Mm3w2u1 + E2
+1m3M2w2u1 + cE1km3P1w2u1
++ 2cE2km3P1w2u1 − ckMm3P1w2u1 + ckm3M2P1w2u1 − E2
+1m2
+3w2u2 − 2E1E2m2
+3w2u2
++ E1Mm2
+3w2u2 − E1m2
+3M2w2u2 − E2
+1k2m3v1 − 2E1E2k2m3v1 + E1k2Mm3v1
++ E2
+1Mm2m3v1 − E1k2m3M2v1 + E2
+1m2m3M2v1 − cE1km2m3P1v1 + ckMm2m3P1v1
++ ckm2m3M2P1v1 + E2
+1m2m2
+3v2 − E1Mm2m2
+3v2 − E1m2m2
+3M2v2.
+(A.14)
+– 19 –
+
+Acknowledgments
+The author Q. Li thanks Prof. Fen-Kun Guo of ITP-CAS, and Dr. Xu-Chang Zheng of
+Chongqing Univ., and Dr. Hao Xu of Northwest Normal Univ. for helpful suggestions
+and discussions.
+This work is supported by the National Natural Science Foundation
+of China (NSFC) under Grant Nos. 12005169, 12075301, 11821505, 12047503, 11805024,
+11865001, and 12075073.
+It is also supported by the National Key R&D Program of
+China (2022YFA1604803), the Natural Science Basic Research Program of Shaanxi (Program
+No. 2021JQ-074), and the Fundamental Research Funds for the Central Universities.
+References
+[1] R. Aaij, et al., Phys. Rev. Lett. 122 (22) (2019) 222001. arXiv:1904.03947,
+DOI:10.1103/PhysRevLett.122.222001.
+[2] T. Gershon (6 2022). arXiv:2206.15233.
+[3] J.-J. Wu, R. Molina, E. Oset, B. S. Zou, Phys. Rev. Lett. 105 (2010) 232001.
+arXiv:1007.0573, DOI:10.1103/PhysRevLett.105.232001.
+[4] W. L. Wang, F. Huang, Z. Y. Zhang, B. S. Zou, Phys. Rev. C 84 (2011) 015203.
+arXiv:1101.0453, DOI:10.1103/PhysRevC.84.015203.
+[5] J.-J. Wu, T. S. H. Lee, B. S. Zou, Phys. Rev. C 85 (2012) 044002. arXiv:1202.1036,
+DOI:10.1103/PhysRevC.85.044002.
+[6] Z.-C. Yang, Z.-F. Sun, J. He, X. Liu, S.-L. Zhu, Chin. Phys. C 36 (2012) 6–13.
+arXiv:1105.2901, DOI:10.1088/1674-1137/36/1/002,10.1088/1674-1137/36/3/006.
+[7] X.-Q. Li, X. Liu, Eur. Phys. J. C 74 (12) (2014) 3198. arXiv:1409.3332,
+DOI:10.1140/epjc/s10052-014-3198-3.
+[8] M. Karliner, J. L. Rosner, Phys. Rev. Lett. 115 (12) (2015) 122001. arXiv:1506.06386,
+DOI:10.1103/PhysRevLett.115.122001.
+[9] R. Chen, X. Liu, X.-Q. Li, S.-L. Zhu, Phys. Rev. Lett. 115 (13) (2015) 132002.
+arXiv:1507.03704, DOI:10.1103/PhysRevLett.115.132002.
+[10] C.-J. Xiao, Y. Huang, Y.-B. Dong, L.-S. Geng, D.-Y. Chen, Phys. Rev. D 100 (1) (2019)
+014022. arXiv:1904.00872, DOI:10.1103/PhysRevD.100.014022.
+[11] M.-Z. Liu, Y.-W. Pan, F.-Z. Peng, M. S´anchez S´anchez, L.-S. Geng, A. Hosaka,
+M. Pavon Valderrama, Phys. Rev. Lett. 122 (24) (2019) 242001. arXiv:1903.11560,
+DOI:10.1103/PhysRevLett.122.242001.
+[12] R. Chen, Z.-F. Sun, X. Liu, S.-L. Zhu, Phys. Rev. D 100 (1) (2019) 011502.
+arXiv:1903.11013, DOI:10.1103/PhysRevD.100.011502.
+[13] C. W. Xiao, J. Nieves, E. Oset, Phys. Rev. D 100 (1) (2019) 014021. arXiv:1904.01296,
+DOI:10.1103/PhysRevD.100.014021.
+[14] J. He, Eur. Phys. J. C 79 (5) (2019) 393. arXiv:1903.11872,
+DOI:10.1140/epjc/s10052-019-6906-1.
+[15] Y.-H. Lin, B.-S. Zou, Phys. Rev. D 100 (5) (2019) 056005. arXiv:1908.05309,
+DOI:10.1103/PhysRevD.100.056005.
+– 20 –
+
+[16] H.-X. Chen, W. Chen, S.-L. Zhu, Phys. Rev. D 100 (5) (2019) 051501. arXiv:1903.11001,
+DOI:10.1103/PhysRevD.100.051501.
+[17] A. Ali, A. Y. Parkhomenko, Phys. Lett. B 793 (2019) 365–371. arXiv:1904.00446,
+DOI:10.1016/j.physletb.2019.05.002.
+[18] L. Meng, B. Wang, G.-J. Wang, S.-L. Zhu, Phys. Rev. D 100 (1) (2019) 014031.
+arXiv:1905.04113, DOI:10.1103/PhysRevD.100.014031.
+[19] T. J. Burns, E. S. Swanson, Phys. Rev. D 100 (11) (2019) 114033. arXiv:1908.03528,
+DOI:10.1103/PhysRevD.100.114033.
+[20] M. B. Voloshin, Phys. Rev. D 100 (3) (2019) 034020. arXiv:1907.01476,
+DOI:10.1103/PhysRevD.100.034020.
+[21] F.-K. Guo, X.-H. Liu, S. Sakai, Prog. Part. Nucl. Phys. 112 (2020) 103757.
+arXiv:1912.07030, DOI:10.1016/j.ppnp.2020.103757.
+[22] H.-W. Ke, M. Li, X.-H. Liu, X.-Q. Li, Phys. Rev. D 101 (1) (2020) 014024.
+arXiv:1909.12509, DOI:10.1103/PhysRevD.101.014024.
+[23] M.-L. Du, V. Baru, F.-K. Guo, C. Hanhart, U.-G. Meißner, J. A. Oller, Q. Wang, Phys. Rev.
+Lett. 124 (7) (2020) 072001. arXiv:1910.11846, DOI:10.1103/PhysRevLett.124.072001.
+[24] Z.-G. Wang, Chin. Phys. C 44 (11) (2020) 113106. arXiv:2006.13028,
+DOI:10.1088/1674-1137/abb080.
+[25] Y. Yamaguchi, H. Garcia-Tecocoatzi, A. Giachino, A. Hosaka, E. Santopinto, S. Takeuchi,
+M. Takizawa, Phys. Rev. D 101 (9) (2020) 091502. arXiv:1907.04684,
+DOI:10.1103/PhysRevD.101.091502.
+[26] H. Xu, Q. Li, C.-H. Chang, G.-L. Wang, Phys. Rev. D 101 (5) (2020) 054037.
+arXiv:2001.02980, DOI:10.1103/PhysRevD.101.054037.
+[27] T. J. Burns, E. S. Swanson, Eur. Phys. J. A 58 (4) (2022) 68. arXiv:2112.11527,
+DOI:10.1140/epja/s10050-022-00723-9.
+[28] C. Fernandez-Ramirez, A. Pilloni, M. Albaladejo, A. Jackura, V. Mathieu, M. Mikhasenko,
+J. A. Silva-Castro, A. P. Szczepaniak, Phys. Rev. Lett. 123 (9) (2019) 092001.
+arXiv:1904.10021, DOI:10.1103/PhysRevLett.123.092001.
+[29] W. Ruangyoo, K. Phumphan, C.-C. Chen, A. Limphirat, Y. Yan, J. Phys. G 49 (7) (2022)
+075001. arXiv:2105.14249, DOI:10.1088/1361-6471/ac58af.
+[30] F. Stancu, Phys. Rev. D 104 (5) (2021) 054050. arXiv:2108.05841,
+DOI:10.1103/PhysRevD.104.054050.
+[31] S. X. Nakamura, Phys. Rev. D 103 (2021) 111503. arXiv:2103.06817,
+DOI:10.1103/PhysRevD.103.L111503.
+[32] S. X. Nakamura, A. Hosaka, Y. Yamaguchi, Phys. Rev. D 104 (9) (2021) L091503.
+arXiv:2109.15235, DOI:10.1103/PhysRevD.104.L091503.
+[33] U. ¨Ozdem, Chin. Phys. C 45 (2) (2021) 023119. DOI:10.1088/1674-1137/abd01c.
+[34] U. ¨Ozdem, Eur. Phys. J. C 81 (4) (2021) 277. arXiv:2102.01996,
+DOI:10.1140/epjc/s10052-021-09070-3.
+[35] Y.-J. Xu, Y.-L. Liu, M.-Q. Huang, Eur. Phys. J. C 81 (5) (2021) 421. arXiv:2008.07937,
+DOI:10.1140/epjc/s10052-021-09211-8.
+– 21 –
+
+[36] S. Sakai, H.-J. Jing, F.-K. Guo, Phys. Rev. D 100 (7) (2019) 074007. arXiv:1907.03414,
+DOI:10.1103/PhysRevD.100.074007.
+[37] Y. Dong, P. Shen, F. Huang, Z. Zhang, Eur. Phys. J. C 80 (4) (2020) 341.
+arXiv:2002.08051, DOI:10.1140/epjc/s10052-020-7890-1.
+[38] G.-J. Wang, L.-Y. Xiao, R. Chen, X.-H. Liu, X. Liu, S.-L. Zhu, Phys. Rev. D 102 (3) (2020)
+036012. arXiv:1911.09613, DOI:10.1103/PhysRevD.102.036012.
+[39] H.-X. Chen, Eur. Phys. J. C 80 (10) (2020) 945. arXiv:2001.09563,
+DOI:10.1140/epjc/s10052-020-08519-1.
+[40] Y.-J. Xu, C.-Y. Cui, Y.-L. Liu, M.-Q. Huang, Phys. Rev. D 102 (3) (2020) 034028.
+arXiv:1907.05097, DOI:10.1103/PhysRevD.102.034028.
+[41] Z.-G. Wang, X. Wang, Chin. Phys. C 44 (2020) 103102. arXiv:1907.04582,
+DOI:10.1088/1674-1137/ababf7.
+[42] R. Aaij, et al., Phys. Rev. D 102 (11) (2020) 112012. arXiv:2007.11292,
+DOI:10.1103/PhysRevD.102.112012.
+[43] Q. Li, C.-H. Chang, S.-X. Qin, G.-L. Wang, Chin. Phys. C 44 (2020) 013102.
+arXiv:1903.02282, DOI:10.1088/1674-1137/44/1/013102.
+[44] Q. Li, C.-H. Chang, S.-X. Qin, G.-L. Wang, Eur. Phys. J. C 82 (2022) 60.
+arXiv:2112.10966, DOI:10.1140/epjc/s10052-022-10006-8.
+[45] Q. Li, C.-H. Chang, G.-L. Wang, T. Wang, Phys. Rev. D 104 (1) (2021) 014018.
+arXiv:2104.12372, DOI:10.1103/PhysRevD.104.014018.
+[46] C.-H. Chang, C. Kim, G.-L. Wang, Phys. Lett. B 623 (2005) 218–226.
+DOI:10.1016/j.physletb.2005.07.059.
+[47] Z.-H. Wang, G.-L. Wang, C.-H. Chang, J. Phys. G: Nucl. Part. Phys. 39 (2012) 015009.
+arXiv:1107.0474, DOI:10.1088/0954-3899/39/1/015009.
+[48] T. Wang, G.-L. Wang, H.-F. Fu, W.-L. Ju, JHEP 07 (2013) 120. arXiv:1305.1067,
+DOI:10.1007/JHEP07(2013)120.
+[49] Q. Li, T. Wang, Y. Jiang, H. Yuan, G.-L. Wang, Eur. Phys. J. C 76 (8) (2016) 454.
+DOI:10.1140/epjc/s10052-016-4306-3.
+[50] Q. Li, T. Wang, Y. Jiang, H. Yuan, T. Zhou, G.-L. Wang, Eur. Phys. J. C 77 (1) (2017) 12.
+DOI:10.1140/epjc/s10052-016-4588-5.
+[51] Q. Li, Y. Jiang, T. Wang, H. Yuan, G.-L. Wang, C.-H. Chang, Eur. Phys. J. C 77 (5) (2017)
+297. arXiv:1701.03252, DOI:10.1140/epjc/s10052-017-4865-y.
+[52] Q. Li, T. Wang, Y. Jiang, G.-L. Wang, C.-H. Chang, Phys. Rev. D 100 (7) (2019) 076020.
+arXiv:1802.06351, DOI:10.1103/PhysRevD.100.076020.
+[53] R. Casalbuoni, A. Deandrea, N. Di Bartolomeo, R. Gatto, F. Feruglio, G. Nardulli, Phys.
+Rept. 281 (1997) 145–238. arXiv:hep-ph/9605342, DOI:10.1016/S0370-1573(96)00027-0.
+[54] T.-M. Yan, H.-Y. Cheng, C.-Y. Cheung, G.-L. Lin, Y. C. Lin, H.-L. Yu, Phys. Rev. D 46
+(1992) 1148–1164. DOI:10.1103/PhysRevD.46.1148.
+[55] H.-Y. Cheng, C.-K. Chua, Phys. Rev. D 75 (2007) 014006. arXiv:hep-ph/0610283,
+DOI:10.1103/PhysRevD.75.014006.
+– 22 –
+
+[56] Y.-R. Liu, M. Oka, Phys. Rev. D 85 (2012) 014015. arXiv:1103.4624,
+DOI:10.1103/PhysRevD.85.014015.
+[57] F.-K. Guo, C. Hanhart, U.-G. Meißner, Q. Wang, Q. Zhao, B.-S. Zou, Rev. Mod. Phys.
+90 (1) (2018) 015004. arXiv:1705.00141, DOI:10.1103/RevModPhys.90.015004.
+[58] M. B. Wise, Phys. Rev. D 45 (7) (1992) R2188. DOI:10.1103/PhysRevD.45.R2188.
+[59] G. Burdman, J. F. Donoghue, Phys. Lett. B 280 (1992) 287–291.
+DOI:10.1016/0370-2693(92)90068-F.
+[60] E. E. Jenkins, M. E. Luke, A. V. Manohar, M. J. Savage, Nucl. Phys. B 390 (1993) 463–473.
+arXiv:hep-ph/9204238, DOI:10.1016/0550-3213(93)90464-Z.
+[61] P. Colangelo, F. De Fazio, T. N. Pham, Phys. Rev. D 69 (2004) 054023.
+arXiv:hep-ph/0310084, DOI:10.1103/PhysRevD.69.054023.
+[62] E. J. Garzon, J.-J. Xie, Phys. Rev. C 92 (3) (2015) 035201. arXiv:1506.06834,
+DOI:10.1103/PhysRevC.92.035201.
+[63] P. A. Zyla, et al., PTEP 2020 (8) (2020) 083C01. DOI:10.1093/ptep/ptaa104.
+[64] R. Mertig, M. Bohm, A. Denner, Comput. Phys. Commun. 64 (1991) 345–359.
+DOI:10.1016/0010-4655(91)90130-D.
+[65] V. Shtabovenko, R. Mertig, F. Orellana, Comput. Phys. Commun. 207 (2016) 432–444.
+arXiv:1601.01167, DOI:10.1016/j.cpc.2016.06.008.
+[66] V. Shtabovenko, R. Mertig, F. Orellana, Comput. Phys. Commun. 256 (2020) 107478.
+arXiv:2001.04407, DOI:10.1016/j.cpc.2020.107478.
+– 23 –
+
diff --git a/nNA0T4oBgHgl3EQfJv9r/content/tmp_files/load_file.txt b/nNA0T4oBgHgl3EQfJv9r/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..e79f0f9da7f148be6e4ce7ffd206050f72f750dd
--- /dev/null
+++ b/nNA0T4oBgHgl3EQfJv9r/content/tmp_files/load_file.txt
@@ -0,0 +1,1465 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf,len=1464
+page_content='Strong decays of P N ψ (4312)+ → J/ψ(ηc)p within the Bethe-Salpeter framework Qiang Li1 Chao-Hsi Chang2,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4 Tianhong Wang5 Guo-Li Wang5,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6 1School of Physical Science and Technology,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Northwestern Polytechnical University,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xi’an 710072,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' China 2CAS Key Laboratory of Theoretical Physics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Institute of Theoretical Physics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chinese Academy of Sciences,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Beijing 100190,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' China 3School of Physical Sciences,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' University of Chinese Academy of Sciences,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Beijing 100049,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' China 4School of Physical Science and Technology,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lanzhou University,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lanzhou 730000,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' China 5School of Physics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Harbin Institute of Technology,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Harbin 150001,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' China 6Department of Physics,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hebei University,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Baoding 071002,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' China E-mail: liruo@nwpu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='cn, zhangzx@itp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='cn, thwang@hit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='cn, gl wang@hit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='edu.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='cn Abstract: Based on the effective Lagrangian in the heavy quark limit, we calculate the one-boson-exchange interaction kernel of ¯DΣc in the isospin-1 2 state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' We present the Bethe- Salpeter equation and wave function for the constituent particles to be a (pseudo)scalar meson and a 1 2 baryon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' By using the Bethe-Salpeter equation, we can obtain P N ψ (4312)+ as the ¯DΣc molecular state with JP = ( 1 2)−.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Combining the effective Lagrangian and the obtained BS wave function, the partial decay widths of P N ψ (4312)+ to J/ψp and ηcp are calculated to be 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11−0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06 MeV and 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='9 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6 × 10−2 MeV, respectively, which are consistent with the LHCb experimental measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our results indicate the fraction of J/ψp channel amounts to ∼ 1% of P N ψ (4312)+, and can reach to ∼ 30% when the interaction kernel is reduced by half.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our results favor the interpretation of P N ψ (4312)+ as the ¯DΣc molecular state with JP = ( 1 2)− and isospin I = 1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='02094v1 [hep-ph] 5 Jan 2023 Contents 1 Introduction 1 2 P N ψ (4312)+ as the ¯DΣc molecular state 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 Bethe-Salpeter equation of a JP = 0− meson and a 1 2 + baryon 3 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 Interaction kernel from the one-boson exchange 5 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 Salpeter wave function for the JP = 1 2 − pentaquark states 7 3 Strong decay of P N ψ (4312)+ → J/ψ(ηc)p within the BS wave function 7 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 Amplitude for P N ψ (4312)+ → J/ψp 9 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 Amplitude with D exchange 9 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 Amplitude with D∗ exchange 11 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 Amplitude for P N ψ (4312)+ → ηcp 12 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 Partial decay width 13 4 Numerical results and discussions 14 A Expressions of the decay form factors 17 1 Introduction In 2019, a narrow pentaquark state Pc(4312)+ is first observed in the J/ψp invariant mass spectrum [1] by the LHCb collaboration, which indicates this state at least to contain five valence quarks, namely, [c¯cuud] quark contents.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' This pentaquark state will be labeled as P N ψ (4312)+ in this work following the new naming scheme proposed by the LHCb collabora- tion [2], where the superscript N denotes the isotopic spin I = 1 2 and the subscript ψ denotes the hidden charm flavor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The measured mass and total width are MP N ψ (4312)+ = 4312 MeV and ΓP N ψ (4312)+ = 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7+3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7 −4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 MeV [1] respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The proximity to the ¯DΣc threshold of the observed narrow peak suggests that they play an important role in the dynamics of P N ψ (4312)+ state, and makes the ¯DΣc molecular state picture a natural interpretation to this exotic particle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The hidden charm molecular pentaquark states have been proposed before the exper- imental confirmation [3–9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' After the LHCb discoveries, lots of literature explored these newly observed pentaquark states from different aspects within different approaches, such as Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [10–27].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Although the properties of the P N ψ (4312)+ are most likely to be the S- wave combination of ¯DΣc with I(JP ) = 1/2(1/2−) [10–16, 26, 27], the contrary view [28], or the possibilities of the compact pentaquark state [29, 30] or kinematical effects [31, 32] still exist.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Though suggested by the LHCb to be labeled as P N ψ (4312)+, the essence of this pentaquark state is still an open question.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 1 – Besides the spectrum or electromagnetic properties [33–35], the strong decay properties play important roles in determining the nature of the pentaquark states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The decay to J/ψp is the discovery channel and also the only detected decay mode of P N ψ (4312)+ so far, and hence this decay channel should be payed more attention to explore the property of P N ψ (4312)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Several approaches are used to study the decay properties of these pentaquark states [10, 15, 30, 36–41], including the effective Lagrangian methods [10, 15], the flavor-spin and heavy quark spin symmetry [30, 36], the chiral constituent quark model [37], QCD sum rules [40, 41], etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Most of the previous studies are based on the nonrelativistic Schrodinger or Lippmann-Schwenger equation and the results are dependent on several introduced free parameters, especially the cutoff value in the form factors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' These undetermined parameters weaken the prediction power of the theories and bring ambiguity in interpreting the nature of P N ψ (4312)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Some researches also suggest the ηcp channel can be an important decay mode of P N ψ (4312)+ [37, 38], especially, the methods by using the heavy quark symmetry predict that the decay ratio of P N ψ (4312)+ to ηcp over J/ψp can reach about three [20, 36].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' However, no experimental evidence is reported in a recent search for pentaquark state P N ψ (4312)+ in Λ0 b → ηcpK− decay channel [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' More studies on the decay behaviors of P N ψ (4312)+ can be important and helpful to explore its inner structure and dynamics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In this work, we will calculate the partial decay widths of P N ψ (4312)+ to J/ψp and ηcp by combing the Bethe-Salpeter (BS) framework with the effective Lagrangian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The Bethe- Salpeter equation(BSE) is a relativistic two-body bound state equation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Another advantage is that the constructed BS wave functions only depend on the good quantum number spin- parity and Lorentz covariance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The BS methods have already been successfully used to cope with mass spectra of the doubly heavy baryons [43, 44], producing the recently observed molecular pentaquarks [26] and the fully heavy tetraquark TQQ ¯Q ¯Q states [45], and also the hadronic transitions and decays[46–52].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The theoretical calculations from BS methods achieve satisfactory consistences with the experimental measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' This paper is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' After the introduction, we start with the Bethe- Salpeter equation for P N ψ (4312)+ as the molecular state of a (pseudo)scalar meson and a baryon, including the interaction kernel and the relevant Salpeter wave function (Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2), then we calculate the strong decay widths of P N ψ (4312)+ → J/ψ(ηc)p (Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' We finally present the numerical results, discussion and summaries in Sect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2 P N ψ (4312)+ as the ¯DΣc molecular state In this part, we will first briefly review Bethe-Salpeter equation of a scalar meson and a baryon under the instantaneous approximation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then we introduce the pentaquark interaction kernel based on the one-boson exchange.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The relativistic BS wave functions of the JP = ( 1 2)− P N ψ state will be introduced and solved numerically to prepare for the next decay calculations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 2 – P N ψ , P P N ψ , P ¯D, p1 Σc, p2 ¯D, k1 Σc, k2 p1 p2 i i′′ i′ i = K Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 1: Bethe-Salpeter equation of the molecular state consisting of the constituent (pseudo)scalar meson and J = 1 2 baryon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The orange letters denote the Dirac indices.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The blue symbols P, p1(k1), p2(k2) denote the momenta of the pentaquark, constituent meson, and the constituent baryon respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 Bethe-Salpeter equation of a JP = 0− meson and a 1 2 + baryon Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 1 schematically depicts the Bethe-Salpeter equation for the bound state consisting of a constituent meson and a constituent baryon, which can be expressed as Γ(P, q, r) = � d4k (2π)4 (−i)K(k, q)[S(k2)Γ(P, k, r)D(k1)], (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1) where Γ(P, q, r) denotes the vertex of the pentaquark, constituent meson and baryon;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' we used P, q, and r to represent the pentaquark total momentum, inner relative momentum, and spin state respectively;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' the inner relative momentum q and k are defined as q ≡ α2p1 − α1p2, k ≡ α2k1 − α1k2, with α1(2) ≡ m1(2) m1+m2 , k1(2) denoting the momentum of the constituent meson (baryon), and m1(2) is the corresponding mass;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S(k2) = i /k2−m2 is the free Dirac propagator of the baryon;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D(k1) = i k2 1−m2 1 denotes the usual scalar propagator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The iϵ should be implied in all the propagators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Since both the two constituent particles, namely ¯D and Σc, contain a heavy charm quark, the relative velocity would be small.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then the interaction kernel K(k, q) is assumed to be instantaneous and is not dependent on the time component of the exchanged momentum (k − q), namely, K(k, q) ∼ K(k⊥ − q⊥), where k⊥ = k − kP ˆP with kP ≡ k · ˆP and ˆP = P M , and M is the pentaquark mass.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The spacelike momentum q⊥ is defined similarly.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Throughout this work, this instantaneous approximation is assumed for the pentaquark kernel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The four-dimensional Bethe-Salpeter wave function is defined as ψ(q) = S(p2)Γ(q)D(p1), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2) where the dependence on P and r omitted for simplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Since the interaction kernel K(k⊥ − q⊥) is instantaneous, the integral over the time component of q can be absorbed into the wave function and it is useful to define the three-dimensional Salpeter wave function as ϕ(q⊥) ≡ −i � dqP 2π ψ(q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3) – 3 – where the factor (−i) is just a convention for later convenience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Performing the contour integral over qP on both sides of Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2), we can obtain the Salpeter equation (SE) for meson-baryon bound state [26], ϕ(q⊥) = 1 2w1 � Λ+(p2⊥) M − w1 − w2 + Λ−(p2⊥) M + w1 + w2 � Γ(q⊥), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4) where wi = � m2 i − p2 i⊥ �1/2 (i = 1, 2) denotes the kinetic energy of the constituent meson and baryon respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The projector operators Λ±(p2⊥) are defined as Λ±(p2⊥) = 1 2 [1 ± H2(p2⊥)] γ0, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5) H2(p2⊥) = 1 w2 (/p2⊥ + m2) γ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6) Notice that H2(p2⊥) is just the corresponding Dirac Hamiltonian divided by the kinetic energy w2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Using the projector operator, we can further define the positive and negative energy wave functions as ϕ± ≡ Λ±γ0ϕ, and we also have ϕ = ϕ+ + ϕ−.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The SE above can be further rewritten as the following type Mϕ = (w1 + w2)H2(p2⊥)ϕ + 1 2w1 γ0Γ(q⊥).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7) where the vertex Γ(q⊥) is now expressed as the integral of the Salpeter wave function, Γ(q⊥) = � d3k⊥ (2π)3 K(k⊥ − q⊥)ϕ(k⊥).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8) The Salpeter Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7) is in fact an eigenvalue equation about the Salpeter wave function ϕα(q⊥), where the pentaquark mass M behaves as the eigenvalue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The three-dimensional BSE, namely, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7), indicates that the mass of the pentaquark state consists of two parts, the kinetic energy and the potential energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The normalization condition of the BS wave function is generally expressed as, − i � � d4q (2π)4 d4k (2π)4 ¯ψ(P, q, ¯r) ∂ ∂P 0 I(P, k, q)ψ(P, k, r) = 2Mδr¯r, where ¯ψ = ψ†γ0 and δr¯r is the Kronecker symbol;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' the integral kernel I in the normalization condition reads, I(P, k, q) = (2π)2δ4(k − q)S−1(p2)D−1(p1) + iK(P, k, q).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Notice in this work, under the instantaneous approximation, the interaction kernel has no dependence on P 0, which indicates the normalization would only involve the inverses of the two propagators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Performing the contour integral, the normalization condition can be – 4 – further expressed by the Salpeter wave function as � d3q⊥ (2π)3 (2w1) ¯ϕ(q⊥, ¯r)γ0ϕ(q⊥, r) = 2Mδr¯r, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='9) where ¯ϕ = ϕ†γ0 and the symbol ¯r just denotes the spin state;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' also notice both the BS wave function ψ(q) and the Salpeter wave function ϕ(q⊥) are four-component spinor.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 Interaction kernel from the one-boson exchange The P N ψ (4312)+ are consistent with the ¯DΣc molecular state with isospin I = 1 2 and I3 = + 1 2, which can be expressed in the uncoupled representation as | 1 2, 1 2⟩ = √ 2 √ 3 |1, +1⟩ | 1 2, − 1 2⟩ − 1 √ 3 |1, 0⟩ | 1 2, 1 2⟩ = √ 2 √ 3 |Σ++ c ⟩ |D−⟩ − 1 √ 3 |Σ+ c ⟩ | ¯D0⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='10) In the molecular state scenario of P N ψ (4312)+, the interaction kernel between the two constituents Σc and ¯D can be realized by the one-boson exchange.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Notice the usual one- pion exchange is not possible in the ¯DΣc bound state for the parity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' We only need to consider the light scalar and vector meson exchange.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Considering the heavy quark spin-flavor symmetry, hidden local symmetry and the light quark chiral symmetry, the involved Lagrangian describing the charmed anti-heavy- light meson and a light scalar and vector meson reads [6, 53] LHHV = −ρV1 ⟨ ¯H¯cvαVαH¯c⟩ − ρT1 ⟨ ¯H¯cσαβ(∂αVβ − ∂βVα)H¯c⟩ + σ1 ⟨ ¯H¯cσH¯c⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11) Here ⟨ ⟩ denotes taking the Dirac trace, and σ denotes field of the light scalar meson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' ρV 1, ρT1, and σ1 denote the corresponding coupling constants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' H¯c represents the field of the ( ¯D, ¯D∗) doublet in the heavy quark limit, H¯c = � ¯D∗µγµ + i ¯Dγ5 � 1 − /v 2 , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12) where ¯D = ( ¯D0, D−, D− s ) denotes anti-charmed heavy-light meson fields in flavor triplet, and ¯D∗µ is the corresponding vector state;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' ¯H¯c = γ0H† ¯cγ0 is the usual conjunction in Dirac space;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' and v denotes four-velocity of the heavy-light meson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The symbol V denotes the 3 × 3 matrix consisting of the 9 light vector meson fields [6, 53] V = � ��� (ρ0+ω) √ 2 ρ+ K∗+ ρ− − (ρ0−ω) √ 2 K∗0 K∗− ¯K∗0 φ � ��� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='13) Considering the heavy quark symmetry, hidden local symmetry and chiral symmetry, the effective Lagrangian of the heavy-light baryon and light mesons reads [6, 54–56] LB6B6V = ρV2 ⟨¯SµvαVαSµ⟩ + iρT2 ⟨¯Sµ(∂µVν − ∂νVµ)Sν⟩ + σ2 ⟨¯SµσSµ⟩ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='14) – 5 – Here ⟨ ⟩ denotes taking trace in the 3 × 3 flavor space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The baryon spin doublet are incorporated in field Sµ = − 1 √ 3(γµ + vµ)γ5B6 + B∗ 6µ, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='15) where the systematic baryon sextet B6 in 3 × 3 matrix reads B6 = � ��� Σ++ c 1 √ 2Σ+ c 1 √ 2Ξ′+ c 1 √ 2Σ+ c Σ0 c 1 √ 2Ξ′0 c 1 √ 2Ξ′+ c 1 √ 2Ξ′0 c Ω0 c � ��� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='16) The conjugation defines as usual for the spinor field ¯Smn µ = (Smn µ )†γ0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' An asterisk on the symbol denotes the corresponding spin-3 2 baryon, which is not involved in this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Using above relevant Lagrangian and based on the one-boson exchange, we calculate the interaction kernel of ¯DΣc in isospin-1 2 as K(s⊥) = F 2(s2 ⊥) � V1 + V2 /s⊥ |⃗s | � , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='17) where F(s2 ⊥) denotes the regulator in the heavy hadron ( ¯D or Σc here) vertex;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' and the potential V1 and V2 is specifically expressed as, V1 = −2σ1σ2MD 1 E2σ + ρV1ρV2MD � 1 E2ρ − 1 2E2ω � , V2 = −1 3ρV1ρT2MD|s| � 2 E2ρ − 1 E2ω � , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='18) where Eρ = (s2 + m2 ρ)1/2 denotes the energy of the inter-mediator ρ meson, and similar for Eσ and Eω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The influence of the potential strength on the decay widths will be discussed later.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' There is no general method to choose the regulator functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In this work, we use the following propagator-type form factor, namely, F(s2) = m2 Λ s2 + m2 Λ , (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='19) where mΛ is the introduced cutoff parameter to characterize the regulator function.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Notice mΛ is the only free parameter in this analysis and can be determined by fitting bound state mass to the experimental data, which is found to be mΛ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='25 GeV for P N ψ (4312)+ and close to the mass scale of the exchanged particle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In the limit s2 → 0, the heavy hadron is seen by the inter-mediator mesons as a point-like particle, and hence the form factor is normalized to 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The cutoff value mΛ is usually believed to be much larger than the typical energy scale √2µϵ ∼ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 GeV for P N ψ (4312)+ [15, 57], where µ = m1m2 m1+m2 is the reduced mass of the two-hadron system and ϵ = (m1 +m2 −M) denotes the bound energy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 6 – Our determined cutoff value is consistent with this universal estimation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The obtained V1 and V2 for isospin-1 2 are displayed graphically in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3(a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 Salpeter wave function for the JP = 1 2 − pentaquark states According to the spin-parity properties, and also considering the proper Lorentz structures, the Salpeter wave function of JP = 1 2 − pentaquarks consisting of a 0− meson and 1 2 + baryon can be generally constructed as ϕ(P, q⊥, r) = � f1 + f2 /q⊥ q � γ5u(P, r), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='20) where the radial wave function f1(2)(|⃗q |) only explicitly depend on |⃗q |;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' u(P, r) denotes Dirac spinor with spin state r.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In terms of the spherical harmonics Y m l , the wave function can be rewritten as ϕ(P, q⊥, r) = 2√π � f1Y 0 0 + 1 √ 3f2 � Y 1 1 γ− + Y −1 1 γ+ − Y 0 1 γ3�� γ5u(P, r), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='21) where γ± = ∓ 1 √ 2(γ1 ± iγ2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then it is obvious to see that f1 and f2 represent the S- and P-wave components, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Inserting the wave function into Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='22), we obtain the normalization satisfied by the radial wave functions as � d3q⊥ (2π)3 2w1 � f2 1 + f2 2 � = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='22) Inserting the Salpeter wave function Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='20) into the Salpeter equation (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7), elim- inating the spinor, calculating the trace, we can obtain two coupled eigenvalue equations with the pentaquark mass M as the eigenvalue and f1(2) as the eigen wave functions (see Ref.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [26, 43, 44] for details).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Solving the eigenvalue equations numerically, we can obtain the corresponding mass spectra and numerical wave functions, which are also graphically displayed in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3(b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3 Strong decay of P N ψ (4312)+ → J/ψ(ηc)p within the BS wave function In this section, we first present the relevant effective Lagrangian;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' then we give the decay amplitude by using the BS wave function combining with the effective Lagrangian;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' finally, the expressions of the partial decay widths are presented in terms of the relevant form factors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' For P N ψ (4312)+ → J/ψ(ηc)p, the involved interactions are J/ψDD(∗), ηcDD∗, and ΣcND(∗), which involve the Lagrangian of the doubly heavy meson and the heavy-light meson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The heavy-light charmed mesons in S-wave can be represented by [53, 58, 59] Hc = 1 + /v 2 (D∗µγµ + iDγ5), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1) where D∗µ and D denote the corresponding vector and pseudoscalar charmed D mesons respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The anti-heavy-light meson doublet H¯c has been presented in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 7 – For doubly heavy mesons, the heavy quark flavor symmetry does not hold any longer, while the heavy quark spin symmetry still holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In the ground states, the charmonium forms a doublet consisting of a pseudoscalar ηc and a vector state J/ψ, which can be represented by [60] R = 1 + /v 2 (ψµγµ + iηcγ5)1 − /v 2 , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2) where ψµ and ηc denotes the fields of the corresponding mesons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Here all the hadron fields in above equations contain a factor of √MH with MH the corresponding meson mass.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' By assuming the invariance under independent rotations of the heavy quark spins, it is possible to write down the effective coupling between the S-wave charmonia and the heavy-light mesons as [61] L2 = g2Tr � R ¯H¯c ←→ /∂ ¯Hc � + H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3) which is invariant under independent heavy quark spin symmetry;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' and the notation A←→ ∂ B ≡ A∂B −∂AB is used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Consequently, we obtain the following effective Lagrangian describing J/ψ and ηc coupling to the DD∗, L2 = + gψDDψ†µ ¯D←→ ∂µD − igψDD∗ 1 Mψ ϵµναβ∂µψ† ν( ¯D←→ ∂αD∗ β + ¯D∗ α ←→ ∂β D) + gψD∗D∗ψ†µ( ¯D∗ν←→ ∂ν D∗ µ + ¯D∗ µ ←→ ∂ν D∗ν − ¯D∗ν←→ ∂µD∗ ν) + gDD∗ηcη† c(∂µ ¯DD∗µ − ¯D∗µ∂µD) + igD∗D∗ηc 1 Mηc ϵµναβ∂µη† c ¯D∗ ν ←→ ∂αD∗ β + H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4) where we have divide a meson mass in the second and the last Lagrangians to keep all the coupling constants dimensionless.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The symbol ϵµναβ denotes the totally antisymmetric Levi-Civita tensor with ϵµναβ = −ϵµναβ and convention ϵ0123 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' All these coupling con- stants are related to a single coupling g2, which is determined to be g2 = �Mψ/(2MDfψ) with fψ denoting the J/ψ decay constant [61].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then all other coupling constants can also be expressed in terms of the gψDD as gψDD = Mψ fψ , gψDD∗ = � MD∗ MD �1/2 gψDD, gψD∗D∗ = � MD∗ MD � gψDD, gDD∗ηc = � MηcMD∗ MψMD �1/2 gψDD, gD∗D∗ηc = � Mηc Mψ �1/2 MD∗ MD gψDD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5) – 8 – In next section, we will also discuss the effects of these coupling constants on the final decay widths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 Amplitude for P N ψ (4312)+ → J/ψp P N ψ (4312)+ as the ¯DΣc molecular state can decay to J/ψp by exchanging either a D or a D∗ virtual meson, and the total amplitude is the sum of the two.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Pc, P ¯D, k1 Σc, k2 J/ψ, p1 p, p2 D, k3 Pc, P ¯D, k1 Σc, k2 J/ψ, p1 p, p2 µ ν D∗, k3 Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2: Strong decay of P N ψ (4312)+ to the J/ψp by exchanging a virtual mediator D (left panel) and D∗ (right panel).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' P, k1, k2, P1, P2 denote the momenta of P N ψ , constituent meson, constituent baryon, the final J/ψ, and the final p respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 Amplitude with D exchange The left panel of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2 shows the Feynman diagram of P N ψ (4312)+ → J/ψp by exchanging the D meson.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Besides the pentaquark vertex, we also need two other effective Lagrangian to obtain the decay width.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' From above results, the effective Lagrangian describing the DDJ/ψ interaction read LψDD = gψDDψ†µ( ¯D∂µD − ∂µ ¯DD), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6) where D = (D0, D+, D+ s )T represents the charmed meson fields in flavor triplet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Whereas the effective Lagrangian for NDΣc interaction behaves as [10, 62] LNDΣc = −igNDΣc ¯Nγ5Σc ¯D† + H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7) where N stands for nucleon field doublet;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Σc = σ · Σc with σ denoting the Pauli matrix and Σc denoting the Σc baryon isospin triplet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The invariant amplitude for P N ψ (4312)+ → J/ψp by exchanging a D can then be expressed by the Bethe-Salpeter vertex as A1 = � d4k (2π)4 ¯u2(−igNDΣc)γ5 [S(k2)Γ(P, k, r)D(k1)] (gψDD)D(k3)(i)e∗ 1 · (k3 + k1), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8) where u2 is short for u(r2)(P2) with r2 representing the proton spin state;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' e1 is short for e(r1)(P1) representing the polarization vector of the final J/ψ with P1 denoting the J/ψ – 9 – momentum and r1 = 0, ±1 representing the 3 possible polarization states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The polarization vector e1 fulfills the Lorentz condition eα 1 P1α = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='9) The momentum of the exchanged virtual charmed meson is denoted as k3 = (k1 − P1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' We will use M1 and M2 to denote the masses of the final J/ψ and proton respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Also notice k1 = (α1P + k) and k3 are involved in the four-dimensional integration over k.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' To simplify this amplitude, first we strip off the triangle amplitude involved the integral over k as A1T;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='αu(P, r) = γ5 � d4k (2π)4 [S(k2)Γ(P, k, r)D(k1)] D(k3)(2k1α), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='10) where the Lorentz condition of the vector meson is utilized, and we also strip off the spinor u(P, r) for later convenience.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The decay amplitude A can then be simplified as A1 = (gNDΣcgψDD)e∗α 1 ¯u2A1T;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='αu(P, r).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11) Then we perform the contour integral over kP on Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='10), and obtain Aα 1T u(P, r) = γ5 � dk3 ⊥ (2π)3 1 w3 (aα 1 ϕ+ + aα 2 ϕ−), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12) where we used the expression of the positive(negative) energy wave functions ϕ± = Λ±γ0ϕ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' the two coefficients a1 and a2 behaves as aα 1 = c1xα 1 + c2xα 2 + c3xα 3 , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='13) aα 2 = c4xα 4 + c5xα 5 + c6xα 6 , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='14) where xi = k1(kP = kPi) with (i = 1, · · · 6), and kPis are defined as kP1 = ζ+ 1 , kP2 = ζ+ 2 , kP3 = ζ+ 3 , kP4 = ζ− 1 , kP5 = ζ− 2 , kP6 = ζ− 3 , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='15) where the abbreviations ζ± 1 ≡ −(α1M ∓w1), ζ± 2 ≡ (α2M ∓w2), and ζ± 3 ≡ (E1 −α1M ±w3) are used.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The coefficients cis (i = 1, · · · 6) are defined as c1(4) = 1 w1 + w3 ∓ E1 , c2(5) = (−1) w2 + w3 ∓ E2 , c3(6) = (w1 + w2 ∓ M) (w1 + w3 ± E1)(w2 + w3 ∓ E2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='16) Now the amplitude A1T;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='α has been expressed by the three-dimensional Salpeter wave func- – 10 – tion ϕ(k⊥), and can be further simplified as the two form factors A1T;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='α = (s11γα + s12 ˆPα), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='17) Inserting the obtained P N ψ (4312)+ wave function, namely, Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='20), into Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12) and then calculating the three-dimensional integral numerically, we can obtain the amplitude Aα 1T in terms of s11 and s12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In the appendix A we collect the specific expressions of the two form factors in terms of the Salpeter wave functions f1 and f2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The amplitude A1 then behaves as A1 = (gNDΣcgψDD)e∗α 1 ¯u2(s11γα + s12 ˆPα)u(P, r).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='18) 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 Amplitude with D∗ exchange For P N ψ (4312)+ decaying by exchanging D∗ in the lowest level, the relevant Feynman diagram is displayed in the right panel of Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2, and the two involved interaction vertexes are DD∗J/ψ and ΣcD∗p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The DD∗J/ψ interaction is represented by the following effective Lagrangian LψDD∗ = (−i)gψDD∗ϵµναβ 1 Mψ ∂µψ† ν( ¯D∂αD∗ β − ∂α ¯DD∗ β), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='19) Notice here the coupling constant gψDD∗ is defined to have the same dimension with the pseudoscalar coupling constant gψDD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The effective Lagrangian of ND∗Σc reads [62] LND∗Σc = gND∗Σc ¯NγαΣcD∗† α + H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='. (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='20) All the coupling constants in these effective Lagrangian will be specified in next section.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The invariant amplitude for P N ψ (4312)+ → J/ψp by exchanging a D∗ can be expressed by the Bethe-Salpeter vertex as A2 = (gΣcND∗gψDD∗)¯u2γν � d4k (2π)4 [S(k2)Γ(P, k, r)D(k1)] Dµν(k3) 1 M1 ϵP1αβµe∗ 1α(k3 + k1)β, (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='21) where the propagator of the exchanged D∗ meson behaves as Dµν(k3) = i−gµν + k3µk3ν/m2 3 k2 3 − m2 3 + iϵ , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='22) Here the propagator mass m3 is MD∗.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Notice the contraction with Levi-Civita tensor forces the momentum part in numerator of Dµν(k3) to be zero.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The amplitude can be further simplified as A2 = (gΣcND∗gψDD∗)e∗α 1 ¯u2γµ 1 M1 ϵP1αβµAβ 2T u(P, r), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='23) – 11 – where we have stripped off the amplitude involved the integral over k as before Aβ 2T u(P, r) = − � d4k (2π)4 [S(k2)Γ(P, k, r)D(k1)] D(k3)(2kβ 1 ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='24) In order to express the amplitude by the three-dimensional Salpeter wave function, we perform the contour integration over kP on Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='24) as usual and obtain Aβ 2T u(P, r) = − � d3k⊥ (2π)3 1 w3 � aβ 1ϕ+ + aβ 2ϕ−� .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='25) Combining Aβ 2T u(P, r) with γµ 1 M1 ϵP1αβµ, and using the following identity of the Levi-Civita symbol, iγµϵµαβν = γ5(γαγβγν − γαgβν + γβgαν − γνgαβ), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='26) we can express the decay amplitude A2 for D∗ exchange by the following form factors A2 = (gψDD∗gΣcND∗)e∗α 1 ¯u2(s21γα + s22 ˆPα)u(P, r).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='27) Namely, the amplitude A2 can be expressed by the same form as A1, which is just what it should be.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In above equations, the specific expressions of s21 and s22 can be obtained by in- serting the Salpeter wave functions into Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='25) and performing the integral numerically.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The specific expressions are presented in the appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 Amplitude for P N ψ (4312)+ → ηcp Since ηc with JP = 0−, the decay to ηcp can only happen by exchanging a D∗ while the mode of exchanging a D is forbidden.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' From Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4), the effective Lagrangian responsible for DD∗ηc interaction reads LDD∗ηc = gDD∗ηcη† c∂µ ¯DD∗µ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='28) The effective Lagrangian describing ND∗Σc interaction has been presented in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='20).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The corresponding Feynman diagram is similar with that for the decay to J/ψp with D∗ exchange.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The decay amplitude for P N ψ (4312)+ → ηcp behaves as A = gND∗Σc ¯u2γα � d4k (2π)4 [S(k2)Γ(k, r)D(k1)]Dαβ(k3)gD∗Dηc(−iP β 1 ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='29) As usual, it is convenient to strip off the part involved the integral over k as AT u(P, r) = � d4k (2π)4 (γαP β 1 )[S(k2)Γ(k, r)D(k1)]Dαβ(k3), – 12 – Performing the contour integral over kP , we can express AT by the three-dimensional Salpeter wave function AT u(P, r) = � d3k⊥ (2π)3 1 2w3 (γαP β 1 ) 3 � i=1 � cidαβ(yi)ϕ+ + ci+3dαβ(yi+3)ϕ−� , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='30) where the positive (negative) energy wave function is related to the Salpeter wave function by ϕ± = Λ±γ0ϕ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' and we define dαβ and yi as dαβ(yi) = −gαβ + yiαyiβ m2 3 , yi = k3(kP = kPi) = xiP ˆP + k⊥ − P1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='31) Inserting the Salpeter wave function Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='20) of P N ψ (4312)+, we obtain AT expressed by one form factor, AT = s0γ5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='32) Finally, we obtain the amplitude for decay to ηcp by form factor s0 with a simple form A = −i (gND∗ΣcgηcDD∗) s0[¯u2γ5u(P, r)].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='33) The expression of s0 is also listed in appendix A as the integral over Salpeter wave functions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 Partial decay width Combing the two amplitudes from D and D∗ mediators together, we obtain the full invari- ant amplitude for P N ψ (4312)+ → J/ψp decay by two form factors, A = A1 + A2 = e∗α 1 ¯u2 � s1γα + s2 ˆPα � u(P, r).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='34) where s1 and s2 are related to the coupling constants and are expressed as s1 = gψDDgNDΣcs11 + gψDD∗gND∗Σcs21, s2 = gψDDgNDΣcs12 + gψDD∗gND∗Σcs22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='35) Squaring the amplitude, summing all the polarization states, we obtain � r1,r2,r |A|2 = � −gαβ + P α 1 P β 1 M2 1 � Tr �/P2 + M2 � AT;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='α �/P + M � ¯ AT;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='β, (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='36) where AT;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='α = � s1γα + s2 ˆPα � , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='37) – 13 – and ¯ AT;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='β = γ0A† T;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='βγ0 is defined as the usual conjugation variable;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' we also used the rela- tionship of the summation over the vector polarization states r1, � r1 eα (r1)eβ (r1) = −gαβ + P α 1 P β 1 M2 1 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='38) and the summation over the polarization states of the spinors � r2 u(r2)(P2)¯u(r2)(P2) = (/P2 + M2), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='39) � r u(P, r)¯u(P, r) = (/P + M).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='40) For P N ψ (4312)+ → ηcp decay, the squared amplitude behaves as � r2,r |A|2 = Tr �/P2 + M2 � A �/P + M � ¯ A = 4(gND∗ΣcgηcD∗D)2s2 0M(E2 − M2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='41) Namely, the squared amplitude is proportional to the kinetic energy (E2 − M2) of the proton.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Finally, the partial decay width of P N ψ (4312)+ → J/ψ(ηc)p is expressed as Γ[P N ψ (4312)+ →J/ψ(ηc)p] = |P1| 8πM2 1 2 � r,r1,r2 |A|2, (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='42) where the three momentum of J/ψ(ηc) is given by |P1| = 1 2M �� M2 − (M1 + M2)2�� M2 − (M1 − M2)2�� 1 2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='43) 4 Numerical results and discussions Before giving the decay widths, we first specify the effective interaction coefficients we used in above effective Lagrangian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The interaction coefficients between the heavy hadron and the light bosons read [6, 9, 12, 14, 26]: ρV1 = βgV √ 2 = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='75, ρT1 = λgV √ 2 = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='34 GeV−1, and σ1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='76;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' ρV2 = βSgV √ 2 = 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='26, ρT2 = λSgV √ 2 = 13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='81 GeV−1, and σ2 = 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In the heavy quark limit, the coupling constants between the heavy hadrons read [61] gDDψ = Mψ fψ and gDD∗ψ = ( MD∗ MD )1/2gDDψ with the J/ψ decay constant fψ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='416 GeV estimated from the dilepton decay width [63];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' the DD∗ηc coupling constant reads gDD∗ηc = ( MηcMD∗ MψMD )1/2gDDψ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Combined with the total amplitude Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='34), it can be found that the partial decay width is proportional to 1 f2 ψ .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The coupling constants related to the baryons used are gNDΣc = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='69 and gND∗Σc = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='0 [10, 62].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' These values are the standard parameters used in this work, and we will also vary the standard parameters to explore their influence on the wave functions and the final decay widths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 14 – GeV s 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 3 1 GeV 2 F iV 25 − 20 − 15 − 10 − 5 − 0 1 V 2 V (a) GeV q 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4 2 BS radial wave function GeV 0 10 20 30 40 1f 2f (b) GeV q 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 2 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 3 2 BS radial wave function GeV 0 1 2 3 4 ) n V (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 1f ) n V (0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 2f (c) GeV q 0 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8 1 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4 2 BS radial wave function GeV 0 10 20 30 40 ) n V (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 1f ) n V (1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 2f (d) Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3: The figure (a) shows the isospin-1 2 potentials VnF 2 (n = 1, 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Figure (b) is the Bethe-Salpeter radial wave function f1 and f2 for P N ψ (4312)+ as the ¯DΣc molecular state based on the one-boson exchange;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' while (c) and (d) are the radial wave functions when the interaction potential Vn in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='18) is reduced and increased by 50% based, respectively, where the corresponding regulator values are mΛ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='288 GeV and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='73 GeV respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The only free parameter in this work is the regulator mΛ in the form factor F(s2) in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='19).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' All the other parameters have been determined by the previous experimental data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' By solving the relevant BS eigenvalue equation, we find proper cutoff values of mΛ can produce bound state of ¯DΣc based on the one-boson exchange kernel in isospin-1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then by fitting the bound state mass M to the experimental measurement MP N ψ (4312)+ = 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='312 GeV, we fix mΛ in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='19) to be 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='87 GeV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then the obtained V1 and V2 in the interaction kernel are graphically shown in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3(a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3(b) we show the obtained BS wave functions f1 and f2 for P N ψ (4312)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' On the other hand, the obtained radial wave functions depend on the obtained potential V1(2), which is directly related to the coupling constants σ1(2), ρV1(2) and ρT2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' To reflect the influence of these parameters on the wave functions and decay widths, we vary the nu- merical values of V1(2) under standard parameters by ∓50%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Under these variations, the – 15 – obtained regulator values are then mΛ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='288 GeV and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='73 GeV respectively, and the corresponding wave functions obtained are displayed in Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3(c) and Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 3(d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' As V1(2) decreases, the fitted regulator parameter mΛ increases, and also the role of wave function f1 becomes more important.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our results of the mass spectra for I(JP ) = 1/2(1/2)− ¯DΣc molecule indicate that there only exists one bound state, namely, P N ψ (4312)+ as the ground state of ¯DΣc molecule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our results do not support any radially excited states.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' This conclusion is robust even under the ±50% change of the interaction kernel V1(2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The obtained numerical values of the form factors in amplitudes are s0 = 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 × 10−3 for decay to ηcp channel in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='33);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' and form factors for decay to J/ψp channel in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='34) are s11 = 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 × 10−3, s12 = 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7 × 10−3, s21 = −8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2 × 10−3, s22 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 × 10−2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Inserting above form factors into the decay width expressions, we obtain the partial decay widths as Γ(P N ψ (4312)+ →J/ψp) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11 MeV and Γ(P N ψ (4312)+ →ηcp) = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='056 MeV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The obtained partial width for decay to J/ψp is ∼ 1% of the total width of Γ = 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8 ± 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7+3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7 −4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 MeV [1] reported by the LHCb collaboration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The J/ψp channel is also the only observed decay mode of P N ψ (4312)+ currently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' While the decay fraction of P N ψ (4312)+ to ηcp is about 50% smaller than the J/ψp channel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' There is still no evident signal in recent experimental search of P N ψ (4312)+ in ηcp channel [42].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Notice the obtained results are totally predictive and there are no any free adjustable parameters since the regulator mΛ has been fixed by the mass of P N ψ (4312)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' When the interaction kernel V1(2) varies by ±50% based on the standard parameters, the obtained decay widths are 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='89 MeV and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='025 MeV for J/ψp channel, respectively;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' while for ηcp channel, the results are 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='012 MeV and 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='0024 MeV, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Namely, our results indicate that the fraction of J/ψp decay channel can amount to ∼ 30% when the strength of V1(2) is reduced by half.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' I: Comparison of partial decay width of P N ψ (4312)+ → J/ψ(ηc)p with other works in units of MeV, where our theoretical uncertainties are induced by varying the relevant coupling constants by ±10% in the effective Lagrangian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our results also indicate that the decay width of J/ψp channel can amount to 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='89 MeV when the strength of V1(2) is reduced by half.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Channel This [38] [40] [15] [37] [10] [24] J/ψp 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11−0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='32 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='08 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='67+0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='92 −0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='56 10−3 ∼ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='033 (3 ∼ 8) 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3+19.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5 −9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 ηcp 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='056−0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='019 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='026 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='98 ± 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='25 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='54+0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='75 −0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='50 10−2 ∼ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='066 − 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='26−0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='24 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='55 A comparison of our results with other works is listed in Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our obtained partial decay widths are roughly consistent with those in Refs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [15, 37, 38].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Notice the theoretical results for decay widths of P N ψ (4312)→J/ψ(ηc)p are quite different from each other for the – 16 – complication of this problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' More researches are needed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Since the obtained partial decay widths are also directly dependent on the coupling constants gψDD, gψDD∗, gNDΣc, gND∗Σc, and gDD∗ηc in the relevant effective Lagrangian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' To see the sensitivity of the our partial decay widths on these parameters, we calculate the theoretical uncertainties by varying the every coupling constant by ±10%, and then searching the parameter space to find the maximum deviation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The obtained theoretical errors are also listed in above Tab.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' I, where the width uncertainties induced from the coupling constants amount to about ∼ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='05 MeV for channel J/ψp, and ∼ 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='02 MeV for channel ηcp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Finally, we give a brief summary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' In this work, firstly, based on the effective Lagrangian in the the heavy quark limit, we calculate the one-boson-exchange interaction kernel of ¯DΣc in the isospin-1 2 state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Then by using the Bethe-Salpeter equation, we obtain the mass spectrum and wave functions of the experimental P N ψ (4312)+ as the ¯DΣc molecular state with JP = ( 1 2)−.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Finally combining the effective Lagrangian and the obtained BS wave function, we calculate the partial decay width to be 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11−0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04 +0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06 MeV for P N ψ (4312)+ → J/ψ, and 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='9 +2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6 × 10−2 MeV for ηcp channel.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The obtained results indicate that the fraction of P N ψ (4312)+ → J/ψp amounts to ∼ 1%, and can even reach to ∼ 30% when the interaction kernel is reduced by half.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Our results are roughly consistent with some other calculations and the LHCb experimental measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' However, more theoretical analysis and experimental measurements are necessary to determine the properties of the pentaquark state P N ψ (4312)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The interpretation of P N ψ (4312)+ as the ¯DΣc molecular state with JP = ( 1 2)− and isospin I = 1 2 is favored by this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' A Expressions of the decay form factors For completeness, we list the specific expressions of the relevant form factors here, which are all represented by the integral over the radial Salpeter wave functions f1 and f2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Parts of following expressions are calculated with the help of the FeynCalc package [64–66].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' The four form factors for decay P N ψ (4312)+ →J/ψp in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='35) are s11 = − � d3k⊥ (2π)3 1 w2 c22k [kv0f1 + (w2u0 + m2v0) f2] , s12 = � d3k⊥ (2π)3 1 w2P 2 1 (Y1f1 + Y2f2) , s21 = − � d3k⊥ (2π)3 1 w2P 2 1 (Z1f1 + Z2f2) , s22 = − � d3k⊥ (2π)3 1 w2P 2 1 (Z3f1 + Z4f2) , (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1) where Y1, Y2, and Z1 ∼ Z4 read Y1 =P 2 1 w2u1 + c1E1kP1v1 + c1kM2P1v1 − c1kMP1v1 − m2P 2 1 v1 − c1E1kP1w2u0 − c21E2 1k2v0 − c21E1k2M2v0 + c21E1k2Mv0 + c1E1km2P1v0 + c22k2P 2 1 v0, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2) – 17 – Y2 =c1E1P1w2u1 − c1MP1w2u1 + c1M2P1w2u1 − c1E1m2P1v1 − 2c1E2m2P1v1 + c1m2M2P1v1 + c1m2MP1v1 + kP 2 1 v1 − c21E2 1kw2u0 + c21E1kMw2u0 − c21E1kM2w2u0 + c22kP 2 1 w2u0 + c21E2 1km2v0 + 2c21E1E2km2v0 − c1E1k2P1v0 − c21E1km2M2v0 − c21E1km2Mv0 + c22km2P 2 1 v0, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3) Z1 = − E1P 2 1 w2u1 + MP 2 1 w2u1 + M2P 2 1 w2u1 − c1E2 1kP1v1 + c1kM2 1 P1v1 + E1m2P 2 1 v1 − m2MP 2 1 v1 − m2M2P 2 1 v1 + c1E2 1kP1w2u0 − c1E1kMP1w2u0 − c1E1kM2P1w2u0 + c21E3 1k2v0 − c1E2 1km2P1v0 − c22E1k2P 2 1 v0 − c21E1k2M2 1 v0 + c1E1km2MP1v0 + c1E1km2M2P1v0 − c22k2MP 2 1 v0 − c22k2M2P 2 1 v0, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4) Z2 =c1M2 1 P1w2u1 + c1E2 1m2P1v1 + 2c1E1E2m2P1v1 − 2c1E1m2MP1v1 + c1m2M2 1 P1v1 + kMP 2 1 v1 + kM2P 2 1 v1 + c21E3 1kw2u0 + c1E2 1k2P1v0 − c21E1kM2 1 w2u0 − c22E1kP 2 1 w2u0 − c22kMP 2 1 w2u0 − c22kM2P 2 1 w2u0 + 2c21E2 1km2Mv0 − c1E1k2MP1v0 − c1E1k2M2P1v0 + c22E1km2P 2 1 v0 − c21E1km2M2 1 v0 + 2c22E2km2P 2 1 v0 − 3c22km2MP 2 1 v0 − c22km2M2P 2 1 v0 − 2c21E2 1E2km2v0 − c21E3 1km2v0 − E1kP 2 1 v1 − c1E2 1P1w2u1, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='5) Z3 = − MP 2 1 w2u1 − M2P 2 1 w2u1 + c1E1kMP1v1 − c1E1kM2P1v1 − c1kM2 1 P1v1 + m2M2P 2 1 v1 + c1E1kMP1w2u0 + c1E1kM2P1w2u0 − c21E2 1k2Mv0 + c21E2 1k2M2v0 + c21E1k2M2 1 v0 − c1E1km2MP1v0 − c1E1km2M2P1v0 + m2MP 2 1 v1 − c22k2M2P 2 1 v0 + 3c22k2MP 2 1 v0, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='6) Z4 =c1E1MP1w2u1 − c1E1M2P1w2u1 − c1M2 1 P1w2u1 + c1E1m2MP1v1 − c1m2M2 1 P1v1 − kMP 2 1 v1 − kM2P 2 1 v1 − c21E2 1kMw2u0 + c21E2 1kM2w2u0 + c21E1kM2 1 w2u0 + 3c22kMP 2 1 w2u0 − c22kM2P 2 1 w2u0 − c21E2 1km2Mv0 + c21E2 1km2M2v0 + c1E1k2MP1v0 + c1E1k2M2P1v0 + c21E1km2M2 1 v0 + 3c22km2MP 2 1 v0 − c22km2M2P 2 1 v0 − c1E1m2M2P1v1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='7) In above expressions, P1 = |P1|, and c = cos θ, c21 = 1 2(3 cos2 θ − 1), c22 = 1 2(cos2 θ − 1), (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='8) where θ denotes the angle between k and P1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' We also define un and vn (n = 0, 1, 2) for later convenience un = (c1xn 1P + c2xn 2P + c3xn 3P ) + (c4xn 4P + c5xn 5P + c6xn 6P ), vn = (c1xn 1P + c2xn 2P + c3xn 3P ) − (c4xn 4P + c5xn 5P + c6xn 6P ).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='9) The expressions of ci are listed in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='16).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 18 – The form factor s0 in Eq.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='33) for P N ψ (4312)+ → ηcp decay behaves s0 = � d3k⊥ (2π)3 1 4m2 3w2w3P 2 1 � (P 2 1 X1 + kc1X3)f1 + (kP 2 1 X2 + c1X4)f2 � , (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='10) where X1 ∼ X4 read X1 = − ckMP1w2u0 − ckM2P1w2u0 + m2 3Mw2u0 + m2 3M2w2u0 − MM2 1 w2u0 − M2 1 M2w2u0 + ckm3P1w2u1 + E1m3Mw2u1 + E1m3M2w2u1 + m3M2 1 w2u1 − E1m2 3w2u2 + ck3P1v0 + k2M2 1 v0 + ckm2MP1v0 + ckm2M2P1v0 − ckm2m3P1v1 − m2m2 3Mv0 + m2MM2 1 v0 − m2m2 3M2v0 + m2M2 1 M2v0 − E1k2m3v1 − E1m2m3Mv1 − E1m2m3M2v1 + E1m2m2 3v2 − m2m3M2 1 v1, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11) X2 =ckP1w2u0 + M2 1 w2u0 − E1m3w2u1 + ckm2P1v0 + ckm3P1v1 + m3M2 1 v1 − ckMP1v0 − ckM2P1v0 + m2 3Mv0 + m2 3M2v0 + m2M2 1 v0 − MM2 1 v0 − M2 1 M2v0 + E1m3Mv1 + E1m3M2v1 − E1m2m3v1 − E1m2 3v2, (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='X3 = − cE1kP1w2u0 − E1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 w2u0 + ckMP1w2u0 + ckM2P1w2u0 + MM2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 w2u0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ M2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 w2u0 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m3w2u1 − E1m3Mw2u1 − E1m3M2w2u1 + cE1km2P1v0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− cE1km3P1v1 − cE1kMP1v0 − cE1kM2P1v0 + E1m2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 + E1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3Mv0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ E1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2v0 − E1MM2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 − E1M2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 + ckm3MP1v1 + ckm3M2P1v1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− ckm2MP1v0 − ckm2M2P1v0 + ckM2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 P1v0 − m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 − m2MM2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− m2M2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 + M4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v0 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m3Mv1 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m3M2v1 − E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m2m3v1 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3v2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− 2E1m3M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v1 + E1m2m3Mv1 + E1m2m3M2v1 − E1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3Mv2 − E1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2v2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ m3MM2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v1 + m3M2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 v1,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='13) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='X4 =v0m2M4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + u0w2M4 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 − v0m2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + E1k2v0M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + 2E2k2v0M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + m2m3M2v1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− k2Mv0M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 − E1Mv0m2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + k2v0M2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 − E1v0m2M2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + ckv0m2P1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− u0m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3w2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 − E1Mu0w2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 − E1u0M2w2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + cku0P1w2M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + 2E2m3w2u1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− Mm3w2u1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + m3M2w2u1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 − 2E1m2m3v1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 + Mm2m3v1M2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ E1Mv0m2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3 + E1v0m2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2 + cE1k3v0P1 + 2cE2k3v0P1 − ck3Mv0P1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− cE1kMv0m2P1 + ck3v0M2P1 − cE1kv0m2M2P1 + E1Mu0m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3w2 + E1u0m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2w2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='− cE1kMu0P1w2 − cE1ku0M2P1w2 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1Mm3w2u1 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m3M2w2u1 + cE1km3P1w2u1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ 2cE2km3P1w2u1 − ckMm3P1w2u1 + ckm3M2P1w2u1 − E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3w2u2 − 2E1E2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3w2u2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ E1Mm2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3w2u2 − E1m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2w2u2 − E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1k2m3v1 − 2E1E2k2m3v1 + E1k2Mm3v1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1Mm2m3v1 − E1k2m3M2v1 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m2m3M2v1 − cE1km2m3P1v1 + ckMm2m3P1v1 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='+ ckm2m3M2P1v1 + E2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1m2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3v2 − E1Mm2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3v2 − E1m2m2 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3M2v2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' (A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='14) – 19 – Acknowledgments The author Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li thanks Prof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Fen-Kun Guo of ITP-CAS, and Dr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xu-Chang Zheng of Chongqing Univ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', and Dr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hao Xu of Northwest Normal Univ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' for helpful suggestions and discussions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' This work is supported by the National Natural Science Foundation of China (NSFC) under Grant Nos.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 12005169, 12075301, 11821505, 12047503, 11805024, 11865001, and 12075073.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' It is also supported by the National Key R&D Program of China (2022YFA1604803), the Natural Science Basic Research Program of Shaanxi (Program No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 2021JQ-074), and the Fundamental Research Funds for the Central Universities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' References [1] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Aaij, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 122 (22) (2019) 222001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='03947, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='122.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='222001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [2] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Gershon (6 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='15233.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [3] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wu, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Molina, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Oset, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zou, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 105 (2010) 232001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='0573, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='232001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [4] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Huang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhang, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zou, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 84 (2011) 015203.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='0453, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='015203.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [5] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wu, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lee, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zou, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 85 (2012) 044002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1036, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='044002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [6] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Sun, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' He, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhu, Chin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 36 (2012) 6–13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2901, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1674-1137/36/1/002,10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1674-1137/36/3/006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [7] X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 74 (12) (2014) 3198.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1409.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='3332, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-014-3198-3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [8] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Karliner, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rosner, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 115 (12) (2015) 122001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1506.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06386, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='115.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='122001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [9] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 115 (13) (2015) 132002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1507.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='03704, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='115.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='132002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [10] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xiao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Huang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Geng, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (1) (2019) 014022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='00872, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [11] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Pan, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Peng, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S´anchez S´anchez, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Geng, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hosaka, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Pavon Valderrama, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 122 (24) (2019) 242001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11560, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='122.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='242001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [12] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Sun, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (1) (2019) 011502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11013, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='011502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [13] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xiao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Nieves, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Oset, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (1) (2019) 014021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='01296, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [14] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' He, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 79 (5) (2019) 393.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11872, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-019-6906-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [15] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lin, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zou, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (5) (2019) 056005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1908.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='05309, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='056005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 20 – [16] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (5) (2019) 051501.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11001, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='051501.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [17] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Ali, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Parkhomenko, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' B 793 (2019) 365–371.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='00446, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='physletb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='05.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [18] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Meng, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (1) (2019) 014031.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1905.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04113, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014031.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [19] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Burns, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Swanson, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (11) (2019) 114033.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1908.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='03528, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='114033.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [20] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Voloshin, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (3) (2019) 034020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1907.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='01476, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='034020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [21] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Guo, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Sakai, Prog.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 112 (2020) 103757.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1912.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='07030, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='ppnp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='103757.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [22] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Ke, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 101 (1) (2020) 014024.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1909.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12509, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014024.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [23] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Du, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Baru, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Guo, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hanhart, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Meißner, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Oller, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 124 (7) (2020) 072001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1910.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11846, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='124.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='072001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [24] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Chin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 44 (11) (2020) 113106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='13028, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1674-1137/abb080.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [25] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yamaguchi, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Garcia-Tecocoatzi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Giachino, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hosaka, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Santopinto, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Takeuchi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Takizawa, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 101 (9) (2020) 091502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1907.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04684, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='091502.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [26] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xu, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 101 (5) (2020) 054037.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='02980, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='054037.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [27] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Burns, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Swanson, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' A 58 (4) (2022) 68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11527, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epja/s10050-022-00723-9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [28] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Fernandez-Ramirez, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Pilloni, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Albaladejo, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jackura, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Mathieu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Mikhasenko, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Silva-Castro, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Szczepaniak, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 123 (9) (2019) 092001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='10021, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevLett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='123.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='092001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [29] W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Ruangyoo, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phumphan, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Limphirat, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yan, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' G 49 (7) (2022) 075001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='14249, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1361-6471/ac58af.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [30] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Stancu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 104 (5) (2021) 054050.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2108.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='05841, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='054050.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [31] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Nakamura, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 103 (2021) 111503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06817, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='L111503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [32] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Nakamura, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hosaka, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yamaguchi, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 104 (9) (2021) L091503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2109.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='15235, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='L091503.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [33] U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' ¨Ozdem, Chin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 45 (2) (2021) 023119.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1674-1137/abd01c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [34] U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' ¨Ozdem, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 81 (4) (2021) 277.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='01996, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-021-09070-3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [35] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Huang, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 81 (5) (2021) 421.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='07937, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-021-09211-8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 21 – [36] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Sakai, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jing, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Guo, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (7) (2019) 074007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1907.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='03414, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='074007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [37] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Dong, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Shen, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Huang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhang, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 80 (4) (2020) 341.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='08051, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-020-7890-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [38] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xiao, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 102 (3) (2020) 036012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1911.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='09613, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='036012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [39] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chen, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 80 (10) (2020) 945.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='09563, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-020-08519-1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [40] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Cui, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Huang, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 102 (3) (2020) 034028.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1907.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='05097, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='034028.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [41] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Chin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 44 (2020) 103102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1907.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04582, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1674-1137/ababf7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [42] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Aaij, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 102 (11) (2020) 112012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='11292, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='112012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [43] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Qin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Chin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 44 (2020) 013102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1903.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='02282, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/1674-1137/44/1/013102.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [44] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Qin, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 82 (2022) 60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='10966, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-022-10006-8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [45] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 104 (1) (2021) 014018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='12372, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [46] C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Kim, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' B 623 (2005) 218–226.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='physletb.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='07.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='059.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [47] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' G: Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 39 (2012) 015009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1107.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='0474, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1088/0954-3899/39/1/015009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [48] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Fu, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Ju, JHEP 07 (2013) 120.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1305.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1067, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1007/JHEP07(2013)120.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [49] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jiang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yuan, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 76 (8) (2016) 454.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-016-4306-3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [50] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jiang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yuan, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhou, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 77 (1) (2017) 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-016-4588-5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [51] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jiang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yuan, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, Eur.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 77 (5) (2017) 297.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1701.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='03252, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1140/epjc/s10052-017-4865-y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [52] Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Li, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jiang, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chang, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 100 (7) (2019) 076020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1802.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06351, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='076020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [53] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Casalbuoni, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Deandrea, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Di Bartolomeo, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Gatto, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Feruglio, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Nardulli, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rept.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 281 (1997) 145–238.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:hep-ph/9605342, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/S0370-1573(96)00027-0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [54] T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Cheng, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Cheung, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lin, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lin, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Yu, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 46 (1992) 1148–1164.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1148.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [55] H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Cheng, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Chua, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 75 (2007) 014006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:hep-ph/0610283, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 22 – [56] Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Liu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Oka, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 85 (2012) 014015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1103.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='4624, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='014015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [57] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Guo, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Hanhart, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Meißner, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wang, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zhao, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zou, Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Mod.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 90 (1) (2018) 015004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1705.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='00141, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/RevModPhys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='015004.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [58] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Wise, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 45 (7) (1992) R2188.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='R2188.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [59] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Burdman, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Donoghue, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Lett.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' B 280 (1992) 287–291.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/0370-2693(92)90068-F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [60] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Jenkins, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Luke, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Manohar, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Savage, Nucl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' B 390 (1993) 463–473.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:hep-ph/9204238, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/0550-3213(93)90464-Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [61] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Colangelo, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' De Fazio, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Pham, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' D 69 (2004) 054023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:hep-ph/0310084, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='054023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [62] E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Garzon, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Xie, Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Rev.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' C 92 (3) (2015) 035201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1506.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06834, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1103/PhysRevC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='035201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [63] P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Zyla, et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=', PTEP 2020 (8) (2020) 083C01.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1093/ptep/ptaa104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [64] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Mertig, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Bohm, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Denner, Comput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 64 (1991) 345–359.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/0010-4655(91)90130-D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [65] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Shtabovenko, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Mertig, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Orellana, Comput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 207 (2016) 432–444.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:1601.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='01167, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='cpc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='06.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' [66] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Shtabovenko, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Mertig, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Orellana, Comput.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' Commun.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' 256 (2020) 107478.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' arXiv:2001.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='04407, DOI:10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='1016/j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='cpc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content='107478.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
+page_content=' – 23 –' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/nNA0T4oBgHgl3EQfJv9r/content/2301.02094v1.pdf'}
diff --git a/nNE2T4oBgHgl3EQfJgZo/content/2301.03692v1.pdf b/nNE2T4oBgHgl3EQfJgZo/content/2301.03692v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..16df0a42b209a2965a892dbc0bcd591a01fb4eff
--- /dev/null
+++ b/nNE2T4oBgHgl3EQfJgZo/content/2301.03692v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fa2c1d8cbe65107f772afc9a87200c3cb83b1127df1a596c8e6ba2407e9bc4a
+size 1141962
diff --git a/nNE2T4oBgHgl3EQfJgZo/vector_store/index.faiss b/nNE2T4oBgHgl3EQfJgZo/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..a3af89a7910b4c57fdb61aa9ed246c5f6f4b646e
--- /dev/null
+++ b/nNE2T4oBgHgl3EQfJgZo/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49503f620d7e4b9a3c9c390121bf933761b3f9aff63d78585b6b48ceae3b1e83
+size 1507373
diff --git a/nNE2T4oBgHgl3EQfJgZo/vector_store/index.pkl b/nNE2T4oBgHgl3EQfJgZo/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..f846e6139d47775499d7b07c5749d6fbdca2703d
--- /dev/null
+++ b/nNE2T4oBgHgl3EQfJgZo/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36b26f2c646be343766a30b13d9816396f746e567740cb493c51bd2f4820464c
+size 53378
diff --git a/o9FAT4oBgHgl3EQfeB1V/vector_store/index.faiss b/o9FAT4oBgHgl3EQfeB1V/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..b2accde5d4cca08db35b1b239426698287b31980
--- /dev/null
+++ b/o9FAT4oBgHgl3EQfeB1V/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44241471b4281916eb43e0dfdd81157161b5535de847db6e2b78dc8e3f8c5687
+size 1179693
diff --git a/o9FAT4oBgHgl3EQfeB1V/vector_store/index.pkl b/o9FAT4oBgHgl3EQfeB1V/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..18ce65b5a24db844eedc4f8a7e74d74ade0c8360
--- /dev/null
+++ b/o9FAT4oBgHgl3EQfeB1V/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1add0084885213dd3cde6ac5896a24d64d352dc617b40593a3d20e1cb1765b5
+size 41398
diff --git a/otE1T4oBgHgl3EQfOwPh/content/2301.03020v1.pdf b/otE1T4oBgHgl3EQfOwPh/content/2301.03020v1.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..d87911962feabe3b20b7a85c712e523ae1076a0a
--- /dev/null
+++ b/otE1T4oBgHgl3EQfOwPh/content/2301.03020v1.pdf
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5be9ae24eee4805a78eeee50fc5287a35ee719988afb932b3d55fa7d9d167ba0
+size 266879
diff --git a/otE1T4oBgHgl3EQfOwPh/vector_store/index.faiss b/otE1T4oBgHgl3EQfOwPh/vector_store/index.faiss
new file mode 100644
index 0000000000000000000000000000000000000000..70666e2ed4350b9d9ccac8b0edc93840eecfadf1
--- /dev/null
+++ b/otE1T4oBgHgl3EQfOwPh/vector_store/index.faiss
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a16cf1fc24e6bfcb837c52c7da28c56b8a7f6c3241bee1f191f5c3855bf3e5e5
+size 3604525
diff --git a/p9FKT4oBgHgl3EQfzC7B/content/tmp_files/2301.11910v1.pdf.txt b/p9FKT4oBgHgl3EQfzC7B/content/tmp_files/2301.11910v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..8270da127eadf30e42d6dae371389d9763a888ae
--- /dev/null
+++ b/p9FKT4oBgHgl3EQfzC7B/content/tmp_files/2301.11910v1.pdf.txt
@@ -0,0 +1,602 @@
+arXiv:2301.11910v1 [math.GR] 27 Jan 2023
+REVERSIBILITY AND REAL ADJOINT ORBITS OF LINEAR MAPS
+KRISHNENDU GONGOPADHYAY, TEJBIR LOHAN AND CHANDAN MAITY
+Dedicated to the 80th birthday of Norbert A’Campo.
+Abstract. We extend classical results on the classification of reversible elements of the group
+GL(n, C) (and GL(n, R)) to GL(n, H) using an infinitesimal version of the classical reversibility,
+namely adjoint reality in the Lie algebra set-up. We also provide a new proof of such a classi-
+fication for the general linear groups over R and C. Further, we classify the real adjoint orbits
+in the Lie algebra gl(n, D) for D = R, C or H.
+1. Introduction
+Reversing and time-reversing symmetries are important classes of symmetries that appear
+in the natural science. Especially they arise in many physically motivated dynamical systems.
+There is an extensive literature discussing such symmetries in different areas of physics and dy-
+namical systems, see e.g. [La], [LR]. In a mathematical terminology, the motions of a dynamical
+system may be associated with a group G, and such symmetries correspond to the “reversible”
+and “strongly reversible” elements in G. An element g in a group G is called reversible if g
+and g−1 are conjugate in G , that is, if there exists h ∈ G such that hgh−1 = g−1. An element
+g in a group G is called strongly reversible if g is conjugate to g−1 by an involution (i.e., an
+element of order at most two) in G. Equivalently, strongly reversible elements are products of
+two involutions. Some authors have called strongly reversible elements “bireflectional”.
+Reversible maps have appeared from different perspectives in the literature, e.g. [AA], [De],
+[Se], [FS]. From a group-theoretical perspective, a classical theorem of Frobenius and Schur
+asserts that the number of real-valued complex irreducible characters of a finite simple group
+G is equal to the number of reversible conjugacy classes of G. With this motivation, many
+mathematicians have used the terminology “real” and “strongly real” elements.
+A strongly
+reversible element is reversible, but the converse is not true in general. It is a problem of potential
+interest to classify reversible and strongly reversible elements in different groups of interest. In
+the theory of finite groups, the classification of such elements is relatively well-understood in the
+literature. However, a complete classification of reversible classes is not available other than for
+a few families of infinite groups. Some of the infinite groups where it has been classified include
+compact Lie groups, real rank one classical groups, and isometry groups of Hermitian spaces,
+see e.g. [FS, BG, GL].
+The idea of reversibility is apparent in the work of A’Campo [A’C] as well. A’Campo investi-
+gated the monodromy of real isolated singularities using the fact that the complex conjugation
+on complex space permutes the level sets of a real polynomial function and induces involu-
+tions on level sets corresponding to real values. In particular, it was proved that the geometric
+monodromy is the composition of the involution induced by complex conjugation and another
+involution. In other words, the geometric monodromies are strongly reversible. A’Campo called
+2020 Mathematics Subject Classification. Primary 20E45;
+Secondary 15B33, 22E60.
+Key words and phrases. General linear group, reversibility, adjoint reality.
+1
+
+2
+K. GONGOPADHYAY, T. LOHAN AND C. MAITY
+the corresponding singularity strongly invertible.
+Further, it follows that any two geometric
+monodromies are “linked” by the involution that comes from complex conjugation.
+Recently, the concept of reversibility has been extended to semisimple Lie algebras using the
+adjoint representations of Lie groups in [GM]. The infinitesimal notion that has been introduced
+for the Lie algebras is called adjoint reality. Understanding the adjoint orbits of a semisimple
+Lie group is an active research theme, see the survey [CM]. However, the exploration of adjoint
+reality properties has been the object of attention only very recently, cf. [GM], [GLM]. As
+an application of adjoint reality for nilpotent orbits in simple Lie algebras, the reversible and
+strongly reversible unipotent elements in the classical simple Lie groups have been completely
+classified, see [GM].
+Let D = R, C or H. In this chapter, we revisit the classification of reversible elements in the
+general linear group GL(n, D). The classification of reversible elements in GL(n, D) and their
+equivalence with the strongly reversible elements are well known in the literature for D = R or
+C, cf. [Wo], [FS]. Extending these results over the quaternions is not straightforward due to the
+non-commutativity of H. We shall overcome such difficulties by approaching this problem using
+adjoint reality in the Lie algebra gl(n, D). We recall the notion of adjoint reality below.
+Consider the adjoint action of the general linear group G := GL(n, D) on its Lie algebra
+g := gl(n, D). Recall that gl(n, D) ≃ M(n, D), the algebra of n × n matrices over D. In this case,
+the adjoint action is given by the conjugation. Consider the conjugacy action of GL(n, D) on
+gl(n, D): Ad(g)X := gXg−1. An element X ∈ g is called AdG-real if −X = gXg−1 for some
+g ∈ G. An AdG-real element X ∈ g is called strongly AdG-real
+if −X = τXτ −1 for some
+involution (i.e., element of order at most two) τ ∈ G; see [GM, Definition 1.1]. Observe that if
+X ∈ g is AdG-real, then exp(X) is reversible in G, but the converse may not be true.
+We classify the adjoint real elements in gl(n, D) and investigate their equivalence with the
+strongly adjoint real elements in gl(n, D); see Theorems 4.1, 4.2, 4.4. Using these ideas and
+applying some of these results, we shall prove that some particular types of Jordan forms in
+GL(n, D) are strongly reversible. This will be used to classify the reversible and strongly re-
+versible elements in GL(n, D).
+This approach not only reduces the complexity of the computations but also gives a better
+understanding of reversibility in the group GL(n, D). This also provides a uniform treatment
+over D = R, C or H. We show by a counterexample in Example 5.3 that, unlike the field case, the
+notion of reversible and strongly reversible elements are not equivalent in GL(n, H) in general.
+We give a sufficient criterion for equivalence of the two notions in GL(n, H), see Theorem 5.4.
+The chapter is organized as follows. In §2, we fix some notation and recall some background
+related to Jordan canonical forms. In §3, reversibility and adjoint reality of certain Jordan forms
+are described. In §4, we deal with the adjoint reality in the Lie algebra gl(n, D). We revisit
+the reversibility problem of GL(n, D) in §5 and provide a much simpler proof of earlier obtained
+results.
+Acknowledgement. It has been a highly rewarding experience to know Norbert A’Campo.
+Interaction with him has always been rich in mathematical and non-mathematical ideas! It is a
+pleasure to dedicate this small contribution to the volume in honor of Norbert. We wish Norbert
+a very good health and a happy life ahead.
+Gongopadhyay is partially supported by the SERB core research grant CRG/2022/003680.
+Lohan acknowledges support from the CSIR SRF grant, File No. : 09/947(0113)/2019-EMR-I.
+Maity is supported by an NBHM PDF during this work.
+
+REVERSIBILITY IN GL(n, D)
+3
+2. Preliminaries
+In this section, we will recall some necessary background. Recall that H := R + Ri+ Rj+ Rk
+denotes the division algebra of Hamilton’s quaternions. We consider Hn as a right H-module.
+We refer to [Ro] for a nice exposition on quaternion linear algebra.
+Definition 2.1. Let A ∈ M(n, H). A non-zero vector v ∈ Hn is said to be a right eigenvector
+of A corresponding to a right eigenvalue λ ∈ H if the equality Aλ = vλ holds.
+Eigenvalues of A ∈ M(n, H) occur in similarity classes, i.e., if v is an eigenvector corresponding
+to λ, then vµ ∈ vH is an eigenvector corresponding to µ−1λµ. Each similarity class of eigen-
+values contains a unique complex number with a non-negative imaginary part. Here, instead
+of similarity classes of eigenvalues, we will consider the unique complex representative with a
+non-negative imaginary part.
+Let ψ: C −→ M(2, R) be the embedding given by ψ(z) :=
+�
+Re(z)
+Im(z)
+−Im(z)
+Re(z)
+�
+. This induces
+the embedding Ψ: M(n, C) −→ M(2n, R) defined as
+Ψ((zi,j)n×n) :=
+�
+ψ(zi,j)
+�
+2n×2n .
+(2.1)
+It follows from the definition of the exponential map exp : M(n, C) −→ GL(n, C) that
+Ψ(exp(X)) = exp(Ψ(X))
+∀ X ∈ M(n, C).
+(2.2)
+Definition 2.2 (cf. [Ro, p. 94]). A Jordan block J(λ, m) is an m × m matrix with λ ∈ D on
+the diagonal entries, 1 on all of the super-diagonal entries and zero elsewhere. We will refer to
+a block diagonal matrix where each block is a Jordan block as a Jordan form.
+We also consider the following block matrix as a Jordan form over R, cf. [GLR, p. 364], which
+corresponds to the case when the eigenvalues of a matrix over R belong to C \ R. Recall the
+embedding Ψ as in (2.1). Let K := Ψ(µ + iν) =
+�
+µ
+ν
+−ν
+µ
+�
+∈ M(2, R), where µ, ν are real
+numbers with ν > 0. Then define
+JR(µ ± iν, 2n) := Ψ(J(µ + iν, n)) =
+
+
+
+
+
+
+
+K
+I2
+K
+I2
+...
+K
+I2
+K
+
+
+
+
+
+
+
+∈ M(2n, R),
+(2.3)
+where I2 denotes the 2×2 identity matrix; see [Ro, Theorem 15.1.1], [GLR, Chapter 12]. Further,
+we also define
+JR(µ ∓ iν, 2n) := Ψ(J(µ − iν, n)) = σ
+�
+Ψ(J(µ + iν, n))
+�
+σ−1,
+(2.4)
+where σ = diag(1, −1, 1, −1, . . . , (−1)2n−1 )2n×2n.
+Remark 2.3. We will follow the notation JR(µ±iν, n) and JR(µ∓iν, n) as defined in (2.3) and
+(2.4) throughout this chapter. Note that {JR(µ ± iν, n)} is a singleton by the above definition.
+Next we recall the Jordan form in M(n, D), see [Ro, Theorem 15.1.1, Theorem 5.5.3].
+Lemma 2.4 (Jordan form in M(n, D), cf. [Ro]). For every A ∈ M(n, D), there is an invertible
+matrix S ∈ GL(n, D) such that SAS−1 has the following form:
+
+4
+K. GONGOPADHYAY, T. LOHAN AND C. MAITY
+(1) For D = R, SAS−1 = J(λ1, m1) ⊕ · · · ⊕ J(λk, mk)
+�
+JR(µ1 ± iν1, 2ℓ1) ⊕ · · · ⊕ JR(µq ± iνq, 2ℓq),
+(2.5)
+where λ1, . . . , λk; µ1, . . . , µq ; ν1, . . . , νq are (not necessarily distinct) real numbers and
+ν1, . . . , νq are positive.
+(2) For D = C,
+SAS−1 = J(λ1, m1) ⊕ · · · ⊕ J(λk, mk),
+(2.6)
+where λ1, . . . , λk are (not necessarily distinct) complex numbers.
+(3) For D = H,
+SAS−1 = J(λ1, m1) ⊕ · · · ⊕ J(λk, mk),
+(2.7)
+where λ1, . . . , λk are (not necessarily distinct) complex numbers and have non-negative
+imaginary parts.
+The forms (2.5), (2.6) and (2.7) are uniquely determined by A up to a permutation of Jordan
+blocks.
+3. Strong Reversibility of Jordan forms
+In this section, we investigate the adjoint reality in gl(n, D) and reversibility in GL(n, D) for
+certain types of Jordan forms.
+3.1. Strong adjoint reality of Jordan forms in gl(n, D). Here, we will consider some par-
+ticular types of Jordan forms in gl(n, D) and show that they are strongly AdGL(n,D)-real by
+explicitly constructing a suitable reversing involution.
+Lemma 3.1. Let D = R, C or H, and X := J(0, n) be the nilpotent element in gl(n, D). Then
+X is strongly AdGL(n,D)-real.
+Proof. Let g := diag(1, −1, 1, −1, . . . , (−1)n−1 )n×n. Then g is an involution in GL(n, D)
+such that gXg−1 = −X. This completes the proof.
+□
+Lemma 3.2. Let X := J(λ, n) ⊕ J(−λ, n) be the Jordan form in gl(2n, D), where λ ∈ D \ {0},
+for D = R or C, and for D = H, λ ∈ C \ {0} with non-negative imaginary part such that the real
+part of λ ̸= 0. Then X is strongly AdGL(2n,D)-real.
+Proof. Write X =
+�
+J(λ, n)
+J(−λ, n)
+�
+. Let τ := diag(1, −1, 1, −1, . . . , (−1)n−1 )n×n. Note
+that J(λ, n) τ = −τ J(−λ, n). Consider g =
+�
+τ
+τ
+�
+∈ GL(2n, D). Then g is an involution in
+GL(2n, D) such that gXg−1 = −X. This completes the proof.
+□
+Lemma 3.3. Let X := J(µi, n) ⊕ J(µi, n) be the Jordan form in gl(2n, H), where µ ∈ R, µ > 0.
+Then X is strongly AdGL(2n,H)-real.
+Proof. Let τ := diag(j, −j, j, −j, . . . , (−1)n−1j)n×n. Then τ 2 = −In and τ J(µi, n) τ −1 =
+−J(µi, n). Consider the involution g =
+�
+τ
+−τ
+�
+in GL(2n, H). Then g is an involution in
+GL(2n, H) such that gXg−1 = −X. This proves the lemma.
+□
+Recall that the matrix JR(µ ± iν, 2n) is defined in (2.3).
+Lemma 3.4. Let X := JR(0 ± iν, 2n) be the Jordan block in gl(2n, R), where ν ∈ R such that
+ν > 0. Then X is strongly AdGL(2n,R)-real.
+
+REVERSIBILITY IN GL(n, D)
+5
+Proof. Let g := diag(I1,1, −I1,1, I1,1, −I1,1, . . . , (−1)n−1I1,1 )2n×2n, where I1,1 :=
+�
+1
+0
+0
+−1
+�
+.
+Then g is an involution in GL(2n, R) such that gXg−1 = −X. This completes the proof.
+□
+We refer to (2.4) for the notation JR(µ ∓ iν, 2n).
+Lemma 3.5. Let X := JR(µ ± iν, 2n) ⊕ JR(−µ ∓ iν, 2n) be the Jordan form gl(4n, R), where
+µ, ν ∈ R such that µ ̸= 0 and ν > 0. Then X is strongly AdGL(4n,R)-real.
+Proof. Write X =
+�
+JR(µ ± iν, 2n)
+JR(−µ ∓ iν, 2n)
+�
+. Consider g =
+�
+τ
+τ
+�
+, where τ :=
+diag(I2, −I2, I2, −I2, . . . , (−1)n−1I2 )2n×2n. Observe that JR(µ ± iν, 2n) τ = −τ JR(−µ ∓ iν, 2n).
+Then g is an involution in GL(4n, R) such that gXg−1 = −X. This completes the proof.
+□
+3.2. Strong reversibility of Jordan forms in GL(n, D). We will apply the results obtained
+in §3.1 to provide a different proof of strong reversibility of some particular types of Jordan
+forms in GL(n, D). Such results are known for D = R, C. We shall extend these results over H.
+These results will be used in the Proposition 5.2 and the Theorem 5.4.
+Lemma 3.6. Let D = R, C or H, and A := J(µ, n) be the Jordan block in GL(n, D), where
+µ ∈ {±1}. Then A is strongly reversible in GL(n, D).
+Proof. First, we will consider µ = 1. Then J(1, n) = In + J(0, n). By setting N := J(0, n),
+we have gNg−1 = −N, g2 = In, where g is as in Lemma 3.1. This implies g eNg−1 = e−N. Since
+the Jordan form of eN is J(1, n), J(1, n) = τeNτ −1 for some τ ∈ GL(n, D). Now
+(τgτ −1)J(1, n)(τg−1τ −1) = τgeNg−1τ −1 = τe−Nτ −1 = (J(1, n))−1 .
+Hence, J(1, n) is strongly reversible. The case when µ = −1 follows from the fact that −J(−1, n)
+has Jordan form J(1, n) and J(1, n) is strongly reversible. This completes the proof.
+□
+Lemma 3.7. Let A := J(λ, n) ⊕ J(λ−1, n) be the Jordan form in GL(2n, D), where λ ∈ D \
+{±1, 0}, for D = R or C, and for D = H, λ ∈ C \ {0} with non-negative imaginary part such
+that |λ| ̸= 1. Then A is strongly reversible in GL(2n, D).
+Proof. Let µ ∈ C such that eµ = λ. Note that exp(J(µ, n)) = exp(µIn) · exp(J(0, n)) =
+λ exp(J(0, n)). Let P ∈ GL(n, D) so that λ exp(J(0, n)) = PJ(λ, n) P −1. Thus
+exp(J(µ, n)) = PJ(λ, n) P −1.
+(3.1)
+Similarly, there exists Q ∈ GL(n, D) such that
+exp(J(−µ, n)) = Q J(λ−1, n) Q−1.
+(3.2)
+Now, let σ := diag(1, −1, . . . , (−1)n−1) be an involution in GL(n, D) so that −J(−µ, n) =
+σ J(µ, n) σ−1. This implies
+σ exp(J(µ, n)) σ−1 =
+�
+exp(J(−µ, n))
+�−1
+.
+(3.3)
+Using (3.1), (3.2) and (3.3), we have
+σ P J(λ, n) P −1 σ−1 =
+�
+Q J(λ−1, n) Q−1�−1
+.
+(3.4)
+Consider the involution g :=
+�
+P
+Q
+�−1 �
+σ
+σ−1
+� �
+P
+Q
+�
+in GL(2n, D). Then gAg−1 = A−1
+if and only if
+(Q−1σ P) J(λ, n) (Q−1σ P)−1 =
+�
+J(λ−1, n)
+�−1
+.
+(3.5)
+Now the proof follows from Equation (3.4).
+□
+
+6
+K. GONGOPADHYAY, T. LOHAN AND C. MAITY
+Lemma 3.8. Let A := J(µ, n) ⊕ J(µ, n) be the Jordan block in GL(2n, H), where µ ∈ C\{±1},
+with non-negative imaginary part such that |µ| = 1. Then A is strongly reversible in GL(2n, H).
+Proof. Recall that j Z = Z j for all Z ∈ GL(n, C), where Z is the matrix obtained by taking
+the conjugate of each entry of the complex matrix Z. We can assume that A = J(eiθ, n) ⊕
+J(eiθ, n), where θ ∈ (0, π). Write A =
+�
+P
+P
+�
+, where P = J(eiθ, n) ∈ GL(n, C). To show A is
+strongly reversible, it is sufficient to find a g =
+�
+X
+X−1
+�
+∈ GL(2n, H) such that P −1X = XP,
+where X ∈ GL(n, H). Further, If X = Y j for some Y ∈ GL(n, C), then we require P −1Y = Y P,
+i.e.,
+�
+J(eiθ, n)
+�−1
+Y = Y J(e−iθ, n). Using the construction as done in the proof of Lemma 3.7,
+we can find such Y in GL(n, C), see (3.5). This completes the proof.
+□
+Lemma 3.9. Let A := JR(µ ± iν, 2n) be the Jordan block in GL(2n, R) as in (2.3). If µ2 +ν2 =
+1, then A is strongly reversible in GL(2n, R).
+Proof. Let K :=
+�
+µ
+ν
+−ν
+µ
+�
+where µ2 +ν2 = 1. Then K ∈ SO(2). Let Y =
+�
+0
+a
+−a
+0
+�
+∈ so(2)
+such that exp(Y ) = K.
+Note that for I1,1 := diag(1, −1), we have I1,1 Y (I1,1)−1 = −Y .
+Consider
+X := JR(0 ± ia, 2n) =
+
+
+
+
+
+Y
+I2
+...
+...
+Y
+I2
+Y
+
+
+
+
+ ∈ GL(2n, R).
+Let g := diag(I1,1, −I1,1, . . . , (−1)n−1I1,1) ∈ GL(2n, R). Then gXg−1 = −X, and hence exp(X)
+is strongly reversible. To conclude the proof, it is sufficient to show that exp(X) and JR(µ ±
+iν, 2n) are conjugate. To see this recall that Ψ(J(µ+iν, n)) = JR(µ±iν, 2n), and Ψ(J(ia, n)) =
+X, where the map Ψ is as in (2.1). Since exp(Y ) = K, we have eia = µ + iν. Then
+exp(J(ia, n)) = (µ + iν) exp(J(0, n)) = h J(µ + iν, n) h−1 ,
+(3.6)
+for some h ∈ GL(n, C). By using Equations (2.2) and (3.6), we have
+exp(X) = Ψ(exp(J(ia, n)) = h
+�
+Ψ(J(µ + iν, n))
+�
+h−1 = h JR(µ ± iν, 2n) h−1.
+Therefore, JR(µ ± iν, 2n) is strongly reversible in GL(2n, R). This completes the proof.
+□
+Lemma 3.10. Let A := JR(µ ± iν, 2n) ⊕ JR
+�
+µ
+µ2+ν2 ∓ i
+ν
+µ2+ν2, 2n
+�
+be the Jordan form in
+GL(4n, R), where µ2 + ν2 ̸= 1. Then A is strongly reversible in GL(4n, R).
+Proof. Let z ∈ C such that ez = µ+iν and e−z =
+µ
+µ2+ν2 −i
+ν
+µ2+ν2. Recall that the embedding
+Ψ is as in (2.1). Using (2.3) and (2.4), we have
+Ψ(J(ez, n)) = JR(µ ± iν, 2n) and Ψ(J(e−z, n)) = JR(
+µ
+µ2 + ν2 ∓ i
+ν
+µ2 + ν2 , 2n).
+Therefore, we can write A as
+A = Ψ(P), where P =
+�
+J(ez, n)
+J(e−z, n)
+�
+∈ GL(2n, C).
+(3.7)
+The proof now follows from Lemma 3.7 and the fact that Ψ is an embedding.
+□
+
+REVERSIBILITY IN GL(n, D)
+7
+4. Adjoint reality in gl(n, D)
+In this section, we investigate adjoint reality in gl(n, D).
+First we classify AdGL(n,D)-real
+elements in the Lie algebra gl(n, D).
+Theorem 4.1. Let D = R, C or H. An element X ∈ gl(n, D) with Jordan canonical form as
+given in Lemma 2.4 is AdGL(n,D)-real if and only if the following hold:
+(1) For D = R, the blocks can be partitioned into pairs {JR(µ ± iν, 2t), JR(−µ ∓ iν, 2t)},
+{J(λ, s), J(−λ, s)} or, singletons {JR(0±iν, 2ℓ)}, {J(0, m)}, where λ, µ, ν ∈ R and λ, µ ̸=
+0, ν > 0.
+(2) For D = C, the blocks can be partitioned into pairs {J(λ, s), J(−λ, s)} or, singletons
+{J(0, m)}, where λ ∈ C and λ ̸= 0.
+(3) For D = H, the blocks can be partitioned into pairs {J(λ, s), J(−λ, s)} or, singletons
+{J(µ, m)}, where λ, µ ∈ C with non-negative imaginary parts such that real part of λ ̸= 0
+and real part of µ = 0.
+Proof. Consider the case D = R. Using Lemma 2.4, X is conjugate to −X if and only if
+{J(λ1, m1), . . . , J(λk, mk), JR(µ1 ± iν1, 2ℓ1), . . . , JR(µq ± iνq, 2ℓq)}
+= {J(−λ1, m1), . . . , J(−λk, mk), JR(−µ1 ∓ iν1, 2ℓ1), . . . , JR(−µq ∓ iνq, 2ℓq)},
+and the result follows immediately for the case D = R. Recall the fact that for a unique complex
+representative λ of an eigenvalue class of X, [λ] = [−λ] if and only if the real part of λ = 0.
+Using the same line of argument as we used in the D = R case, the result follows for the case
+D = C or H.
+□
+Recall that every strongly AdGL(n,D)-real element in gl(n, D) is AdGL(n,D)-real. In the next
+result, we will prove that the converse holds when D = R or C.
+Theorem 4.2. Let D = R or C. An element A of gl(n, D) is AdGL(n,D)-real if and only if it is
+strongly AdGL(n,D)-real.
+Proof. Without loss of generality, we can assume that A is in Jordan form as in Lemma 2.4.
+Suppose A is AdGL(n,D)-real. Then Jordan blocks of A can be partitioned as in Theorem 4.1.
+In view of Lemma 3.1, Lemma 3.2, Lemma 3.4, and Lemma 3.5, we can explicitly construct an
+involution g in GL(n, D) such that gAg−1 = −A. Since the converse is trivial, this completes
+the proof.
+□
+The following example shows that the above result does not hold for gl(n, H).
+Example 4.3. Let A := (i) ∈ gl(1, H). Then gAg−1 = −A, where g = (j) ∈ GL(1, H). So
+A is AdGL(1,H)-real. Suppose that A is strongly AdGL(1,H)-real. Let g = (a) ∈ GL(1, H) be
+an involution such that gAg−1 = −A. Then we get ai = −ia. This implies a = wj for some
+w ∈ C, w ̸= 0. Since g is an involution, a2 = (wj)2 = 1, i.e, |w|2 = −1. This is a contradiction.
+Therefore, A is AdGL(1,H)-real but not strongly AdGL(1,H)-real.
+□
+The next result gives a sufficient criterion for the AdGL(n,H)-real elements in gl(n, H) to be
+strongly AdGL(n,H)-real.
+Theorem 4.4. Let A ∈ gl(n, H) be an arbitrary AdGL(n,H)-real element. Suppose that in the
+Jordan decomposition (2.7) of A, every Jordan block corresponding to a non-zero eigenvalue
+class with purely imaginary complex representative has even multiplicity. Then A is strongly
+AdGL(n,H)-real.
+Proof. In view of Theorem 4.1, the proof follows from Lemma 3.1, Lemma 3.2 and Lemma 3.3.
+□
+
+8
+K. GONGOPADHYAY, T. LOHAN AND C. MAITY
+5. Reversibility in GL(n, D)
+As before, D := R, C or H. Here, we will consider reversibility in the Lie group GL(n, D). The
+classification of reversible elements in GL(n, C) is given in [FS, Theorem 4.2]. We have included
+the case D = C here as it will be used in Proposition 5.2.
+Theorem 5.1. Let D := R, C or H. An element A ∈ GL(n, D) with Jordan form as given in
+Lemma 2.4 is reversible if and only if the following hold:
+(1) For D = R, the blocks can be partitioned into pairs {JR(µ±iν, 2t), JR(
+µ
+µ2+ν2 ∓ i
+ν
+µ2+ν2, 2t)},
+{J(λ, s), J(λ−1, s)} or, singletons {JR(α±iβ, 2ℓ)}, {J(γ, m)}, where λ, µ, ν ∈ R such that
+λ, γ ̸= 0, ν, β > 0 and λ ̸= ±1, µ2 + ν2 ̸= 1, γ = ±1, α2 + β2 = 1.
+(2) For D = C, the blocks can be partitioned into pairs {J(λ, s), J(λ−1, s)} or, singletons
+{J(µ, m)}, where λ, µ ∈ C \ {0} and λ ̸= ±1, µ = ±1.
+(3) For D = H, the blocks can be partitioned into pairs {J(λ, s), J(λ−1, s)} or, singletons
+{J(µ, m)}, where λ, µ ∈ C \ {0} with non-negative imaginary parts such that |λ| ̸=
+1, |µ| = 1.
+Proof. In the case of D = H, for a unique complex representative λ ∈ C of an eigenvalue
+class of A, [λ] = [λ−1] if and only if |λ| = 1, i.e., λ−1 = λ. Using Lemma 2.4, A is conjugate
+to A−1 if and only if A and A−1 has same Jordan decomposition. Now the proof of (1) and (3)
+follows from the same line of arguments as done in Theorem 4.1. For the proof of (2), we refer
+to [FS, Theorem 4.2].
+□
+In GL(n, D), every reversible element is strongly reversible for the case D = R or C.
+Proposition 5.2 (cf. [FS, Theorems 4.7]). Let A ∈ GL(n, D), where D = R or C. Then A is
+reversible in GL(n, D) if and only if A is strongly reversible in GL(n, D).
+Proof. In view of Theorem 5.1, the proof follows from Lemma 3.6, Lemma 3.7, Lemma 3.9,
+and Lemma 3.10.
+□
+The next example shows that Proposition 5.2 does not hold in the case D = H.
+Example 5.3. Let A := (i) ∈ GL(1, H). Then A is reversible in GL(n, H) but not strongly
+reversible.
+□
+The next result gives a sufficient criterion for the reversible elements in GL(n, H) to be strongly
+reversible.
+Theorem 5.4. Let A ∈ GL(n, H) be an arbitrary reversible element.
+Suppose that in the
+Jordan decomposition of A, every Jordan block corresponding to non-real eigenvalue classes of
+unit modulus has even multiplicity. Then A is strongly reversible in GL(n, H).
+Proof. In view of Theorem 5.1, the proof follows from Lemma 3.6, Lemma 3.7 and Lemma 3.8.
+□
+References
+[A’C]
+Norbert A’Campo, Monodromy of real isolated singularities. Topology, 42, no. 6, 2003, 1229–1240.
+[AA]
+V. I. Arnol′d, A. Avez, Ergodic problems of classical mechanics, Translated from the French by A. Avez
+W. A. Benjamin, Inc., New York–Amsterdam, 1968, ix+286 pp.
+[BG]
+S. Bhunia, K. Gongopadhyay, Reversible quaternionic hyperbolic isometries, Linear Algebra Appl., 591,
+2020, 268–283.
+[CM]
+D. H. Collingwood and W. M. McGovern, Nilpotent orbits in semisimple Lie algebras, Van Nostrand
+Reinhold Mathematics Series, Van Nostrand Reinhold Co., New York, 1993.
+
+REVERSIBILITY IN GL(n, D)
+9
+[De]
+R. L. Devaney, Reversible diffeomorphisms and flows, Trans. Amer. Math. Soc., 218, 1976, 89–113.
+[FS]
+A. G. O’Farrell, I. Short, Reversibility in Dynamics and Group Theory, London Mathematical Society
+Lecture Note Series, vol.416, Cambridge University Press, Cambridge, 2015.
+[GL]
+K. Gongopadhyay, T. Lohan, Reversibility of Hermitian isometries,
+Linear Algebra Appl., 639, 2022,
+159–176.
+[GLM]
+K. Gongopadhyay, T. Lohan, C. Maity, Real adjoint orbits of special linear groups, preprint,
+arXiv:2204.03624.
+[GLR]
+I. Gohberg, P. Lancaster, and L. Rodman., Invariant Subspaces of Matrices with Applications, John
+Wiley, New York, 1986; republication SIAM, Philadelphia, 2006.
+[GM]
+K.
+Gongopadhyay,
+C.
+Maity,
+Reality
+of
+unipotent
+elements
+in
+simple
+Lie
+groups,
+preprint,
+arXiv:2101.02732.
+[La]
+J. S. W. Lamb, Reversing symmetries in dynamical systems. J. Phys. A, 25, no. 4, 1992, 925–937 pp..
+[LR]
+J.S.W. Lamb and J. A. G. Roberts, Time-reversal symmetry in dynamical systems: a survey. Time-
+reversal symmetry in dynamical systems (Coventry, 1996). Phys. D, 112, no. 1-2, 1998, 1–39.
+[Ro]
+L. Rodman, Topics in quaternion linear algebra. Princeton Series in Applied Mathematics. Princeton
+University Press, Princeton, NJ, 2014.
+[Se]
+M.B. Sevryuk, Reversible systems, Lecture Notes in Mathematics, 1211. Springer-Verlag, Berlin, 1986,
+vi+319 pp.
+[Wo]
+M. J. Wonenburger, Transformations which are products of two involutions, J. Math. Mech, 16, 1966,
+327–338.
+Indian Institute of Science Education and Research (IISER) Mohali, Knowledge City, Sector
+81, S.A.S. Nagar 140306, Punjab, India
+Email address: krishnendug@gmail.com, krishnendu@iisermohali.ac.in
+Indian Institute of Science Education and Research (IISER) Mohali, Knowledge City, Sector
+81, S.A.S. Nagar 140306, Punjab, India
+Email address: tejbirlohan70@gmail.com, ph18028@iisermohali.ac.in
+Indian Institute of Science Education and Research (IISER) Mohali, Knowledge City, Sector
+81, S.A.S. Nagar 140306, Punjab, India
+Email address: maity.chandan1@gmail.com , cmaity@iisermohali.ac.in
+
diff --git a/p9FKT4oBgHgl3EQfzC7B/content/tmp_files/load_file.txt b/p9FKT4oBgHgl3EQfzC7B/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..a2e9744e1aeb74abed6b44e7c65f34ac8c4b1013
--- /dev/null
+++ b/p9FKT4oBgHgl3EQfzC7B/content/tmp_files/load_file.txt
@@ -0,0 +1,614 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf,len=613
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='11910v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='GR] 27 Jan 2023 REVERSIBILITY AND REAL ADJOINT ORBITS OF LINEAR MAPS KRISHNENDU GONGOPADHYAY, TEJBIR LOHAN AND CHANDAN MAITY Dedicated to the 80th birthday of Norbert A’Campo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We extend classical results on the classification of reversible elements of the group GL(n, C) (and GL(n, R)) to GL(n, H) using an infinitesimal version of the classical reversibility, namely adjoint reality in the Lie algebra set-up.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We also provide a new proof of such a classi- fication for the general linear groups over R and C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Further, we classify the real adjoint orbits in the Lie algebra gl(n, D) for D = R, C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Introduction Reversing and time-reversing symmetries are important classes of symmetries that appear in the natural science.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Especially they arise in many physically motivated dynamical systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' There is an extensive literature discussing such symmetries in different areas of physics and dy- namical systems, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [La], [LR].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In a mathematical terminology, the motions of a dynamical system may be associated with a group G, and such symmetries correspond to the “reversible” and “strongly reversible” elements in G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An element g in a group G is called reversible if g and g−1 are conjugate in G , that is, if there exists h ∈ G such that hgh−1 = g−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An element g in a group G is called strongly reversible if g is conjugate to g−1 by an involution (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', an element of order at most two) in G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Equivalently, strongly reversible elements are products of two involutions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Some authors have called strongly reversible elements “bireflectional”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Reversible maps have appeared from different perspectives in the literature, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [AA], [De], [Se], [FS].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' From a group-theoretical perspective, a classical theorem of Frobenius and Schur asserts that the number of real-valued complex irreducible characters of a finite simple group G is equal to the number of reversible conjugacy classes of G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' With this motivation, many mathematicians have used the terminology “real” and “strongly real” elements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A strongly reversible element is reversible, but the converse is not true in general.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' It is a problem of potential interest to classify reversible and strongly reversible elements in different groups of interest.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In the theory of finite groups, the classification of such elements is relatively well-understood in the literature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' However, a complete classification of reversible classes is not available other than for a few families of infinite groups.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Some of the infinite groups where it has been classified include compact Lie groups, real rank one classical groups, and isometry groups of Hermitian spaces, see e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [FS, BG, GL].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The idea of reversibility is apparent in the work of A’Campo [A’C] as well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A’Campo investi- gated the monodromy of real isolated singularities using the fact that the complex conjugation on complex space permutes the level sets of a real polynomial function and induces involu- tions on level sets corresponding to real values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In particular, it was proved that the geometric monodromy is the composition of the involution induced by complex conjugation and another involution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In other words, the geometric monodromies are strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A’Campo called 2020 Mathematics Subject Classification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Primary 20E45;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Secondary 15B33, 22E60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Key words and phrases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' General linear group, reversibility, adjoint reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 1 2 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' GONGOPADHYAY, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' LOHAN AND C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' MAITY the corresponding singularity strongly invertible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Further, it follows that any two geometric monodromies are “linked” by the involution that comes from complex conjugation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recently, the concept of reversibility has been extended to semisimple Lie algebras using the adjoint representations of Lie groups in [GM].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The infinitesimal notion that has been introduced for the Lie algebras is called adjoint reality.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Understanding the adjoint orbits of a semisimple Lie group is an active research theme, see the survey [CM].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' However, the exploration of adjoint reality properties has been the object of attention only very recently, cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [GM], [GLM].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' As an application of adjoint reality for nilpotent orbits in simple Lie algebras, the reversible and strongly reversible unipotent elements in the classical simple Lie groups have been completely classified, see [GM].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let D = R, C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In this chapter, we revisit the classification of reversible elements in the general linear group GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The classification of reversible elements in GL(n, D) and their equivalence with the strongly reversible elements are well known in the literature for D = R or C, cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [Wo], [FS].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Extending these results over the quaternions is not straightforward due to the non-commutativity of H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We shall overcome such difficulties by approaching this problem using adjoint reality in the Lie algebra gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We recall the notion of adjoint reality below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider the adjoint action of the general linear group G := GL(n, D) on its Lie algebra g := gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recall that gl(n, D) ≃ M(n, D), the algebra of n × n matrices over D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In this case, the adjoint action is given by the conjugation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider the conjugacy action of GL(n, D) on gl(n, D): Ad(g)X := gXg−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An element X ∈ g is called AdG-real if −X = gXg−1 for some g ∈ G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An AdG-real element X ∈ g is called strongly AdG-real if −X = τXτ −1 for some involution (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', element of order at most two) τ ∈ G;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' see [GM, Definition 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Observe that if X ∈ g is AdG-real, then exp(X) is reversible in G, but the converse may not be true.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We classify the adjoint real elements in gl(n, D) and investigate their equivalence with the strongly adjoint real elements in gl(n, D);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' see Theorems 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2, 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Using these ideas and applying some of these results, we shall prove that some particular types of Jordan forms in GL(n, D) are strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This will be used to classify the reversible and strongly re- versible elements in GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This approach not only reduces the complexity of the computations but also gives a better understanding of reversibility in the group GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This also provides a uniform treatment over D = R, C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We show by a counterexample in Example 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3 that, unlike the field case, the notion of reversible and strongly reversible elements are not equivalent in GL(n, H) in general.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We give a sufficient criterion for equivalence of the two notions in GL(n, H), see Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The chapter is organized as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In §2, we fix some notation and recall some background related to Jordan canonical forms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In §3, reversibility and adjoint reality of certain Jordan forms are described.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In §4, we deal with the adjoint reality in the Lie algebra gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We revisit the reversibility problem of GL(n, D) in §5 and provide a much simpler proof of earlier obtained results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Acknowledgement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' It has been a highly rewarding experience to know Norbert A’Campo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Interaction with him has always been rich in mathematical and non-mathematical ideas!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' It is a pleasure to dedicate this small contribution to the volume in honor of Norbert.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We wish Norbert a very good health and a happy life ahead.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Gongopadhyay is partially supported by the SERB core research grant CRG/2022/003680.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lohan acknowledges support from the CSIR SRF grant, File No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' : 09/947(0113)/2019-EMR-I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Maity is supported by an NBHM PDF during this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' REVERSIBILITY IN GL(n, D) 3 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Preliminaries In this section, we will recall some necessary background.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recall that H := R + Ri+ Rj+ Rk denotes the division algebra of Hamilton’s quaternions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We consider Hn as a right H-module.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We refer to [Ro] for a nice exposition on quaternion linear algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A ∈ M(n, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A non-zero vector v ∈ Hn is said to be a right eigenvector of A corresponding to a right eigenvalue λ ∈ H if the equality Aλ = vλ holds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Eigenvalues of A ∈ M(n, H) occur in similarity classes, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', if v is an eigenvector corresponding to λ, then vµ ∈ vH is an eigenvector corresponding to µ−1λµ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Each similarity class of eigen- values contains a unique complex number with a non-negative imaginary part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Here, instead of similarity classes of eigenvalues, we will consider the unique complex representative with a non-negative imaginary part.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let ψ: C −→ M(2, R) be the embedding given by ψ(z) := � Re(z) Im(z) −Im(z) Re(z) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This induces the embedding Ψ: M(n, C) −→ M(2n, R) defined as Ψ((zi,j)n×n) := � ψ(zi,j) � 2n×2n .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1) It follows from the definition of the exponential map exp : M(n, C) −→ GL(n, C) that Ψ(exp(X)) = exp(Ψ(X)) ∀ X ∈ M(n, C).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2) Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2 (cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [Ro, p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 94]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A Jordan block J(λ, m) is an m × m matrix with λ ∈ D on the diagonal entries, 1 on all of the super-diagonal entries and zero elsewhere.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We will refer to a block diagonal matrix where each block is a Jordan block as a Jordan form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We also consider the following block matrix as a Jordan form over R, cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [GLR, p.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 364], which corresponds to the case when the eigenvalues of a matrix over R belong to C \\ R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recall the embedding Ψ as in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let K := Ψ(µ + iν) = � µ ν −ν µ � ∈ M(2, R), where µ, ν are real numbers with ν > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then define JR(µ ± iν, 2n) := Ψ(J(µ + iν, n)) = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ec \uf8ec \uf8ed K I2 K I2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' K I2 K \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f7 \uf8f8 ∈ M(2n, R), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3) where I2 denotes the 2×2 identity matrix;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' see [Ro, Theorem 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1], [GLR, Chapter 12].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Further, we also define JR(µ ∓ iν, 2n) := Ψ(J(µ − iν, n)) = σ � Ψ(J(µ + iν, n)) � σ−1, (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4) where σ = diag(1, −1, 1, −1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)2n−1 )2n×2n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Remark 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We will follow the notation JR(µ±iν, n) and JR(µ∓iν, n) as defined in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3) and (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4) throughout this chapter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Note that {JR(µ ± iν, n)} is a singleton by the above definition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Next we recall the Jordan form in M(n, D), see [Ro, Theorem 15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4 (Jordan form in M(n, D), cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [Ro]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' For every A ∈ M(n, D), there is an invertible matrix S ∈ GL(n, D) such that SAS−1 has the following form: 4 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' GONGOPADHYAY, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' LOHAN AND C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' MAITY (1) For D = R, SAS−1 = J(λ1, m1) ⊕ · · · ⊕ J(λk, mk) � JR(µ1 ± iν1, 2ℓ1) ⊕ · · · ⊕ JR(µq ± iνq, 2ℓq), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5) where λ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , λk;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' µ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , µq ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' ν1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , νq are (not necessarily distinct) real numbers and ν1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , νq are positive.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (2) For D = C, SAS−1 = J(λ1, m1) ⊕ · · · ⊕ J(λk, mk), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6) where λ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , λk are (not necessarily distinct) complex numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3) For D = H, SAS−1 = J(λ1, m1) ⊕ · · · ⊕ J(λk, mk), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7) where λ1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , λk are (not necessarily distinct) complex numbers and have non-negative imaginary parts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The forms (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5), (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6) and (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7) are uniquely determined by A up to a permutation of Jordan blocks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Strong Reversibility of Jordan forms In this section, we investigate the adjoint reality in gl(n, D) and reversibility in GL(n, D) for certain types of Jordan forms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Strong adjoint reality of Jordan forms in gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Here, we will consider some par- ticular types of Jordan forms in gl(n, D) and show that they are strongly AdGL(n,D)-real by explicitly constructing a suitable reversing involution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let D = R, C or H, and X := J(0, n) be the nilpotent element in gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then X is strongly AdGL(n,D)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let g := diag(1, −1, 1, −1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1 )n×n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then g is an involution in GL(n, D) such that gXg−1 = −X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let X := J(λ, n) ⊕ J(−λ, n) be the Jordan form in gl(2n, D), where λ ∈ D \\ {0}, for D = R or C, and for D = H, λ ∈ C \\ {0} with non-negative imaginary part such that the real part of λ ̸= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then X is strongly AdGL(2n,D)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Write X = � J(λ, n) J(−λ, n) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let τ := diag(1, −1, 1, −1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1 )n×n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Note that J(λ, n) τ = −τ J(−λ, n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider g = � τ τ � ∈ GL(2n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then g is an involution in GL(2n, D) such that gXg−1 = −X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let X := J(µi, n) ⊕ J(µi, n) be the Jordan form in gl(2n, H), where µ ∈ R, µ > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then X is strongly AdGL(2n,H)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let τ := diag(j, −j, j, −j, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1j)n×n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then τ 2 = −In and τ J(µi, n) τ −1 = −J(µi, n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider the involution g = � τ −τ � in GL(2n, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then g is an involution in GL(2n, H) such that gXg−1 = −X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This proves the lemma.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Recall that the matrix JR(µ ± iν, 2n) is defined in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let X := JR(0 ± iν, 2n) be the Jordan block in gl(2n, R), where ν ∈ R such that ν > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then X is strongly AdGL(2n,R)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' REVERSIBILITY IN GL(n, D) 5 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let g := diag(I1,1, −I1,1, I1,1, −I1,1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1I1,1 )2n×2n, where I1,1 := � 1 0 0 −1 � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then g is an involution in GL(2n, R) such that gXg−1 = −X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ We refer to (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4) for the notation JR(µ ∓ iν, 2n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let X := JR(µ ± iν, 2n) ⊕ JR(−µ ∓ iν, 2n) be the Jordan form gl(4n, R), where µ, ν ∈ R such that µ ̸= 0 and ν > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then X is strongly AdGL(4n,R)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Write X = � JR(µ ± iν, 2n) JR(−µ ∓ iν, 2n) � .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider g = � τ τ � , where τ := diag(I2, −I2, I2, −I2, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1I2 )2n×2n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Observe that JR(µ ± iν, 2n) τ = −τ JR(−µ ∓ iν, 2n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then g is an involution in GL(4n, R) such that gXg−1 = −X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Strong reversibility of Jordan forms in GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We will apply the results obtained in §3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1 to provide a different proof of strong reversibility of some particular types of Jordan forms in GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Such results are known for D = R, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We shall extend these results over H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' These results will be used in the Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2 and the Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let D = R, C or H, and A := J(µ, n) be the Jordan block in GL(n, D), where µ ∈ {±1}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is strongly reversible in GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' First, we will consider µ = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then J(1, n) = In + J(0, n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' By setting N := J(0, n), we have gNg−1 = −N, g2 = In, where g is as in Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This implies g eNg−1 = e−N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Since the Jordan form of eN is J(1, n), J(1, n) = τeNτ −1 for some τ ∈ GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Now (τgτ −1)J(1, n)(τg−1τ −1) = τgeNg−1τ −1 = τe−Nτ −1 = (J(1, n))−1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Hence, J(1, n) is strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The case when µ = −1 follows from the fact that −J(−1, n) has Jordan form J(1, n) and J(1, n) is strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A := J(λ, n) ⊕ J(λ−1, n) be the Jordan form in GL(2n, D), where λ ∈ D \\ {±1, 0}, for D = R or C, and for D = H, λ ∈ C \\ {0} with non-negative imaginary part such that |λ| ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is strongly reversible in GL(2n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let µ ∈ C such that eµ = λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Note that exp(J(µ, n)) = exp(µIn) · exp(J(0, n)) = λ exp(J(0, n)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let P ∈ GL(n, D) so that λ exp(J(0, n)) = PJ(λ, n) P −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Thus exp(J(µ, n)) = PJ(λ, n) P −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1) Similarly, there exists Q ∈ GL(n, D) such that exp(J(−µ, n)) = Q J(λ−1, n) Q−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2) Now, let σ := diag(1, −1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1) be an involution in GL(n, D) so that −J(−µ, n) = σ J(µ, n) σ−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This implies σ exp(J(µ, n)) σ−1 = � exp(J(−µ, n)) �−1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3) Using (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1), (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2) and (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3), we have σ P J(λ, n) P −1 σ−1 = � Q J(λ−1, n) Q−1�−1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4) Consider the involution g := � P Q �−1 � σ σ−1 � � P Q � in GL(2n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then gAg−1 = A−1 if and only if (Q−1σ P) J(λ, n) (Q−1σ P)−1 = � J(λ−1, n) �−1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5) Now the proof follows from Equation (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ 6 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' GONGOPADHYAY, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' LOHAN AND C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' MAITY Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A := J(µ, n) ⊕ J(µ, n) be the Jordan block in GL(2n, H), where µ ∈ C\\{±1}, with non-negative imaginary part such that |µ| = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is strongly reversible in GL(2n, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recall that j Z = Z j for all Z ∈ GL(n, C), where Z is the matrix obtained by taking the conjugate of each entry of the complex matrix Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We can assume that A = J(eiθ, n) ⊕ J(eiθ, n), where θ ∈ (0, π).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Write A = � P P � , where P = J(eiθ, n) ∈ GL(n, C).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' To show A is strongly reversible, it is sufficient to find a g = � X X−1 � ∈ GL(2n, H) such that P −1X = XP, where X ∈ GL(n, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Further, If X = Y j for some Y ∈ GL(n, C), then we require P −1Y = Y P, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', � J(eiθ, n) �−1 Y = Y J(e−iθ, n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Using the construction as done in the proof of Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7, we can find such Y in GL(n, C), see (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A := JR(µ ± iν, 2n) be the Jordan block in GL(2n, R) as in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' If µ2 +ν2 = 1, then A is strongly reversible in GL(2n, R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let K := � µ ν −ν µ � where µ2 +ν2 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then K ∈ SO(2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let Y = � 0 a −a 0 � ∈ so(2) such that exp(Y ) = K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Note that for I1,1 := diag(1, −1), we have I1,1 Y (I1,1)−1 = −Y .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider X := JR(0 ± ia, 2n) = \uf8eb \uf8ec \uf8ec \uf8ec \uf8ed Y I2 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Y I2 Y \uf8f6 \uf8f7 \uf8f7 \uf8f7 \uf8f8 ∈ GL(2n, R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let g := diag(I1,1, −I1,1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , (−1)n−1I1,1) ∈ GL(2n, R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then gXg−1 = −X, and hence exp(X) is strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' To conclude the proof, it is sufficient to show that exp(X) and JR(µ ± iν, 2n) are conjugate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' To see this recall that Ψ(J(µ+iν, n)) = JR(µ±iν, 2n), and Ψ(J(ia, n)) = X, where the map Ψ is as in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Since exp(Y ) = K, we have eia = µ + iν.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then exp(J(ia, n)) = (µ + iν) exp(J(0, n)) = h J(µ + iν, n) h−1 , (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6) for some h ∈ GL(n, C).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' By using Equations (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2) and (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6), we have exp(X) = Ψ(exp(J(ia, n)) = h � Ψ(J(µ + iν, n)) � h−1 = h JR(µ ± iν, 2n) h−1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Therefore, JR(µ ± iν, 2n) is strongly reversible in GL(2n, R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A := JR(µ ± iν, 2n) ⊕ JR � µ µ2+ν2 ∓ i ν µ2+ν2, 2n � be the Jordan form in GL(4n, R), where µ2 + ν2 ̸= 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is strongly reversible in GL(4n, R).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let z ∈ C such that ez = µ+iν and e−z = µ µ2+ν2 −i ν µ2+ν2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recall that the embedding Ψ is as in (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Using (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3) and (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4), we have Ψ(J(ez, n)) = JR(µ ± iν, 2n) and Ψ(J(e−z, n)) = JR( µ µ2 + ν2 ∓ i ν µ2 + ν2 , 2n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Therefore, we can write A as A = Ψ(P), where P = � J(ez, n) J(e−z, n) � ∈ GL(2n, C).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7) The proof now follows from Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7 and the fact that Ψ is an embedding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ REVERSIBILITY IN GL(n, D) 7 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Adjoint reality in gl(n, D) In this section, we investigate adjoint reality in gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' First we classify AdGL(n,D)-real elements in the Lie algebra gl(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let D = R, C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An element X ∈ gl(n, D) with Jordan canonical form as given in Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4 is AdGL(n,D)-real if and only if the following hold: (1) For D = R, the blocks can be partitioned into pairs {JR(µ ± iν, 2t), JR(−µ ∓ iν, 2t)}, {J(λ, s), J(−λ, s)} or, singletons {JR(0±iν, 2ℓ)}, {J(0, m)}, where λ, µ, ν ∈ R and λ, µ ̸= 0, ν > 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (2) For D = C, the blocks can be partitioned into pairs {J(λ, s), J(−λ, s)} or, singletons {J(0, m)}, where λ ∈ C and λ ̸= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3) For D = H, the blocks can be partitioned into pairs {J(λ, s), J(−λ, s)} or, singletons {J(µ, m)}, where λ, µ ∈ C with non-negative imaginary parts such that real part of λ ̸= 0 and real part of µ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Consider the case D = R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Using Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4, X is conjugate to −X if and only if {J(λ1, m1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , J(λk, mk), JR(µ1 ± iν1, 2ℓ1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , JR(µq ± iνq, 2ℓq)} = {J(−λ1, m1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , J(−λk, mk), JR(−µ1 ∓ iν1, 2ℓ1), .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' , JR(−µq ∓ iνq, 2ℓq)}, and the result follows immediately for the case D = R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Recall the fact that for a unique complex representative λ of an eigenvalue class of X, [λ] = [−λ] if and only if the real part of λ = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Using the same line of argument as we used in the D = R case, the result follows for the case D = C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ Recall that every strongly AdGL(n,D)-real element in gl(n, D) is AdGL(n,D)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In the next result, we will prove that the converse holds when D = R or C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let D = R or C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An element A of gl(n, D) is AdGL(n,D)-real if and only if it is strongly AdGL(n,D)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Without loss of generality, we can assume that A is in Jordan form as in Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Suppose A is AdGL(n,D)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then Jordan blocks of A can be partitioned as in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In view of Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4, and Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='5, we can explicitly construct an involution g in GL(n, D) such that gAg−1 = −A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Since the converse is trivial, this completes the proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ The following example shows that the above result does not hold for gl(n, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Example 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A := (i) ∈ gl(1, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then gAg−1 = −A, where g = (j) ∈ GL(1, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' So A is AdGL(1,H)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Suppose that A is strongly AdGL(1,H)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let g = (a) ∈ GL(1, H) be an involution such that gAg−1 = −A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then we get ai = −ia.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This implies a = wj for some w ∈ C, w ̸= 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Since g is an involution, a2 = (wj)2 = 1, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='e, |w|2 = −1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' This is a contradiction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Therefore, A is AdGL(1,H)-real but not strongly AdGL(1,H)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ The next result gives a sufficient criterion for the AdGL(n,H)-real elements in gl(n, H) to be strongly AdGL(n,H)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A ∈ gl(n, H) be an arbitrary AdGL(n,H)-real element.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Suppose that in the Jordan decomposition (2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7) of A, every Jordan block corresponding to a non-zero eigenvalue class with purely imaginary complex representative has even multiplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is strongly AdGL(n,H)-real.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In view of Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, the proof follows from Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2 and Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ 8 K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' GONGOPADHYAY, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' LOHAN AND C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' MAITY 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Reversibility in GL(n, D) As before, D := R, C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Here, we will consider reversibility in the Lie group GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' The classification of reversible elements in GL(n, C) is given in [FS, Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' We have included the case D = C here as it will be used in Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let D := R, C or H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' An element A ∈ GL(n, D) with Jordan form as given in Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4 is reversible if and only if the following hold: (1) For D = R, the blocks can be partitioned into pairs {JR(µ±iν, 2t), JR( µ µ2+ν2 ∓ i ν µ2+ν2, 2t)}, {J(λ, s), J(λ−1, s)} or, singletons {JR(α±iβ, 2ℓ)}, {J(γ, m)}, where λ, µ, ν ∈ R such that λ, γ ̸= 0, ν, β > 0 and λ ̸= ±1, µ2 + ν2 ̸= 1, γ = ±1, α2 + β2 = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (2) For D = C, the blocks can be partitioned into pairs {J(λ, s), J(λ−1, s)} or, singletons {J(µ, m)}, where λ, µ ∈ C \\ {0} and λ ̸= ±1, µ = ±1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' (3) For D = H, the blocks can be partitioned into pairs {J(λ, s), J(λ−1, s)} or, singletons {J(µ, m)}, where λ, µ ∈ C \\ {0} with non-negative imaginary parts such that |λ| ̸= 1, |µ| = 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In the case of D = H, for a unique complex representative λ ∈ C of an eigenvalue class of A, [λ] = [λ−1] if and only if |λ| = 1, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', λ−1 = λ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Using Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4, A is conjugate to A−1 if and only if A and A−1 has same Jordan decomposition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Now the proof of (1) and (3) follows from the same line of arguments as done in Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' For the proof of (2), we refer to [FS, Theorem 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ In GL(n, D), every reversible element is strongly reversible for the case D = R or C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2 (cf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [FS, Theorems 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A ∈ GL(n, D), where D = R or C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is reversible in GL(n, D) if and only if A is strongly reversible in GL(n, D).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In view of Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, the proof follows from Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='9, and Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ The next example shows that Proposition 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='2 does not hold in the case D = H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Example 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A := (i) ∈ GL(1, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is reversible in GL(n, H) but not strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ The next result gives a sufficient criterion for the reversible elements in GL(n, H) to be strongly reversible.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Let A ∈ GL(n, H) be an arbitrary reversible element.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Suppose that in the Jordan decomposition of A, every Jordan block corresponding to non-real eigenvalue classes of unit modulus has even multiplicity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Then A is strongly reversible in GL(n, H).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' In view of Theorem 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='1, the proof follows from Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='6, Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='7 and Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' □ References [A’C] Norbert A’Campo, Monodromy of real isolated singularities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Topology, 42, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 6, 2003, 1229–1240.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [AA] V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Arnol′d, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Avez, Ergodic problems of classical mechanics, Translated from the French by A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Avez W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Benjamin, Inc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', New York–Amsterdam, 1968, ix+286 pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [BG] S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Bhunia, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Gongopadhyay, Reversible quaternionic hyperbolic isometries, Linear Algebra Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', 591, 2020, 268–283.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [CM] D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Collingwood and W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' McGovern, Nilpotent orbits in semisimple Lie algebras, Van Nostrand Reinhold Mathematics Series, Van Nostrand Reinhold Co.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', New York, 1993.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' REVERSIBILITY IN GL(n, D) 9 [De] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Devaney, Reversible diffeomorphisms and flows, Trans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Amer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Soc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', 218, 1976, 89–113.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [FS] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' O’Farrell, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Short, Reversibility in Dynamics and Group Theory, London Mathematical Society Lecture Note Series, vol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='416, Cambridge University Press, Cambridge, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [GL] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Gongopadhyay, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lohan, Reversibility of Hermitian isometries, Linear Algebra Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', 639, 2022, 159–176.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [GLM] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Gongopadhyay, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lohan, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Maity, Real adjoint orbits of special linear groups, preprint, arXiv:2204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='03624.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [GLR] I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Gohberg, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lancaster, and L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Rodman.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=', Invariant Subspaces of Matrices with Applications, John Wiley, New York, 1986;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' republication SIAM, Philadelphia, 2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [GM] K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Gongopadhyay, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Maity, Reality of unipotent elements in simple Lie groups, preprint, arXiv:2101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='02732.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [La] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lamb, Reversing symmetries in dynamical systems.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A, 25, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 4, 1992, 925–937 pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='. [LR] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Lamb and J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Roberts, Time-reversal symmetry in dynamical systems: a survey.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Time- reversal symmetry in dynamical systems (Coventry, 1996).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Phys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' D, 112, no.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' 1-2, 1998, 1–39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [Ro] L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Rodman, Topics in quaternion linear algebra.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Princeton Series in Applied Mathematics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Princeton University Press, Princeton, NJ, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [Se] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Sevryuk, Reversible systems, Lecture Notes in Mathematics, 1211.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Springer-Verlag, Berlin, 1986, vi+319 pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' [Wo] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Wonenburger, Transformations which are products of two involutions, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Mech, 16, 1966, 327–338.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Indian Institute of Science Education and Research (IISER) Mohali, Knowledge City, Sector 81, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Nagar 140306, Punjab, India Email address: krishnendug@gmail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='com, krishnendu@iisermohali.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='in Indian Institute of Science Education and Research (IISER) Mohali, Knowledge City, Sector 81, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Nagar 140306, Punjab, India Email address: tejbirlohan70@gmail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='com, ph18028@iisermohali.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='in Indian Institute of Science Education and Research (IISER) Mohali, Knowledge City, Sector 81, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content=' Nagar 140306, Punjab, India Email address: maity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='chandan1@gmail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='com , cmaity@iisermohali.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='ac.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
+page_content='in' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/p9FKT4oBgHgl3EQfzC7B/content/2301.11910v1.pdf'}
diff --git a/qNFAT4oBgHgl3EQfex2-/vector_store/index.pkl b/qNFAT4oBgHgl3EQfex2-/vector_store/index.pkl
new file mode 100644
index 0000000000000000000000000000000000000000..09194905ffd164bd9c96b2548c68164efd65c561
--- /dev/null
+++ b/qNFAT4oBgHgl3EQfex2-/vector_store/index.pkl
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d80bb443d7e756f167ca121f8cf2c2aa320a881f162fe6548c6f8da95e7c6ea
+size 118346
diff --git a/qNFJT4oBgHgl3EQfaCwn/content/tmp_files/2301.11534v1.pdf.txt b/qNFJT4oBgHgl3EQfaCwn/content/tmp_files/2301.11534v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..53db75b0e584cd7e98615d2564d2307a435a3b02
--- /dev/null
+++ b/qNFJT4oBgHgl3EQfaCwn/content/tmp_files/2301.11534v1.pdf.txt
@@ -0,0 +1,1029 @@
+arXiv:2301.11534v1 [math.GN] 27 Jan 2023
+Variations of star selection principles on Hyperspaces∗
+JAVIER CASAS-DE LA ROSA
+Abstract
+In this paper we define some combinatorial principles to characterize
+spaces X whose hyperspace satisfies some variation of some classical star
+selection principle. Specifically, the variations characterized are the selec-
+tive and absolute versions of the star selection principles for the Menger
+and Rothberger cases; also, the hyperspaces considered in these charac-
+terizations are CL(X), K(X), F(X) and CS(X) in both cases, endowed
+with either the Fell topology or the Vietoris topology.
+Key words. Hyperspaces, Fell topology, Vietoris topology, star selection princi-
+ples, (absolutely) strongly star-Menger, (selectively) strongly star-Menger, (ab-
+solutely) strongly star-Rothberger, (selectively) strongly star-Rothberger.
+Mathematics Subject Classification: Primary 54B20, 54D20; Secondary 54A05,
+54A25.
+1
+Introduction and preliminaries
+Many branches from Selection Principles Theory have arisen after a systematic
+research made in [17] by Scheepers.
+Nowadays, this theory has connections
+as well as applications to several areas of mathematics as General Topology,
+Function spaces, Hyperspaces, etc. As an example of it, in [9], [10],[15] and [16],
+the authors studied the fundamental problem on Hyperspaces which consists
+in establishing dualities between some topological properties or, as it is in this
+case, between some classical selection principles under different hyperspaces
+topologies. In other words, given two selection principles P and Q, one must
+determine if it holds that a topological space X satisfies the principle P if and
+only if its hyperspace satisfies the principle Q. Hence, this duality problem can
+be viewed as a method to characterize some classical selection principles of X
+with selection principles on different hyperspaces of X, using several well-known
+topologies.
+On the other hand, several versions of the original selection principles have
+been defined from its beginning. One of the most important versions that its
+investigation has rapidly increased is the star versions of the classical selection
+∗The author was supported for this research by Postdoctoral Fellowship Program at
+UNAM.
+1
+
+principles. These star versions were defined by Koˇcinac in [13] and they gave
+rise to the star selection principles theory (see [14] and [12]).
+The study of the duality problem involving some star selection principles on
+hyperspaces initiated very recently in [4] and continued in [3], [6], [7] and [8] have
+dealt with classical star selection principles only. In particular, in these works
+several combinatorial principles have been defined to establish characterizations
+for the (strongly) star-Menger property and the (strongly) star-Rothberger prop-
+erty on several hyperspaces under different topologies. In this paper we continue
+this line of investigation for some variations of some classical star selection prin-
+ciples. In Section 2, the selective and absolute versions of Menger-type and
+Rothberger-type star selection principle on several hyperspaces with the Vi-
+etoris topology are characterized. In Section 3, analogous characterizations are
+given for several hyperspaces with the Fell topology.
+1.1
+Hyperspaces
+All spaces are assumed to be Hausdorff noncompact and, even, nonparacom-
+pact. Given a topological space X , the hyperspace of X, denoted by CL(X),
+is the set of all nonempty closed subsets of X. By K(X) (F(X)), we denote the
+family of all nonempty compact (all nonempty finite) subsets of X. Also, by
+CS(X) we denote the family of all convergent sequences of X.
+We denote by ω the first infinite cardinal and for a set A, [A]<ω denotes the
+set of all finite subsets of A. For a subset U ⊆ X and a family U of subsets of
+X, we write:
+U −
+=
+{A ∈ CL(X) : A ∩ U ̸= ∅};
+U +
+=
+{A ∈ CL(X) : A ⊆ U};
+U c
+=
+X\U;
+Uc
+=
+{U c : U ∈ U};
+U−
+=
+{U − : U ∈ U};
+U+
+=
+{U + : U ∈ U}.
+In the literature there are many topologies that can be defined on CL(X) or
+on a subcollection of it. In this paper we will consider two well-known topolo-
+gies, the Vietoris topology, denoted by V, and the Fell topology, denoted by F.
+A basic open subset of the Fell topology is of the form:
+(
+n�
+i=1
+V −
+i ) ∩ (Kc)+,
+where V1, . . . , Vn are open subsets of X and K is a compact subset of X.
+A basic open subset of the Vietoris topology is a set of the form:
+⟨U1, . . . , Un⟩ = {A ∈ CL(X) : A ⊆
+n
+�
+i=1
+Ui, A ∩ Ui ̸= ∅ for each i ≤ n}
+2
+
+where U1, . . . , Un are open subsets of X, n ∈ ω.
+1.2
+Variations of the classical star selection principles
+For a set A ⊆ X and a collection U of subsets of X, the star of A with respect
+to U, denoted by St(A, U), is the set �{U ∈ U : U ∩ A ̸= ∅}; for A = {x} with
+x ∈ X, we write St(x, U) instead of St({x}, U).
+In recent years, different versions of the classical star selection principles
+have been defined and studied in several articles (see for instance [1], [5] and
+[2]). Some of these new star selection principles involve dense subsets of the
+space and they are called as the absolute and selective versions of classical star
+selection principles (see [14] for more information about the absolute versions
+and see [2] for an overview about the selective versions). The selective versions
+are stronger than the absolute versions and the absolute versions are stronger
+than the classical star selection principles1. The following properties are the
+absolute versions of the strongly star selection principles for the Menger and
+Rothberger cases.
+Definition 1.1 ([5]). We say that a space X is:
+1. absolutely strongly star-Menger (aSSM) if for each sequence {Un : n ∈ ω}
+of open covers of X and each dense subset D of X, there is a sequence
+{Fn : n ∈ ω} of finite sets such that Fn ⊆ D, n ∈ ω, and {St(Fn, Un) :
+n ∈ ω} is an open cover of X.
+2. absolutely strongly star-Rothberger (aSSR) if for each sequence {Un : n ∈
+ω} of open covers of X and each dense subset D of X, there is a sequence
+{xn : n ∈ ω} of points of X such that xn ∈ D, n ∈ ω, and {St(xn, Un) :
+n ∈ ω} is an open cover of X.
+Now, we recall the selective versions of the strongly star selection principles
+for the Menger and Rothberger cases2.
+Definition 1.2 ([2]). We say that a space X is:
+1. selectively strongly star-Menger (selSSM) if for each sequence {Un : n ∈
+ω} of open covers of X and each sequence {Dn : n ∈ ω} of dense sets of
+X, there exists a sequence {Fn : n ∈ ω} of finite sets such that Fn ⊆ Dn,
+n ∈ ω, and {St(Fn, Un) : n ∈ ω} is an open cover of X.
+2. selectively strongly star-Rothberger (selSSR) if for each sequence {Un :
+n ∈ ω} of open covers of X and each sequence {Dn : n ∈ ω} of dense
+sets of X, there exists a sequence {xn : n ∈ ω} of points of X such that
+xn ∈ Dn, n ∈ ω, and {St(xn, Un) : n ∈ ω} is an open cover of X.
+1In [5], the authors introduced the absolute versions of classical star selection principles
+in a general form with a different notation than the one used in [2], where it was defined the
+absolute and selective versions of star selection principles, also given in a general form.
+2The Hurewicz case and some other interesting properties are also given in [2]
+3
+
+2
+Absolute and selective versions on hyperspaces
+with the Vietoris topology
+In this section we present some technical principles that are useful to charac-
+terize some variations of classical star selection principles (selSSM, selSSR,
+aSSM and aSSR) on several hiperspaces with the Vietoris topology. To that
+end, we recall some notation and useful definitions to establish these character-
+izations.
+Using the notation of [16], ζ denotes the family:
+ζ = {(V1, . . . , Vn) : V1, . . . , Vn are open subsets of X,
+n ∈ N}.
+In [16], Li defined the notion of πV -networks in a space X as follows:
+A family ζ is called a πV -network of X if for each open subset U of X, with
+U ̸= X, there exist (V1, . . . , Vn) ∈ ζ and a finite set F with F ∩Vi ̸= ∅ (1 ≤ i ≤ n)
+such that �n
+i=1 V c
+i ⊆ U and F ∩U = ∅. The collection of πV -networks of a space
+X is denoted by ΠV .
+Henceforth, ∆ will denote a subset of CL(X) which is closed under finite
+unions and contains all singletons. Using a family ∆, in [8], it was defined a
+modification of πV -network which is called as πV (∆)-network of X.
+Definition 2.1 ([8]). A family ζ is called a πV (∆)-network of X, if for each
+U ∈ ∆c, there exist a (V1, . . . , Vn) ∈ ζ and F ∈ [X]<ω with F ∩ Vi ̸= ∅ (1 ≤ i ≤
+n) such that �n
+i=1 V c
+i ⊂ U and F ∩U = ∅. The collection of all πV (∆)-networks
+of X is denoted by ΠV (∆).
+As pointed out in [8], if we consider ∆ to be CL(X), then the collections
+ΠV (∆) and ΠV coincide. But this fact not need be true in general. In [8], the
+authors gave an example of a πV (∆)-network that is not a πV -network (on a
+metrizable space X and a specific family ∆). Here we present another example
+in a non-metrizable space1.
+Example 2.2. There exists a family ζ on a certain space X and there exists a
+family ∆ ⊆ CL(X) such that ζ is a πV (∆)-network of X but is not a πV -network
+of X.
+Proof. Let A be an uncountable almost disjoint family on ω with ω = � A. We
+consider the Mr´owka-Isbell space (see [11]), X = Ψ(A) and ∆ = K(X). Let
+ζ = {(V1, . . . , Vn) : Vi = {Ai} ∪ (Ai \ Di) (1 ≤ i ≤ n) where A1, . . . , An ∈ A,
+D1 . . . , Dn ∈ [ω]<ω, n ∈ N}
+1This space was also used in [4] to show that πF (∆)-network and πF -network are different
+notions; see Section 3 for definitions.
+4
+
+Note that ζ is properly defined, that is, for each i ∈ {1, . . . , n}, {Ai} ∪ (Ai \ Di)
+is an open set of X.
+Claim 1: ζ is a πV (K(X))-network of X.
+Let U ∈ [K(X)]c. Suppose that U = X \ K0 for some K0 ∈ K(X). We consider
+the following two cases:
+Case I: K0 is infinite.
+Note that in this case, K0 ∩ A is a nonempty finite set.
+Moreover, the set
+(K0 ∩ ω) \ (�{B : B ∈ K0 ∩ A}) is finite. For each m ∈ (K0 ∩ ω) \ (�{B : B ∈
+K0∩A}), let Bm ∈ A be such that m ∈ Bm. The collection B = (K0∩A)∪{Bm :
+m ∈ (K0 ∩ ω) \ (�{B : B ∈ K0 ∩ A})} is finite. We enumerate this collection
+as B = {A1, . . . , An}.
+We define, for each i ∈ {1, . . ., n}, Vi = {Ai} ∪ Ai.
+Then, each Vi is an open set of X and therefore, (V1, . . . , Vn) ∈ ζ. Note that
+K0 ⊆ �n
+i=1 Vi. Thus, �n
+i=1 V c
+i ⊆ U. Finally, we let F = (K0 ∩ A) ∪ [(K0 ∩
+ω) \ (�{B : B ∈ K0 ∩ A})]. Then, F is a finite set of X. Moreover, note that
+F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) and F ∩ U = ∅. Therefore, ζ is a πV (K(X))-network of
+X.
+Case II: K0 is finite.
+For each m ∈ K0 ∩ ω (if K0 ∩ ω ̸= ∅), let Bm ∈ A such that m ∈ Bm. Then, the
+collection B = (K0 ∩ A) ∪ {Bm : m ∈ K0 ∩ ω} is finite. We enumerate this col-
+lection as B = {A1, . . . , An}. We define, for each i ∈ {1, . . . , n}, Vi = {Ai} ∪ Ai.
+Then, each Vi is an open set of X and therefore, (V1, . . . , Vn) ∈ ζ. Note that
+K0 ⊆ �n
+i=1 Vi. Hence, �n
+i=1 V c
+i
+⊆ U. Let F = K0. Then, F is a finite set
+of X such that F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) and F ∩ U = ∅. Therefore, ζ is a
+πV (K(X))-network of X.
+Claim 2: ζ is not a πV -network of X.
+Let U = ω. Then, U is an open set of X with U ̸= X. Let (V1, . . . , Vn) be any
+element of ζ, where Vi = {Ai} ∪ (Ai \ Di) (1 ≤ i ≤ n) for some A1, . . . , An ∈ A,
+D1 . . . , Dn ∈ [ω]<ω, n ∈ N. Then, �n
+i=1[{Ai} ∪ (Ai \ Di)]c ⊈ U. Since this fact
+holds for any element of ζ, we conclude that ζ is not a πV -network of X.
+From now on, if Jn is an element in ΠV (∆), we put:
+Jn = {(V n
+1,s, . . . , V n
+ms,s) : s ∈ Sn}.
+Another notion, defined also in [8], that involves a family ∆ is the following:
+Definition 2.3 ([8]). Let (X, τ) be a topological space.
+A family U ⊆ ∆c
+is called a cV (∆)-cover of X, if for any open subsets V1, . . . , Vm of X, there
+exists U ∈ U and F ∈ [X]<ω such that for each i ∈ {1, . . ., m}, F ∩ Vi ̸= ∅,
+�m
+i=1 V c
+i ⊆ U and F ∩ U = ∅. The family of all cV (∆)-covers of a space X is
+denoted by CV (∆).
+The following two lemmas will be useful in the proofs of next results in this
+section and the proofs of these lemmas can be easily obtained; we refer the
+reader to [8] for details. The first lemma says how cV (∆)-covers on a space X
+can be viewed as dense subspaces of certain hyperspaces of X.
+5
+
+Lemma 2.4 ([8]). Let X be a topological space and U ⊆ ∆c. Then U is a
+cV (∆)-cover of X if and only if Uc is a dense subset of (∆, V).
+The next lemma says how πV (∆)-networks of a space X can be interpreted
+as open covers of certain hyperspaces of X.
+Lemma 2.5 ([8]). Let X be a topological space and ζ = {(V1, . . . , Vn) : V1, . . . , Vn
+are open subsets of X, n ∈ ω}. Then ζ is a πV (∆)-network of X if and only if
+the collection U = {⟨V1, . . . , Vn⟩ : (V1, . . . , Vn) ∈ ζ} is an open cover of (∆, V).
+The following selection principle will help us to characterize the selectively
+strongly star-Menger property on hyperspaces with the Vietoris topology.
+Definition 2.6. Let X be a topological space. We define:
+SVM(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each
+sequence {Cn : n ∈ N} ⊆ CV (∆), there are finite subsets Vn ⊆ Cn, n ∈ N,
+such that J = �
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈ Jn : there exists Un ∈ Vn such that
+�ms
+i=1(V n
+i,s)c ⊆ Un, V n
+i,s ⊈ Un (1 ≤ i ≤ ms)} is an element of ΠV (∆).
+Theorem 2.7. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, V) is selSSM;
+(2) X satisfies SVM(ΠV (∆), ΠV (∆)).
+Proof. (1) ⇒ (2): Let {Jn : n ∈ N} be a sequence of πV (∆)-networks of X
+and let {Cn : n ∈ N} be a sequence of cV (∆)-covers of X. By Lemma 2.4, the
+collections Dn = Cc
+n are dense subsets of (∆, V), for each n ∈ N. Furthermore,
+if we put, for each n ∈ ω, Jn = {(V n
+1,s, . . . , V n
+ms,s) : s ∈ Sn}, then by Lemma 2.5,
+the collections Un = {⟨V n
+1,s, . . . , V n
+ms,s⟩ : s ∈ Sn} are open covers of (∆, V), for
+each n ∈ N.
+Now, applying (1) to the sequence {Un : n ∈ N} and the sequence {Dn : n ∈
+N}, there is a sequence {An : n ∈ N} of finite sets such that, for each n ∈ N,
+An ⊆ Dn and the collection {St(An, Un) : n ∈ N} is an open cover of (∆, V).
+We put, for each n ∈ N, Vn = {Ac : A ∈ An}. Then, for each n ∈ N, Vn is a
+finite subset of Cn.
+Let us show that the collection J = �
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈ Jn : ∃ Un ∈
+Vn such that �ms
+i=1(V n
+i,s)c ⊆ Un, V n
+i,s ⊈ Un (1 ≤ i ≤ ms)} is a πV (∆)-network
+of X.
+Let U ∈ ∆c. Then U c ∈ ∆ and therefore, there exists n0 ∈ ω such that
+U c ∈ St(An0, Un0). Then, there are ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩ ∈ Un0 and An0 ∈ An0 so
+that {U c, An0} ⊆ ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩. Let Un0 = Ac
+n0. Then Un0 ∈ Vn0. Since
+An0 belongs to ⟨V n0
+1,s0, . . . , V n0
+ms0,s0⟩, it follows that �ms0
+i=1 (V n0
+i,s0)c ⊆ Un0,
+V n0
+i,s0 ⊈
+Un0(1 ≤ i ≤ ms0); hence (V n0
+1,s0, . . . , V n0
+ms0 ,s0) ∈ J . On the other hand, using
+the fact that U c also belongs to ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩, we can take, for each i ∈
+{1, . . ., ms0}, xi ∈ U c ∩ V n0
+i,s0. We put F = {xi : i ∈ {1, . . . , ms0}}. Hence,
+6
+
+F ∈ [X]<ω with F ∩ V n0
+i,s0 ̸= ∅ and F ∩ U = ∅. Moreover, since U c ⊆ �ms0
+i=1 V n0
+i,s0,
+then we obtain that �ms0
+i=1 (V n0
+i,s0)c ⊆ U. We conclude that J ∈ ΠV (∆).
+(2) ⇒ (1): Let {Un : n ∈ N} be a sequence of open covers of (∆, V) and let
+{Dn : n ∈ N} be a sequence of dense subsets of (∆, V). We can assume that
+each open cover Un consists of basic open subsets in CL(X). Thus, put for each
+n ∈ N, Un = {⟨V n
+1,s, . . . , V n
+ms,s⟩ : s ∈ Sn}, where V n
+i,s is an open subset of X, for
+every n ∈ N, s ∈ Sn and i ∈ {1, . . . , ms}. Let Jn = {(V n
+1,s, . . . , V n
+ms,s) : s ∈ Sn}
+for every n ∈ N. By Lemma 2.5, note that each Jn is a πV (∆)-network of X.
+On the other hand, for each n ∈ N, let Cn = Dc
+n. Thus, by Lemma 2.4, each Cn
+is a cV (∆)-cover of X.
+We apply (2) to the sequence of πV (∆)-networks {Jn : n ∈ N} and the se-
+quence of cV (∆)-covers {Cn : n ∈ N} to obtain a sequence {Vn : n ∈ N} such
+that, for each n ∈ N, Vn ∈ [Cn]<ω, and the collection J = �
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈
+Jn : there exists Un ∈ Vn such that �ms
+i=1(V n
+i,s)c ⊆ Un, V n
+i,s ⊈ Un
+(1 ≤ i ≤
+ms)} is a πV (∆)-network of X. For each n ∈ ω, we define An = Vc
+n. It follows
+that An ∈ [Dn]<ω, for each n ∈ N.
+Let us show that the collection {St(An, Un) : n ∈ N} is an open cover of
+(∆, V). Let A ∈ ∆. Since J is a πV (∆)-network of X and Ac ∈ ∆c, there exist
+(V n0
+1,s0, . . . , V n0
+ms0,s0) ∈ J (for some n0 ∈ N and some s0 ∈ Sn0) and a finite set
+F ⊆ X such that for every i ∈ {1, . . . , ms0}, F ∩ V n0
+i,s0 ̸= ∅, �ms0
+i=1 (V n0
+i,s0)c ⊆ Ac
+and F ∩ Ac = ∅. Since (V n0
+1,s0, . . . , V n0
+ms0,s0) ∈ J , there is Un0 ∈ Vn0 such that
+�ms0
+i=1 (V n0
+i,s0)c ⊆ Un0 and for each i ∈ {1, . . . , ms0}, V n0
+i,s0 ⊈ Un0.
+It means
+that {A, An0} ⊆ ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩ ∈ Un0, where An0 = U c
+n0. Since An0 is an
+element of An0, we obtain that A ∈ St(An0, Un0). This shows that the collection
+{St(An, Un) : n ∈ N} is an open cover of (∆, V).
+We obtain the following particular cases by taking different choices of our
+family ∆.
+Corollary 2.8. Let X be a topological space. Then:
+1. (CL(X), V) is selSSM if and only if X satisfies SVM(ΠV , ΠV );
+2. (K(X), V) is selSSM if and only if X satisfies SVM(ΠV (K(X)), ΠV (K(X)));
+3. (CS(X), V) is selSSM if and only if X satisfies SVM(ΠV (CS(X)), ΠV (CS(X)));
+4. (F(X), V) is selSSM if and only if X satisfies SVM(ΠV (F(X)), ΠV (F(X))).
+Let us now define another selection principle to characterize the selectively
+strongly star-Rothberger property on hyperspaces with the Vietoris topology.
+Definition 2.9. Let X be a topological space. We define:
+SVR(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each
+sequence {Cn : n ∈ N} ⊆ CV (∆), there is a sequence {Cn : n ∈ N} with
+Cn ∈ Cn, n ∈ N, such that J = �
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈ Jn : �ms
+i=1(V n
+i,s)c ⊆ Cn,
+V n
+i,s ⊈ Cn (1 ≤ i ≤ ms)} is an element of ΠV (∆).
+7
+
+Theorem 2.10. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, V) is selSSR;
+(2) X satisfies SVR(ΠV (∆), ΠV (∆)).
+Proof. (1) ⇒ (2): By mimicking the first part of (1) ⇒ (2) in the proof of
+Theorem 2.7, we can obtain a sequence {Dn : n ∈ N} of points in (∆, V) such
+that, for each n ∈ N, Dn ∈ Dn and the collection {St(Dn, Un) : n ∈ N} is an
+open cover of (∆, V). We put, for each n ∈ N, Cn = Dc
+n. Then, Cn ∈ Cn, for
+each n ∈ N.
+Let us show that, taking the sequence {Cn : n ∈ N}, the collection J =
+�
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈ Jn : �ms
+i=1(V n
+i,s)c ⊆ Cn, V n
+i,s ⊈ Cn (1 ≤ i ≤ ms)} is a
+πV (∆)-network of X.
+Let U ∈ ∆c. Then there exists n0 ∈ ω such that U c ∈ St(Dn0, Un0). Thus,
+U c, Dn0 ∈ ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩ for some ⟨V n0
+1,s0, . . . , V n0
+ms0,s0⟩ ∈ Un0. Since Cn0 =
+Dc
+n0 and Dn0 ∈ ⟨V n0
+1,s0, . . . , V n0
+ms0,s0⟩, it follows that �ms0
+i=1 (V n0
+i,s0)c ⊆ Cn0 and
+V n0
+i,s0 ⊈ Cn0(1 ≤ i ≤ ms0); hence (V n0
+1,s0, . . . , V n0
+ms0 ,s0) ∈ J . In addition, using
+that U c is an element of ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩, we obtain that �ms0
+i=1 (V n0
+i,s0)c ⊆ U
+and also, we can define a finite subset F of X such that F ∩ V n0
+i,s0 ̸= ∅ and
+F ∩ U = ∅. We conclude that J ∈ ΠV (∆).
+(2) ⇒ (1): In the same way as in the first part of (2) ⇒ (1) in the proof of
+Theorem 2.7, we obtain a sequence {Cn : n ∈ N} such that, for each n ∈ N,
+Cn ∈ Cn, and the collection J = �
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈ Jn : �ms
+i=1(V n
+i,s)c ⊆
+Cn, V n
+i,s ⊈ Cn (1 ≤ i ≤ ms)} is a πV (∆)-network of X. We put Dn = Cc
+n for
+each n ∈ ω. Then, we have that {Dn : n ∈ N} is a sequence of points in (∆, V)
+with Dn ∈ Dn, for each n ∈ N.
+Let us show that the collection {St(Dn, Un) : n ∈ N} is an open cover of
+(∆, V). Let A ∈ ∆. There exists (V n0
+1,s0, . . . , V n0
+ms0,s0) ∈ J (for some n0 ∈ N and
+some s0 ∈ Sn0) and a finite set F ⊆ X such that for every i ∈ {1, . . . , ms0},
+F ∩ V n0
+i,s0 ̸= ∅,
+�ms0
+i=1 (V n0
+i,s0)c ⊆ Ac and F ∩ Ac = ∅.
+Using the fact that
+(V n0
+1,s0, . . . , V n0
+ms0,s0) is an element of J , we get that �ms0
+i=1 (V n0
+i,s0)c ⊆ Cn0 and
+for each i ∈ {1, . . . , ms0}, V n0
+i,s0 ⊈ Cn0. Thus, A, Dn0 are in ⟨V n0
+1,s0, . . . , V n0
+ms0 ,s0⟩
+which is an element of Un0. In other words, A ∈ St(Dn0, Un0). This shows that
+the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, V).
+We obtain the following particular cases by taking different choices of our
+family ∆.
+Corollary 2.11. Let X be a topological space. Then:
+1. (CL(X), V) is selSSR if and only if X satisfies SVR(ΠV , ΠV );
+2. (K(X), V) is selSSR if and only if X satisfies SVR(ΠV (K(X)), ΠV (K(X)));
+3. (CS(X), V) is selSSR if and only if X satisfies SVR(ΠV (CS(X)), ΠV (CS(X)));
+8
+
+4. (F(X), V) is selSSR if and only if X satisfies SVR(ΠV (F(X)), ΠV (F(X))).
+Now, let us deal with the absolute versions of star selection principles for
+Menger and Rothberger cases.
+We start giving the following principle that
+will allow us to characterize the absolutely strongly star-Menger property on
+different hyperspaces with the Vietoris topology.
+Definition 2.12. Let X be a topological space. We define:
+AVM(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each
+C ∈ CV (∆), there is a sequence {Vn : n ∈ N} ⊆ [C]<ω such that J
+=
+�
+n∈N{(V n
+1,s, . . . , V n
+ms,s) ∈ Jn :
+there exists V ∈ Vn such that �ms
+i=1(V n
+i,s)c ⊆
+V, V n
+i,s ⊈ V (1 ≤ i ≤ ms)} is an element of ΠV (∆).
+By doing some light modifications to the proof of Theorem 2.7, one can
+easily prove the following result.
+Theorem 2.13. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, V) is aSSM;
+(2) X satisfies AVM(ΠV (∆), ΠV (∆)).
+Again, some immediate consequences of the previous theorem are obtained
+by taking different choices of our family ∆.
+Corollary 2.14. Let X be a topological space. Then:
+1. (CL(X), V) is aSSM if and only if X satisfies AVM(ΠV , ΠV );
+2. (K(X), V) is aSSM if and only if X satisfies AVM(ΠV (K(X)), ΠV (K(X)));
+3. (CS(X), V) is aSSM if and only if X satisfies AVM(ΠV (CS(X)), ΠV (CS(X)));
+4. (F(X), V) is aSSM if and only if X satisfies AVM(ΠV (F(X)), ΠV (F(X))).
+Next principle allows us to characterize the absolutely strongly star-Rothberger
+property on hyperspaces considering the Vietoris topology.
+Definition 2.15. Let X be a topological space. We define:
+AVR(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each C ∈
+CV (∆), there is a sequence {Cn : n ∈ N} ⊆ C, such that J = �
+n∈N{(V n
+1,s, . . . , V n
+ms,s)
+∈ Jn : �ms
+i=1(V n
+i,s)c ⊆ Cn, V n
+i,s ⊈ Cn (1 ≤ i ≤ ms)} is an element of ΠV (∆).
+Following same idea as in the proof of Theorem 2.10, it is easy to prove the
+following result.
+Theorem 2.16. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, V) is aSSR;
+9
+
+(2) X satisfies AVR(ΠV (∆), ΠV (∆)).
+By taking different choices of the family ∆, we obtain the following particular
+cases.
+Corollary 2.17. Let X be a topological space. Then:
+1. (CL(X), V) is aSSR if and only if X satisfies AVR(ΠV , ΠV );
+2. (K(X), V) is aSSR if and only if X satisfies AVR(ΠV (K(X)), ΠV (K(X)));
+3. (CS(X), V) is aSSR if and only if X satisfies AVR(ΠV (CS(X)), ΠV (CS(X)));
+4. (F(X), V) is aSSR if and only if X satisfies AVR(ΠV (F(X)), ΠV (F(X))).
+3
+Absolute and selective versions on hyperspaces
+with the Fell topology
+In this section we now introduce some other technical principles which will help
+us to make some characterizations of same variations considered in previous
+section on different hyperspaces with the Fell topology. For that end, we will
+show a couple of lemmas which will be often used in the proofs of theorems in
+this section; it is worth to mention that these lemmas are the analogous ones to
+the Lemmas 2.4 and 2.5 mentioned in Section 2.
+Let us first recall the definition of a πF (∆)-network of a space X (see [4]).
+Let ξ denote the family
+ξ = {(K; V1, . . . , Vn) : K is a compact subset of X, V1, . . . , Vn are
+open subsets of X with Vi ∩ Kc ̸= ∅ (1 ≤ i ≤ n) , n ∈ N}.
+Definition 3.1 ([4]). Let (X, τ) be a topological space. A family ξ is called a
+πF (∆)-network of X, if for each U ∈ ∆c, there exist a (K; V1, . . . , Vn) ∈ ξ and
+a finite set F with F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) such that K ⊂ U and F ∩ U = ∅.
+The family of all πF (∆)-network of X is denoted by ΠF (∆).
+Similarly to Definition 2.3, we can define the notion of kF (∆)-cover of a
+space X by doing some light modifications to the notion of kF -cover introduced
+in [16].
+Definition 3.2. Let (X, τ) be a topological space. A family U ⊆ ∆c is called a
+kF (∆)-cover of X, if for any compact subset K of X and open subsets V1, . . . , Vm
+of X, there exists U ∈ U and F ∈ [X]<ω with F ∩ Vi ̸= ∅ (1 ≤ i ≤ m) such
+that K ⊆ U and F ∩ U = ∅. The family of all kF (∆)-covers of X is denoted by
+KF (∆).
+The following lemma says how kF (∆)-covers on a space X can be viewed as
+dense subspaces of certain hyperspaces of X with the Fell topology.
+10
+
+Lemma 3.3. Let X be a topological space and U ⊆ ∆c. Then U is a kF (∆)-
+cover of X if and only if Uc is a dense subset of (∆, F).
+Proof. [→] Let (�n
+i=1 V −
+i ) ∩ (Kc)+ be a nonempty basic open set in (∆, F).
+Since V1, . . . , Vn are open subsets of X, K is a compact subset of X and U is a
+kF (∆)-cover, then there exist U ∈ U and F ∈ [X]<ω with F ∩Vi ̸= ∅ (1 ≤ i ≤ m)
+such that K ⊆ U and F ∩ U = ∅. Put A = U c ∈ ∆. Then, it easy to show that
+A belongs to
+�
+(�n
+i=1 V −
+i ) ∩ (Kc)+�
+∩ Uc. Thus, Uc is dense set in (∆, F).
+[←] Let K be a compact subset of X and let V1, . . . , Vm be open sub-
+sets of X. Using these sets to define the basic open set (�n
+i=1 V −
+i ) ∩ (Kc)+
+of (∆, F), we have that
+�
+(�n
+i=1 V −
+i ) ∩ (Kc)+�
+∩ Uc ̸= ∅.
+So, we take A ∈
+�
+(�n
+i=1 V −
+i ) ∩ (Kc)+�
+∩ Uc. Using this fact, it is easy to define a finite subset of
+X so that F ∩ Vi ̸= ∅ (1 ≤ i ≤ m). Moreover, if we define U = Ac, then K ⊆ U
+and F can be defined so that F ∩ U = ∅. Thus, U is a kF (∆)-cover of X.
+The next lemma says how πF (∆)-networks of a space X can be interpreted
+as open covers of certain hyperspaces of X considering the Fell topology. We
+remark that both directions in the following lemma were used in some proofs
+of theorems in [4]; for an easier reference of the reader, we explicitly state this
+fact and give the proof here.
+Lemma 3.4. Let X be a topological space and ξ = {(K; V1, . . . , Vn) : K is a
+compact subset of X, V1, . . . , Vn are open subsets of X with Vi ∩ Kc ̸= ∅ (1 ≤
+i ≤ n), n ∈ N}. Then ξ is a πF (∆)-network of X if and only if the collection
+U = {(�n
+i=1 V −
+i ) ∩ (Kc)+ : (K; V1, . . . , Vn) ∈ ξ} is an open cover of (∆, F).
+Proof. [→] Let B ∈ ∆. Since ξ is a πF (∆)-network of X and Bc ∈ ∆c, then
+there exists (K; V1, . . . , Vn) ∈ ξ and a finite subset F of X with F ∩ Vi ̸= ∅
+(1 ≤ i ≤ n) such that K ⊆ Bc and F ∩ Bc = ∅.
+These conditions imply
+that B ⊆ Kc and F ⊆ B.
+Since F ∩ Vi ̸= ∅ for each i ∈ {1, . . . , n}, then
+B ∩ Vi ̸= ∅ for each i ∈ {1, . . ., n}. It follows that B ∈ (�n
+i=1 V −
+i ) ∩ (Kc)+ with
+(�n
+i=1 V −
+i ) ∩ (Kc)+ being an element of U. Hence, the collection U is an open
+cover of (∆, F).
+[←] Let U ∈ ∆c. Since U c ∈ ∆ and the collection U = {(�n
+i=1 V −
+i )∩(Kc)+ :
+(K; V1, . . . , Vn) ∈ ξ} is an open cover of (∆, F), there exists (�n
+i=1 V −
+i )∩(Kc)+ ∈
+U such that U c ∈ (�n
+i=1 V −
+i ) ∩ (Kc)+. From this fact, we obtain that K ⊆ U
+and U c ∩ Vi ̸= ∅ for each i ∈ {1 . . ., n}. Hence, we can take xi ∈ U c ∩ Vi for
+each i ∈ {1, . . . , n} and define F = {xi : 1 ≤ i ≤ n}. Thus, F is a finite subset
+of X with F ∩ U = ∅ and F ∩ Vi ̸= ∅ (1 ≤ i ≤ n). Since the respective element
+(K; V1, . . . , Vn) is an element of ξ, we conclude that ξ is a πF (∆)-network of X.
+The following selection principle will help us to characterize the selectively
+strongly star-Menger property on hyperspaces with the Fell topology.
+Definition 3.5. Let X be a topological space. We define:
+SFM(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠF (∆) and each se-
+quence {Kn : n ∈ N} ⊆ KF (∆), there are finite subsets Wn ⊆ Kn, n ∈ N, such
+11
+
+that J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn : there exists W ∈ Wn such that Kn
+s
+⊆ W, V n
+i,s ⊈ W (1 ≤ i ≤ ms)} is an element of ΠF (∆).
+Theorem 3.6. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, F) is selSSM;
+(2) X satisfies SFM(ΠF (∆), ΠF (∆)).
+Proof. (1) ⇒ (2): Let {Jn : n ∈ N} be a sequence of πF (∆)-networks of X and
+let {Kn : n ∈ N} be a sequence of kF (∆)-covers of X. For each n ∈ ω, we
+denote Jn = {(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) : s ∈ Sn}. Applying Lemma 3.4, we obtain
+that, for each n ∈ ω, the collection Un = {(�ms
+i=1(V n
+i,s)−) ∩ ((Kn
+s )c)+ : s ∈ Sn} is
+an open cover of (∆, F). Furthermore, by Lemma 3.3, the collection Dn = Kc
+n
+is a dense subset of (∆, F), for each n ∈ N.
+Thus, applying (1) to the sequence {Un : n ∈ N} and the sequence {Dn : n ∈
+N}, there is a sequence {Fn : n ∈ N} of finite sets such that, for each n ∈ N,
+Fn ⊆ Dn and the collection {St(Fn, Un) : n ∈ N} is an open cover of (∆, F).
+We put, for each n ∈ N, Wn = {F c : F ∈ Fn}. Then, for each n ∈ N, Wn is a
+finite subset of Kn.
+Let us show that the collection J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn :
+there exists W ∈ Wn such that Kn
+s ⊆ W, V n
+i,s ⊈ W (1 ≤ i ≤ ms)} is a πF (∆)-
+network of X.
+Let U ∈ ∆c. Then U c ∈ ∆ and therefore, there exists n0 ∈ ω such that
+U c ∈ St(Fn0, Un0). Thus, there is (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+ ∈ Un0 such that
+�
+(�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+� � Fn0 ̸= ∅ and U c ∈ (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+.
+We fix F0 ∈
+�
+(�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+� � Fn0 and define W0 = F c
+0 . Thus
+W0 ∈ Wn0 and, observe that Kn0
+s0 ⊆ W0 and V n0
+i,s0 ⊈ W0(1 ≤ i ≤ ms0) are
+obtained from the fact that F0 ∈ (�ms0
+i=1 (V n0
+i,s0)−)∩((Kn0
+s0 )c)+. It follows that the
+corresponding element (Kn0
+s0 ; V n0
+1,s0, . . . , V n0
+ms0,s0) belongs to J . In addition, from
+the fact that U c ∈ �ms0
+i=1 (V n0
+i,s0)−, we can fix xi ∈ U c ∩ V n0
+i,s0 (i ∈ {1, . . . , ms0})
+and let M = {xi : i ∈ {1, . . ., ms0}}. So, M is a finite subset of X that satisfies
+M ∩ V n0
+i,s0 ̸= ∅ (1 ≤ i ≤ ms0) and M ∩ U = ∅. On the other hand, Kn0
+s0 ⊆ U
+follows from U c ∈ ((Kn0
+s0 )c)+. We conclude that J ∈ ΠF (∆).
+(2) ⇒ (1):
+Let {Un : n ∈ N} be a sequence of open covers of (∆, F)
+and let {Dn : n ∈ N} be a sequence of dense subsets of (∆, F).
+Without
+loss of generality, suppose that the elements of each open cover Un are ba-
+sic open subsets in CL(X). Therefore, let us denote, for each n ∈ N, Un =
+{(�ms
+i=1(V n
+i,s)−) ∩ ((Kn
+s )c)+ : s ∈ Sn}, where each V n
+i,s is an open subset of X
+and Kn
+s is a compact subset of X (for every n ∈ N, s ∈ Sn and i ∈ {1, . . . , ms}).
+For each n ∈ N, let Jn = {(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) : s ∈ Sn}. Then, by Lemma 3.4,
+we get that each Jn is a πF (∆)-network of X. Moreover, if we define, for each
+n ∈ N, Kn = Dc
+n, then each Kn is a kF (∆)-cover of X by Lemma 3.3.
+Applying (2) to the sequence of πF (∆)-networks {Jn : n ∈ N} and the
+sequence of kF (∆)-covers {Kn : n ∈ N} we obtain a sequence {Wn : n ∈ N}
+12
+
+with Wn a finite subset of Kn, for each n ∈ N, and so that the collection
+J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn : there exists W ∈ Wn such that Kn
+s ⊆
+W, V n
+i,s ⊈ W (1 ≤ i ≤ ms)} is a πF (∆)-network of X. We define, for each
+n ∈ ω, Fn = Wc
+n. Then Fn is a finite subset of Dn, for each n ∈ N.
+Let us show that the collection {St(Fn, Un) : n ∈ N} is an open cover of
+(∆, F). Let A ∈ ∆. Using the fact that J is a πF (∆)-network of X with the set
+Ac ∈ ∆c, we obtain an element (Kn0
+s0 ; V n0
+1,s0, . . . , V n0
+ms0,s0) of J (for some n0 ∈ N
+and some s0 ∈ Sn0) and a finite set F ⊆ X such that for every i ∈ {1, . . . , ms0},
+F ∩ V n0
+i,s0 ̸= ∅,
+Kn0
+s0 ⊆ Ac and F ∩ Ac = ∅. It easily follows from these last
+conditions that A belongs to (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+. On the other hand,
+since (Kn0
+s0 ; V n0
+1,s0, . . . , V n0
+ms0 ,s0) ∈ J , there exists W0 ∈ Wn0 such that Kn0
+s0 ⊆ W0
+and for each i ∈ {1, . . . , ms0}, V n0
+i,s0 ⊈ W0. These conditions imply that the
+element F0 = W c
+0 belongs to (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+. Notice that F0 ∈
+Fn0 and (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+ ∈ Un0. It means that A ∈ St(Fn0, Un0).
+Therefore, {St(Fn, Un) : n ∈ N} is an open cover of (∆, F).
+As a consequence of previous theorem, we can characterize the selectively
+strongly star-Menger property on different hyperspaces considered with the Fell
+topology.
+Corollary 3.7. Let X be a topological space. Then:
+1. (CL(X), F) is selSSM if and only if X satisfies SFM(ΠF , ΠF );
+2. (K(X), F) is selSSM if and only if X satisfies SFM(ΠF (K(X)), ΠF (K(X)));
+3. (CS(X), F) is selSSM if and only if X satisfies SFM(ΠF (CS(X)), ΠF (CS(X)));
+4. (F(X), F) is selSSM if and only if X satisfies SFM(ΠF (F(X)), ΠF (F(X))).
+For the Rothberger case we define the following selection principle to get,
+in the same way, characterizations of the selectively strongly star-Rothberger
+property on hyperspaces with the Fell topology.
+Definition 3.8. Let X be a topological space. We define:
+SFR(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ ω} ⊆ ΠF (∆) and each
+sequence {Kn : n ∈ ω} ⊆ KF (∆), there is a sequence {Kn : n ∈ ω} with
+Kn ∈ Kn, n ∈ ω, such that J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn : Kn
+s ⊆
+Kn, V n
+i,s ⊈ Kn (1 ≤ i ≤ ms)} is an element of ΠF (∆).
+Theorem 3.9. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, F) is selSSR;
+(2) X satisfies SFR(ΠF (∆), ΠF (∆)).
+13
+
+Proof. (1) ⇒ (2): By using Lemmas 3.3 and 3.4 and same idea as in the first part
+of (1) ⇒ (2) in the proof of Theorem 3.6, we can obtain a sequence {Dn : n ∈ N}
+of points in (∆, F) such that, for each n ∈ N, Dn ∈ Dn and the collection
+{St(Dn, Un) : n ∈ N} is an open cover of (∆, F). For each n ∈ N, we define
+Kn = Dc
+n. Then, Kn ∈ Kn, for each n ∈ N.
+Considering the sequence {Kn : n ∈ N}, let us show that the collection
+J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn : Kn
+s ⊆ Kn, V n
+i,s ⊈ Kn (1 ≤ i ≤ ms)} is
+a πF (∆)-network of X.
+Let U ∈ ∆c. Then, for the element U c ∈ ∆, there exists n0 ∈ ω such that
+U c ∈ St(Dn0, Un0). Thus, there is (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+ ∈ Un0 such that
+U c, Dn0 ∈ (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+.
+As Kn0 = Dc
+n0 and Dn0 ∈ (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+, it easily follows that
+Kn0
+s0 ⊆ Kn0 and V n0
+i,s0 ⊈ Kn0(1 ≤ i ≤ ms0). This implies that the corresponding
+(Kn0
+s0 ; V n0
+1,s0, . . . , V n0
+ms0,s0) is an element of J . Furthermore, since U c belongs to
+(�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+, we get that Kn0
+s0 ⊆ U and also, we can define a
+finite subset F of X such that F ∩ V n0
+i,s0 ̸= ∅ and F ∩ U = ∅. We conclude that
+J ∈ ΠF (∆).
+(2) ⇒ (1): Following same arguments and employing Lemmas 3.3 and 3.4
+as in the first part of (2) ⇒ (1) in the proof of Theorem 3.6, we obtain a
+sequence {Kn : n ∈ N} such that, for each n ∈ N, Kn ∈ Kn, and the collection
+J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn : Kn
+s ⊆ Kn, V n
+i,s ⊈ Kn (1 ≤ i ≤ ms)} is a
+πF (∆)-network of X. We define, for each n ∈ N, Dn = Kc
+n. Then, {Dn : n ∈ N}
+is a sequence of points in (∆, F) with Dn ∈ Dn, for each n ∈ N.
+We claim that the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, F).
+Indeed, let A ∈ ∆. Since J is a πF (∆)-network of X and Ac ∈ ∆c, there exists
+(Kn0
+s0 ; V n0
+1,s0, . . . , V n0
+ms0,s0) ∈ J (for some n0 ∈ N and some s0 ∈ Sn0) and a finite
+set F ⊆ X such that F ∩ V n0
+i,s0 ̸= ∅ (1 ≤ i ≤ ms0), Kn0
+s0 ⊆ Ac and F ∩ Ac = ∅.
+The fact (Kn0
+s0 ; V n0
+1,s0, . . . , V n0
+ms0,s0) ∈ J means that Kn0
+s0 ⊆ Kn0 and V n0
+i,s0 ⊈ Kn0
+( 1 ≤ i ≤ ms0) and hence, Dn0 ∈ (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+. Moreover, the
+conditions F ∩ V n0
+i,s0 ̸= ∅ (1 ≤ i ≤ ms0),
+Kn0
+s0 ⊆ Ac and F ∩ Ac = ∅ mean
+that A ∈ (�ms0
+i=1 (V n0
+i,s0)−) ∩ ((Kn0
+s0 )c)+. Thus, A ∈ St(Dn0, Un0). In conclusion,
+{St(Dn, Un) : n ∈ N} is an open cover of (∆, F).
+Now, we can characterize the selectively strongly star-Rothberger property
+on several hyperspaces with the Fell topology.
+Corollary 3.10. Let X be a topological space. Then:
+1. (CL(X), F) is selSSR if and only if X satisfies SFR(ΠF , ΠF );
+2. (K(X), F) is selSSR if and only if X satisfies SFR(ΠF (K(X)), ΠF (K(X)));
+3. (CS(X), F) is selSSR if and only if X satisfies SFR(ΠF (CS(X)), ΠF (CS(X)));
+4. (F(X), F) is selSSR if and only if X satisfies SFR(ΠF (F(X)), ΠF (F(X))).
+14
+
+Now we characterize the absolute versions of the Menger-type and Rothberger-
+type star selection principles with the Fell topology. The following principle al-
+lows us to characterize the absolutely strongly star-Menger property on several
+hyperspaces with the Fell topology.
+Definition 3.11. Let X be a topological space. We define:
+AFM(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ ω} ⊆ ΠF (∆) and each
+K ∈ KF (∆), there is a sequence {Wn : n ∈ ω} ⊆ [K]<ω such that J =
+�
+n∈N{(Kn
+s ; V n
+1,s, . . . , V n
+ms,s) ∈ Jn : there exists W ∈ Wn such that Kn
+s ⊆ W, V n
+i,s
+⊈ W (1 ≤ i ≤ ms)} is an element of ΠF (∆).
+Adapting the proof of Theorem 3.6, one can easily obtain the following result.
+Theorem 3.12. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, F) is aSSM;
+(2) X satisfies AFM(ΠF (∆), ΠF (∆)).
+We obtain the following particular cases by taking different choices of our
+family ∆.
+Corollary 3.13. Let X be a topological space. Then:
+1. (CL(X), F) is aSSM if and only if X satisfies AFM(ΠF , ΠF );
+2. (K(X), F) is aSSM if and only if X satisfies AFM(ΠF (K(X)), ΠF (K(X)));
+3. (CS(X), F) is aSSM if and only if X satisfies AFM(ΠF (CS(X)), ΠF (CS(X)));
+4. (F(X), F) is aSSM if and only if X satisfies AFM(ΠF (F(X)), ΠF (F(X))).
+The following principle will give us characterizations of the absolutely strongly
+star-Rothberger property on hyperspaces considering the Fell topology.
+Definition 3.14. Let X be a topological space. We define:
+AFR(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠF (∆) and each K ∈
+KF (∆), there is a sequence {Kn : n ∈ N} ⊆ K, such that J = �
+n∈N{(Kn
+s ; V n
+1,s, . . . ,
+V n
+ms,s) ∈ Jn : Kn
+s ⊆ Kn, V n
+i,s ⊈ Kn (1 ≤ i ≤ ms)} is an element of ΠF (∆).
+Similarly to the proof of Theorem 3.9, one can prove the following result.
+Theorem 3.15. Given a topological space X, the following conditions are equiv-
+alent:
+(1) (∆, F) is aSSR;
+(2) X satisfies AFR(ΠF (∆), ΠF (∆)).
+The following particular cases are obtained by choosing different families ∆.
+15
+
+Corollary 3.16. Let X be a topological space. Then:
+1. (CL(X), F) is aSSR if and only if X satisfies AFR(ΠF , ΠF );
+2. (K(X), F) is aSSR if and only if X satisfies AFR(ΠF (K(X)), ΠF (K(X)));
+3. (CS(X), F) is aSSR if and only if X satisfies AFR(ΠF (CS(X)), ΠF (CS(X)));
+4. (F(X), F) is aSSR if and only if X satisfies AFR(ΠF (F(X)), ΠF (F(X))).
+Concluding remark.
+It worth remarking that, by doing some slightly
+modifications to definitions given in Sections 2 and 3, it is possible to easily
+obtain hit-and-miss generalizations that involve both the Vietoris topology and
+the Fell topology (similar to how it was done in [6] and [7]). However, for con-
+venience of the reader, the author decided to establish simpler notation results
+depending of which topology is being considered.
+References
+[1] M. Bonanzinga, F. Cammaroto, Lj.D.R. Koˇcinac, M.V. Matveev, On
+weaker forms of Menger, Rothberger and Hurewicz properties, Math. Vesnik
+61 (2009) 13-23.
+[2] J. Casas-de la Rosa, S. A. Garcia-Balan, Variations of star selection prin-
+ciples on small spaces, Filomat (2022), 36:14, 4903-4917.
+[3] J. Casas-de la Rosa, I. Mart´ınez-Ruiz, A. Ram´ırez-P´aramo, Star versions
+of the Menger property on hyperspaces, Houston J. Math., to appear.
+[4] J. Casas-de la Rosa, I. Mart´ınez-Ruiz, A. Ram´ırez-P´aramo, Star versions
+of the Rothberger property on hyperspaces, Topol. Appl. 283 (2020), Art.
+ID 107396, 12 pages.
+[5] A. Caserta, G. Di Maio, Lj.D.R. Koˇcinac, Versions of properties (a) and
+(pp), Topology Appl. 158 (2011) 1360–1368.
+[6] R. Cruz-Castillo, A. Ram´ırez-P´aramo, J.F. Tenorio Menger and Menger-
+type star selection principles for hit-and-miss topology, Topol. Appl. 290
+(2021), Art. ID 107574, 12 pages.
+[7] R. Cruz-Castillo, A. Ram´ırez-P´aramo, J.F. Tenorio Star and strong star-
+type versions of Rothberger and Menger principles for hit-and-miss topology,
+Topol. Appl. 300 (2021), Art. ID 107758, 11 pages.
+[8] J.
+D´ıaz-Reyes,
+A.
+Ram´ırez-P´aramo,
+J.F.
+Tenorio
+Rothberger
+and
+Rothberger-type star selection principles on hyperspaces, Topol. Appl. 287
+(2021), Art. ID 107448, 9 pages.
+[9] G. Di Maio, Lj.D.R. Koˇcinac, E. Meccariello, Selection principles and hy-
+perspace topologies, Topol. Appl. 153 (2005) 912-923.
+16
+
+[10] G. Di Maio, Lj.D.R. Koˇcinac, Some covering properties of hyperspaces,
+Topol. Appl. 155 (2008) 1959-1969.
+[11] F. Hern´andez-Hern´andez, M. Hruˇs´ak, Topology of Mr´owka-Isbell spaces.
+In Pseudocompact Topological Spaces, Eds. Hruˇs´ak, Tamariz, Tkachenko.
+Springer International Publishing AG, 2018.
+[12] Lj.D.R. Koˇcinac, On Star Selection Principles Theory, Axioms, (2023) 12,
+93.
+[13] Lj.D.R. Koˇcinac, Star-Menger and related spaces, Publ. Math. Debrecen 55
+(1999) 421-431.
+[14] Lj.D.R. Koˇcinac, Star selection principles: A survey, Khayyam J. Math. 1
+(2015) No. 1 82-106.
+[15] Lj.D.R. Koˇcinac, The Reznichenko property and the Pytkeev property in
+hyperspaces, Acta Math. Hungar. 107, 3 (2005), 231–239.
+[16] Z. Li, Selection principles of the Fell topology and the Vietoris topology,
+Topol. Appl. 212 (2016) 90-104.
+[17] M. Scheepers, Combinatorics of open covers I: Ramsey theory, Topol. Appl.
+69 (1996) 31-62.
+Departamento de Matem´aticas, Facultad de Ciencias, UNAM,
+Circuito Exterior S/N, Ciudad Universitaria, CP 04510, Ciudad de
+M´exico, M´exico.
+Email address: J. Casas-de la Rosa: olimpico.25@hotmail.com
+17
+
diff --git a/qNFJT4oBgHgl3EQfaCwn/content/tmp_files/load_file.txt b/qNFJT4oBgHgl3EQfaCwn/content/tmp_files/load_file.txt
new file mode 100644
index 0000000000000000000000000000000000000000..7f3cc38268fb42e44038ddad13f34fa780473b45
--- /dev/null
+++ b/qNFJT4oBgHgl3EQfaCwn/content/tmp_files/load_file.txt
@@ -0,0 +1,942 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf,len=941
+page_content='arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='11534v1 [math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='GN] 27 Jan 2023 Variations of star selection principles on Hyperspaces∗ JAVIER CASAS-DE LA ROSA Abstract In this paper we define some combinatorial principles to characterize spaces X whose hyperspace satisfies some variation of some classical star selection principle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Specifically, the variations characterized are the selec- tive and absolute versions of the star selection principles for the Menger and Rothberger cases;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' also, the hyperspaces considered in these charac- terizations are CL(X), K(X), F(X) and CS(X) in both cases, endowed with either the Fell topology or the Vietoris topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Key words.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hyperspaces, Fell topology, Vietoris topology, star selection princi- ples, (absolutely) strongly star-Menger, (selectively) strongly star-Menger, (ab- solutely) strongly star-Rothberger, (selectively) strongly star-Rothberger.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Mathematics Subject Classification: Primary 54B20, 54D20;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Secondary 54A05, 54A25.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1 Introduction and preliminaries Many branches from Selection Principles Theory have arisen after a systematic research made in [17] by Scheepers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Nowadays, this theory has connections as well as applications to several areas of mathematics as General Topology, Function spaces, Hyperspaces, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' As an example of it, in [9], [10],[15] and [16], the authors studied the fundamental problem on Hyperspaces which consists in establishing dualities between some topological properties or, as it is in this case, between some classical selection principles under different hyperspaces topologies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In other words, given two selection principles P and Q, one must determine if it holds that a topological space X satisfies the principle P if and only if its hyperspace satisfies the principle Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hence, this duality problem can be viewed as a method to characterize some classical selection principles of X with selection principles on different hyperspaces of X, using several well-known topologies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' On the other hand, several versions of the original selection principles have been defined from its beginning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' One of the most important versions that its investigation has rapidly increased is the star versions of the classical selection ∗The author was supported for this research by Postdoctoral Fellowship Program at UNAM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1 principles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' These star versions were defined by Koˇcinac in [13] and they gave rise to the star selection principles theory (see [14] and [12]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The study of the duality problem involving some star selection principles on hyperspaces initiated very recently in [4] and continued in [3], [6], [7] and [8] have dealt with classical star selection principles only.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In particular, in these works several combinatorial principles have been defined to establish characterizations for the (strongly) star-Menger property and the (strongly) star-Rothberger prop- erty on several hyperspaces under different topologies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In this paper we continue this line of investigation for some variations of some classical star selection prin- ciples.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In Section 2, the selective and absolute versions of Menger-type and Rothberger-type star selection principle on several hyperspaces with the Vi- etoris topology are characterized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In Section 3, analogous characterizations are given for several hyperspaces with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='1 Hyperspaces All spaces are assumed to be Hausdorff noncompact and, even, nonparacom- pact.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X , the hyperspace of X, denoted by CL(X), is the set of all nonempty closed subsets of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' By K(X) (F(X)), we denote the family of all nonempty compact (all nonempty finite) subsets of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Also, by CS(X) we denote the family of all convergent sequences of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We denote by ω the first infinite cardinal and for a set A, [A]<ω denotes the set of all finite subsets of A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For a subset U ⊆ X and a family U of subsets of X, we write: U − = {A ∈ CL(X) : A ∩ U ̸= ∅};' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' U + = {A ∈ CL(X) : A ⊆ U};' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' U c = X\\U;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Uc = {U c : U ∈ U};' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' U− = {U − : U ∈ U};' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' U+ = {U + : U ∈ U}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In the literature there are many topologies that can be defined on CL(X) or on a subcollection of it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In this paper we will consider two well-known topolo- gies, the Vietoris topology, denoted by V, and the Fell topology, denoted by F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A basic open subset of the Fell topology is of the form: ( n� i=1 V − i ) ∩ (Kc)+, where V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn are open subsets of X and K is a compact subset of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A basic open subset of the Vietoris topology is a set of the form: ⟨U1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Un⟩ = {A ∈ CL(X) : A ⊆ n � i=1 Ui, A ∩ Ui ̸= ∅ for each i ≤ n} 2 where U1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Un are open subsets of X, n ∈ ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='2 Variations of the classical star selection principles For a set A ⊆ X and a collection U of subsets of X, the star of A with respect to U, denoted by St(A, U), is the set �{U ∈ U : U ∩ A ̸= ∅};' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' for A = {x} with x ∈ X, we write St(x, U) instead of St({x}, U).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In recent years, different versions of the classical star selection principles have been defined and studied in several articles (see for instance [1], [5] and [2]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Some of these new star selection principles involve dense subsets of the space and they are called as the absolute and selective versions of classical star selection principles (see [14] for more information about the absolute versions and see [2] for an overview about the selective versions).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The selective versions are stronger than the absolute versions and the absolute versions are stronger than the classical star selection principles1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following properties are the absolute versions of the strongly star selection principles for the Menger and Rothberger cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='1 ([5]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We say that a space X is: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' absolutely strongly star-Menger (aSSM) if for each sequence {Un : n ∈ ω} of open covers of X and each dense subset D of X, there is a sequence {Fn : n ∈ ω} of finite sets such that Fn ⊆ D, n ∈ ω, and {St(Fn, Un) : n ∈ ω} is an open cover of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' absolutely strongly star-Rothberger (aSSR) if for each sequence {Un : n ∈ ω} of open covers of X and each dense subset D of X, there is a sequence {xn : n ∈ ω} of points of X such that xn ∈ D, n ∈ ω, and {St(xn, Un) : n ∈ ω} is an open cover of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Now, we recall the selective versions of the strongly star selection principles for the Menger and Rothberger cases2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='2 ([2]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We say that a space X is: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' selectively strongly star-Menger (selSSM) if for each sequence {Un : n ∈ ω} of open covers of X and each sequence {Dn : n ∈ ω} of dense sets of X, there exists a sequence {Fn : n ∈ ω} of finite sets such that Fn ⊆ Dn, n ∈ ω, and {St(Fn, Un) : n ∈ ω} is an open cover of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' selectively strongly star-Rothberger (selSSR) if for each sequence {Un : n ∈ ω} of open covers of X and each sequence {Dn : n ∈ ω} of dense sets of X, there exists a sequence {xn : n ∈ ω} of points of X such that xn ∈ Dn, n ∈ ω, and {St(xn, Un) : n ∈ ω} is an open cover of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1In [5], the authors introduced the absolute versions of classical star selection principles in a general form with a different notation than the one used in [2], where it was defined the absolute and selective versions of star selection principles, also given in a general form.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2The Hurewicz case and some other interesting properties are also given in [2] 3 2 Absolute and selective versions on hyperspaces with the Vietoris topology In this section we present some technical principles that are useful to charac- terize some variations of classical star selection principles (selSSM, selSSR, aSSM and aSSR) on several hiperspaces with the Vietoris topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' To that end, we recall some notation and useful definitions to establish these character- izations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Using the notation of [16], ζ denotes the family: ζ = {(V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) : V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn are open subsets of X, n ∈ N}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In [16], Li defined the notion of πV -networks in a space X as follows: A family ζ is called a πV -network of X if for each open subset U of X, with U ̸= X, there exist (V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ζ and a finite set F with F ∩Vi ̸= ∅ (1 ≤ i ≤ n) such that �n i=1 V c i ⊆ U and F ∩U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The collection of πV -networks of a space X is denoted by ΠV .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Henceforth, ∆ will denote a subset of CL(X) which is closed under finite unions and contains all singletons.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Using a family ∆, in [8], it was defined a modification of πV -network which is called as πV (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='1 ([8]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A family ζ is called a πV (∆)-network of X, if for each U ∈ ∆c, there exist a (V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ζ and F ∈ [X]<ω with F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) such that �n i=1 V c i ⊂ U and F ∩U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The collection of all πV (∆)-networks of X is denoted by ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' As pointed out in [8], if we consider ∆ to be CL(X), then the collections ΠV (∆) and ΠV coincide.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' But this fact not need be true in general.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In [8], the authors gave an example of a πV (∆)-network that is not a πV -network (on a metrizable space X and a specific family ∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Here we present another example in a non-metrizable space1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Example 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' There exists a family ζ on a certain space X and there exists a family ∆ ⊆ CL(X) such that ζ is a πV (∆)-network of X but is not a πV -network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let A be an uncountable almost disjoint family on ω with ω = � A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We consider the Mr´owka-Isbell space (see [11]), X = Ψ(A) and ∆ = K(X).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let ζ = {(V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) : Vi = {Ai} ∪ (Ai \\ Di) (1 ≤ i ≤ n) where A1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , An ∈ A, D1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Dn ∈ [ω]<ω, n ∈ N} 1This space was also used in [4] to show that πF (∆)-network and πF -network are different notions;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' see Section 3 for definitions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4 Note that ζ is properly defined, that is, for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , n}, {Ai} ∪ (Ai \\ Di) is an open set of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Claim 1: ζ is a πV (K(X))-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let U ∈ [K(X)]c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Suppose that U = X \\ K0 for some K0 ∈ K(X).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We consider the following two cases: Case I: K0 is infinite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Note that in this case, K0 ∩ A is a nonempty finite set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Moreover, the set (K0 ∩ ω) \\ (�{B : B ∈ K0 ∩ A}) is finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For each m ∈ (K0 ∩ ω) \\ (�{B : B ∈ K0∩A}), let Bm ∈ A be such that m ∈ Bm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The collection B = (K0∩A)∪{Bm : m ∈ (K0 ∩ ω) \\ (�{B : B ∈ K0 ∩ A})} is finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We enumerate this collection as B = {A1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , An}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define, for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', n}, Vi = {Ai} ∪ Ai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, each Vi is an open set of X and therefore, (V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ζ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Note that K0 ⊆ �n i=1 Vi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, �n i=1 V c i ⊆ U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Finally, we let F = (K0 ∩ A) ∪ [(K0 ∩ ω) \\ (�{B : B ∈ K0 ∩ A})].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, F is a finite set of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Moreover, note that F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Therefore, ζ is a πV (K(X))-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Case II: K0 is finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For each m ∈ K0 ∩ ω (if K0 ∩ ω ̸= ∅), let Bm ∈ A such that m ∈ Bm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, the collection B = (K0 ∩ A) ∪ {Bm : m ∈ K0 ∩ ω} is finite.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We enumerate this col- lection as B = {A1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , An}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define, for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , n}, Vi = {Ai} ∪ Ai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, each Vi is an open set of X and therefore, (V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ζ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Note that K0 ⊆ �n i=1 Vi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hence, �n i=1 V c i ⊆ U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let F = K0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, F is a finite set of X such that F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Therefore, ζ is a πV (K(X))-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Claim 2: ζ is not a πV -network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let U = ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, U is an open set of X with U ̸= X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let (V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) be any element of ζ, where Vi = {Ai} ∪ (Ai \\ Di) (1 ≤ i ≤ n) for some A1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , An ∈ A, D1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Dn ∈ [ω]<ω, n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, �n i=1[{Ai} ∪ (Ai \\ Di)]c ⊈ U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since this fact holds for any element of ζ, we conclude that ζ is not a πV -network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' From now on, if Jn is an element in ΠV (∆), we put: Jn = {(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) : s ∈ Sn}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Another notion, defined also in [8], that involves a family ∆ is the following: Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3 ([8]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let (X, τ) be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A family U ⊆ ∆c is called a cV (∆)-cover of X, if for any open subsets V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vm of X, there exists U ∈ U and F ∈ [X]<ω such that for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', m}, F ∩ Vi ̸= ∅, �m i=1 V c i ⊆ U and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The family of all cV (∆)-covers of a space X is denoted by CV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following two lemmas will be useful in the proofs of next results in this section and the proofs of these lemmas can be easily obtained;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' we refer the reader to [8] for details.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The first lemma says how cV (∆)-covers on a space X can be viewed as dense subspaces of certain hyperspaces of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 5 Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4 ([8]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space and U ⊆ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then U is a cV (∆)-cover of X if and only if Uc is a dense subset of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The next lemma says how πV (∆)-networks of a space X can be interpreted as open covers of certain hyperspaces of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='5 ([8]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space and ζ = {(V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) : V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn are open subsets of X, n ∈ ω}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then ζ is a πV (∆)-network of X if and only if the collection U = {⟨V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn⟩ : (V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ζ} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following selection principle will help us to characterize the selectively strongly star-Menger property on hyperspaces with the Vietoris topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: SVM(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each sequence {Cn : n ∈ N} ⊆ CV (∆), there are finite subsets Vn ⊆ Cn, n ∈ N, such that J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists Un ∈ Vn such that �ms i=1(V n i,s)c ⊆ Un, V n i,s ⊈ Un (1 ≤ i ≤ ms)} is an element of ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, V) is selSSM;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies SVM(ΠV (∆), ΠV (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (1) ⇒ (2): Let {Jn : n ∈ N} be a sequence of πV (∆)-networks of X and let {Cn : n ∈ N} be a sequence of cV (∆)-covers of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' By Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4, the collections Dn = Cc n are dense subsets of (∆, V), for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Furthermore, if we put, for each n ∈ ω, Jn = {(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) : s ∈ Sn}, then by Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='5, the collections Un = {⟨V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s⟩ : s ∈ Sn} are open covers of (∆, V), for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Now, applying (1) to the sequence {Un : n ∈ N} and the sequence {Dn : n ∈ N}, there is a sequence {An : n ∈ N} of finite sets such that, for each n ∈ N, An ⊆ Dn and the collection {St(An, Un) : n ∈ N} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We put, for each n ∈ N, Vn = {Ac : A ∈ An}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, for each n ∈ N, Vn is a finite subset of Cn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us show that the collection J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : ∃ Un ∈ Vn such that �ms i=1(V n i,s)c ⊆ Un, V n i,s ⊈ Un (1 ≤ i ≤ ms)} is a πV (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let U ∈ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then U c ∈ ∆ and therefore, there exists n0 ∈ ω such that U c ∈ St(An0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, there are ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩ ∈ Un0 and An0 ∈ An0 so that {U c, An0} ⊆ ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let Un0 = Ac n0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then Un0 ∈ Vn0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since An0 belongs to ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0⟩, it follows that �ms0 i=1 (V n0 i,s0)c ⊆ Un0, V n0 i,s0 ⊈ Un0(1 ≤ i ≤ ms0);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' hence (V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0) ∈ J .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' On the other hand, using the fact that U c also belongs to ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩, we can take, for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', ms0}, xi ∈ U c ∩ V n0 i,s0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We put F = {xi : i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hence, 6 F ∈ [X]<ω with F ∩ V n0 i,s0 ̸= ∅ and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Moreover, since U c ⊆ �ms0 i=1 V n0 i,s0, then we obtain that �ms0 i=1 (V n0 i,s0)c ⊆ U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We conclude that J ∈ ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) ⇒ (1): Let {Un : n ∈ N} be a sequence of open covers of (∆, V) and let {Dn : n ∈ N} be a sequence of dense subsets of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We can assume that each open cover Un consists of basic open subsets in CL(X).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, put for each n ∈ N, Un = {⟨V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s⟩ : s ∈ Sn}, where V n i,s is an open subset of X, for every n ∈ N, s ∈ Sn and i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let Jn = {(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) : s ∈ Sn} for every n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' By Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='5, note that each Jn is a πV (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' On the other hand, for each n ∈ N, let Cn = Dc n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, by Lemma 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4, each Cn is a cV (∆)-cover of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We apply (2) to the sequence of πV (∆)-networks {Jn : n ∈ N} and the se- quence of cV (∆)-covers {Cn : n ∈ N} to obtain a sequence {Vn : n ∈ N} such that, for each n ∈ N, Vn ∈ [Cn]<ω, and the collection J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists Un ∈ Vn such that �ms i=1(V n i,s)c ⊆ Un, V n i,s ⊈ Un (1 ≤ i ≤ ms)} is a πV (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For each n ∈ ω, we define An = Vc n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It follows that An ∈ [Dn]<ω, for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us show that the collection {St(An, Un) : n ∈ N} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let A ∈ ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since J is a πV (∆)-network of X and Ac ∈ ∆c, there exist (V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) ∈ J (for some n0 ∈ N and some s0 ∈ Sn0) and a finite set F ⊆ X such that for every i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}, F ∩ V n0 i,s0 ̸= ∅, �ms0 i=1 (V n0 i,s0)c ⊆ Ac and F ∩ Ac = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since (V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) ∈ J , there is Un0 ∈ Vn0 such that �ms0 i=1 (V n0 i,s0)c ⊆ Un0 and for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}, V n0 i,s0 ⊈ Un0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It means that {A, An0} ⊆ ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩ ∈ Un0, where An0 = U c n0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since An0 is an element of An0, we obtain that A ∈ St(An0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' This shows that the collection {St(An, Un) : n ∈ N} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We obtain the following particular cases by taking different choices of our family ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), V) is selSSM if and only if X satisfies SVM(ΠV , ΠV );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), V) is selSSM if and only if X satisfies SVM(ΠV (K(X)), ΠV (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), V) is selSSM if and only if X satisfies SVM(ΠV (CS(X)), ΠV (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), V) is selSSM if and only if X satisfies SVM(ΠV (F(X)), ΠV (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us now define another selection principle to characterize the selectively strongly star-Rothberger property on hyperspaces with the Vietoris topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: SVR(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each sequence {Cn : n ∈ N} ⊆ CV (∆), there is a sequence {Cn : n ∈ N} with Cn ∈ Cn, n ∈ N, such that J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : �ms i=1(V n i,s)c ⊆ Cn, V n i,s ⊈ Cn (1 ≤ i ≤ ms)} is an element of ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 7 Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, V) is selSSR;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies SVR(ΠV (∆), ΠV (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (1) ⇒ (2): By mimicking the first part of (1) ⇒ (2) in the proof of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='7, we can obtain a sequence {Dn : n ∈ N} of points in (∆, V) such that, for each n ∈ N, Dn ∈ Dn and the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We put, for each n ∈ N, Cn = Dc n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, Cn ∈ Cn, for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us show that, taking the sequence {Cn : n ∈ N}, the collection J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : �ms i=1(V n i,s)c ⊆ Cn, V n i,s ⊈ Cn (1 ≤ i ≤ ms)} is a πV (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let U ∈ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then there exists n0 ∈ ω such that U c ∈ St(Dn0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, U c, Dn0 ∈ ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩ for some ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0⟩ ∈ Un0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since Cn0 = Dc n0 and Dn0 ∈ ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0⟩, it follows that �ms0 i=1 (V n0 i,s0)c ⊆ Cn0 and V n0 i,s0 ⊈ Cn0(1 ≤ i ≤ ms0);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' hence (V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0) ∈ J .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In addition, using that U c is an element of ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩, we obtain that �ms0 i=1 (V n0 i,s0)c ⊆ U and also, we can define a finite subset F of X such that F ∩ V n0 i,s0 ̸= ∅ and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We conclude that J ∈ ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) ⇒ (1): In the same way as in the first part of (2) ⇒ (1) in the proof of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='7, we obtain a sequence {Cn : n ∈ N} such that, for each n ∈ N, Cn ∈ Cn, and the collection J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : �ms i=1(V n i,s)c ⊆ Cn, V n i,s ⊈ Cn (1 ≤ i ≤ ms)} is a πV (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We put Dn = Cc n for each n ∈ ω.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, we have that {Dn : n ∈ N} is a sequence of points in (∆, V) with Dn ∈ Dn, for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us show that the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let A ∈ ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' There exists (V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) ∈ J (for some n0 ∈ N and some s0 ∈ Sn0) and a finite set F ⊆ X such that for every i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}, F ∩ V n0 i,s0 ̸= ∅, �ms0 i=1 (V n0 i,s0)c ⊆ Ac and F ∩ Ac = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Using the fact that (V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) is an element of J , we get that �ms0 i=1 (V n0 i,s0)c ⊆ Cn0 and for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}, V n0 i,s0 ⊈ Cn0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, A, Dn0 are in ⟨V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0⟩ which is an element of Un0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In other words, A ∈ St(Dn0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' This shows that the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, V).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We obtain the following particular cases by taking different choices of our family ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), V) is selSSR if and only if X satisfies SVR(ΠV , ΠV );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), V) is selSSR if and only if X satisfies SVR(ΠV (K(X)), ΠV (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), V) is selSSR if and only if X satisfies SVR(ΠV (CS(X)), ΠV (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 8 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), V) is selSSR if and only if X satisfies SVR(ΠV (F(X)), ΠV (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Now, let us deal with the absolute versions of star selection principles for Menger and Rothberger cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We start giving the following principle that will allow us to characterize the absolutely strongly star-Menger property on different hyperspaces with the Vietoris topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: AVM(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each C ∈ CV (∆), there is a sequence {Vn : n ∈ N} ⊆ [C]<ω such that J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists V ∈ Vn such that �ms i=1(V n i,s)c ⊆ V, V n i,s ⊈ V (1 ≤ i ≤ ms)} is an element of ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' By doing some light modifications to the proof of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='7, one can easily prove the following result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, V) is aSSM;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies AVM(ΠV (∆), ΠV (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Again, some immediate consequences of the previous theorem are obtained by taking different choices of our family ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), V) is aSSM if and only if X satisfies AVM(ΠV , ΠV );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), V) is aSSM if and only if X satisfies AVM(ΠV (K(X)), ΠV (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), V) is aSSM if and only if X satisfies AVM(ΠV (CS(X)), ΠV (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), V) is aSSM if and only if X satisfies AVM(ΠV (F(X)), ΠV (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Next principle allows us to characterize the absolutely strongly star-Rothberger property on hyperspaces considering the Vietoris topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: AVR(ΠV (∆), ΠV (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠV (∆) and each C ∈ CV (∆), there is a sequence {Cn : n ∈ N} ⊆ C, such that J = � n∈N{(V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : �ms i=1(V n i,s)c ⊆ Cn, V n i,s ⊈ Cn (1 ≤ i ≤ ms)} is an element of ΠV (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Following same idea as in the proof of Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='10, it is easy to prove the following result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, V) is aSSR;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 9 (2) X satisfies AVR(ΠV (∆), ΠV (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' By taking different choices of the family ∆, we obtain the following particular cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), V) is aSSR if and only if X satisfies AVR(ΠV , ΠV );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), V) is aSSR if and only if X satisfies AVR(ΠV (K(X)), ΠV (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), V) is aSSR if and only if X satisfies AVR(ΠV (CS(X)), ΠV (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), V) is aSSR if and only if X satisfies AVR(ΠV (F(X)), ΠV (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3 Absolute and selective versions on hyperspaces with the Fell topology In this section we now introduce some other technical principles which will help us to make some characterizations of same variations considered in previous section on different hyperspaces with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For that end, we will show a couple of lemmas which will be often used in the proofs of theorems in this section;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' it is worth to mention that these lemmas are the analogous ones to the Lemmas 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4 and 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='5 mentioned in Section 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us first recall the definition of a πF (∆)-network of a space X (see [4]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let ξ denote the family ξ = {(K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) : K is a compact subset of X, V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn are open subsets of X with Vi ∩ Kc ̸= ∅ (1 ≤ i ≤ n) , n ∈ N}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='1 ([4]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let (X, τ) be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A family ξ is called a πF (∆)-network of X, if for each U ∈ ∆c, there exist a (K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ξ and a finite set F with F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) such that K ⊂ U and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The family of all πF (∆)-network of X is denoted by ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Similarly to Definition 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3, we can define the notion of kF (∆)-cover of a space X by doing some light modifications to the notion of kF -cover introduced in [16].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let (X, τ) be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A family U ⊆ ∆c is called a kF (∆)-cover of X, if for any compact subset K of X and open subsets V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vm of X, there exists U ∈ U and F ∈ [X]<ω with F ∩ Vi ̸= ∅ (1 ≤ i ≤ m) such that K ⊆ U and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The family of all kF (∆)-covers of X is denoted by KF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following lemma says how kF (∆)-covers on a space X can be viewed as dense subspaces of certain hyperspaces of X with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 10 Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space and U ⊆ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then U is a kF (∆)- cover of X if and only if Uc is a dense subset of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [→] Let (�n i=1 V − i ) ∩ (Kc)+ be a nonempty basic open set in (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn are open subsets of X, K is a compact subset of X and U is a kF (∆)-cover, then there exist U ∈ U and F ∈ [X]<ω with F ∩Vi ̸= ∅ (1 ≤ i ≤ m) such that K ⊆ U and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Put A = U c ∈ ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, it easy to show that A belongs to � (�n i=1 V − i ) ∩ (Kc)+� ∩ Uc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, Uc is dense set in (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [←] Let K be a compact subset of X and let V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vm be open sub- sets of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Using these sets to define the basic open set (�n i=1 V − i ) ∩ (Kc)+ of (∆, F), we have that � (�n i=1 V − i ) ∩ (Kc)+� ∩ Uc ̸= ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' So, we take A ∈ � (�n i=1 V − i ) ∩ (Kc)+� ∩ Uc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Using this fact, it is easy to define a finite subset of X so that F ∩ Vi ̸= ∅ (1 ≤ i ≤ m).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Moreover, if we define U = Ac, then K ⊆ U and F can be defined so that F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, U is a kF (∆)-cover of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The next lemma says how πF (∆)-networks of a space X can be interpreted as open covers of certain hyperspaces of X considering the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We remark that both directions in the following lemma were used in some proofs of theorems in [4];' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' for an easier reference of the reader, we explicitly state this fact and give the proof here.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space and ξ = {(K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) : K is a compact subset of X, V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn are open subsets of X with Vi ∩ Kc ̸= ∅ (1 ≤ i ≤ n), n ∈ N}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then ξ is a πF (∆)-network of X if and only if the collection U = {(�n i=1 V − i ) ∩ (Kc)+ : (K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ξ} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [→] Let B ∈ ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since ξ is a πF (∆)-network of X and Bc ∈ ∆c, then there exists (K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ξ and a finite subset F of X with F ∩ Vi ̸= ∅ (1 ≤ i ≤ n) such that K ⊆ Bc and F ∩ Bc = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' These conditions imply that B ⊆ Kc and F ⊆ B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since F ∩ Vi ̸= ∅ for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , n}, then B ∩ Vi ̸= ∅ for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', n}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It follows that B ∈ (�n i=1 V − i ) ∩ (Kc)+ with (�n i=1 V − i ) ∩ (Kc)+ being an element of U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hence, the collection U is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [←] Let U ∈ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since U c ∈ ∆ and the collection U = {(�n i=1 V − i )∩(Kc)+ : (K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) ∈ ξ} is an open cover of (∆, F), there exists (�n i=1 V − i )∩(Kc)+ ∈ U such that U c ∈ (�n i=1 V − i ) ∩ (Kc)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' From this fact, we obtain that K ⊆ U and U c ∩ Vi ̸= ∅ for each i ∈ {1 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', n}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hence, we can take xi ∈ U c ∩ Vi for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , n} and define F = {xi : 1 ≤ i ≤ n}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, F is a finite subset of X with F ∩ U = ∅ and F ∩ Vi ̸= ∅ (1 ≤ i ≤ n).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since the respective element (K;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , Vn) is an element of ξ, we conclude that ξ is a πF (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following selection principle will help us to characterize the selectively strongly star-Menger property on hyperspaces with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: SFM(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠF (∆) and each se- quence {Kn : n ∈ N} ⊆ KF (∆), there are finite subsets Wn ⊆ Kn, n ∈ N, such 11 that J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists W ∈ Wn such that Kn s ⊆ W, V n i,s ⊈ W (1 ≤ i ≤ ms)} is an element of ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, F) is selSSM;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies SFM(ΠF (∆), ΠF (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (1) ⇒ (2): Let {Jn : n ∈ N} be a sequence of πF (∆)-networks of X and let {Kn : n ∈ N} be a sequence of kF (∆)-covers of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For each n ∈ ω, we denote Jn = {(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) : s ∈ Sn}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Applying Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4, we obtain that, for each n ∈ ω, the collection Un = {(�ms i=1(V n i,s)−) ∩ ((Kn s )c)+ : s ∈ Sn} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Furthermore, by Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3, the collection Dn = Kc n is a dense subset of (∆, F), for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, applying (1) to the sequence {Un : n ∈ N} and the sequence {Dn : n ∈ N}, there is a sequence {Fn : n ∈ N} of finite sets such that, for each n ∈ N, Fn ⊆ Dn and the collection {St(Fn, Un) : n ∈ N} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We put, for each n ∈ N, Wn = {F c : F ∈ Fn}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, for each n ∈ N, Wn is a finite subset of Kn.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us show that the collection J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists W ∈ Wn such that Kn s ⊆ W, V n i,s ⊈ W (1 ≤ i ≤ ms)} is a πF (∆)- network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let U ∈ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then U c ∈ ∆ and therefore, there exists n0 ∈ ω such that U c ∈ St(Fn0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, there is (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+ ∈ Un0 such that � (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+� � Fn0 ̸= ∅ and U c ∈ (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We fix F0 ∈ � (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+� � Fn0 and define W0 = F c 0 .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus W0 ∈ Wn0 and, observe that Kn0 s0 ⊆ W0 and V n0 i,s0 ⊈ W0(1 ≤ i ≤ ms0) are obtained from the fact that F0 ∈ (�ms0 i=1 (V n0 i,s0)−)∩((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It follows that the corresponding element (Kn0 s0 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) belongs to J .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In addition, from the fact that U c ∈ �ms0 i=1 (V n0 i,s0)−, we can fix xi ∈ U c ∩ V n0 i,s0 (i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}) and let M = {xi : i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', ms0}}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' So, M is a finite subset of X that satisfies M ∩ V n0 i,s0 ̸= ∅ (1 ≤ i ≤ ms0) and M ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' On the other hand, Kn0 s0 ⊆ U follows from U c ∈ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We conclude that J ∈ ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) ⇒ (1): Let {Un : n ∈ N} be a sequence of open covers of (∆, F) and let {Dn : n ∈ N} be a sequence of dense subsets of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Without loss of generality, suppose that the elements of each open cover Un are ba- sic open subsets in CL(X).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Therefore, let us denote, for each n ∈ N, Un = {(�ms i=1(V n i,s)−) ∩ ((Kn s )c)+ : s ∈ Sn}, where each V n i,s is an open subset of X and Kn s is a compact subset of X (for every n ∈ N, s ∈ Sn and i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms}).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For each n ∈ N, let Jn = {(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) : s ∈ Sn}.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, by Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4, we get that each Jn is a πF (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Moreover, if we define, for each n ∈ N, Kn = Dc n, then each Kn is a kF (∆)-cover of X by Lemma 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Applying (2) to the sequence of πF (∆)-networks {Jn : n ∈ N} and the sequence of kF (∆)-covers {Kn : n ∈ N} we obtain a sequence {Wn : n ∈ N} 12 with Wn a finite subset of Kn, for each n ∈ N, and so that the collection J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists W ∈ Wn such that Kn s ⊆ W, V n i,s ⊈ W (1 ≤ i ≤ ms)} is a πF (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define, for each n ∈ ω, Fn = Wc n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then Fn is a finite subset of Dn, for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let us show that the collection {St(Fn, Un) : n ∈ N} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let A ∈ ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Using the fact that J is a πF (∆)-network of X with the set Ac ∈ ∆c, we obtain an element (Kn0 s0 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) of J (for some n0 ∈ N and some s0 ∈ Sn0) and a finite set F ⊆ X such that for every i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}, F ∩ V n0 i,s0 ̸= ∅, Kn0 s0 ⊆ Ac and F ∩ Ac = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It easily follows from these last conditions that A belongs to (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' On the other hand, since (Kn0 s0 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0 ,s0) ∈ J , there exists W0 ∈ Wn0 such that Kn0 s0 ⊆ W0 and for each i ∈ {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , ms0}, V n0 i,s0 ⊈ W0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' These conditions imply that the element F0 = W c 0 belongs to (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Notice that F0 ∈ Fn0 and (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+ ∈ Un0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It means that A ∈ St(Fn0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Therefore, {St(Fn, Un) : n ∈ N} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' As a consequence of previous theorem, we can characterize the selectively strongly star-Menger property on different hyperspaces considered with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), F) is selSSM if and only if X satisfies SFM(ΠF , ΠF );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), F) is selSSM if and only if X satisfies SFM(ΠF (K(X)), ΠF (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), F) is selSSM if and only if X satisfies SFM(ΠF (CS(X)), ΠF (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), F) is selSSM if and only if X satisfies SFM(ΠF (F(X)), ΠF (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For the Rothberger case we define the following selection principle to get, in the same way, characterizations of the selectively strongly star-Rothberger property on hyperspaces with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: SFR(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ ω} ⊆ ΠF (∆) and each sequence {Kn : n ∈ ω} ⊆ KF (∆), there is a sequence {Kn : n ∈ ω} with Kn ∈ Kn, n ∈ ω, such that J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : Kn s ⊆ Kn, V n i,s ⊈ Kn (1 ≤ i ≤ ms)} is an element of ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, F) is selSSR;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies SFR(ΠF (∆), ΠF (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 13 Proof.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (1) ⇒ (2): By using Lemmas 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4 and same idea as in the first part of (1) ⇒ (2) in the proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='6, we can obtain a sequence {Dn : n ∈ N} of points in (∆, F) such that, for each n ∈ N, Dn ∈ Dn and the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' For each n ∈ N, we define Kn = Dc n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, Kn ∈ Kn, for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Considering the sequence {Kn : n ∈ N}, let us show that the collection J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : Kn s ⊆ Kn, V n i,s ⊈ Kn (1 ≤ i ≤ ms)} is a πF (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let U ∈ ∆c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, for the element U c ∈ ∆, there exists n0 ∈ ω such that U c ∈ St(Dn0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, there is (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+ ∈ Un0 such that U c, Dn0 ∈ (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' As Kn0 = Dc n0 and Dn0 ∈ (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+, it easily follows that Kn0 s0 ⊆ Kn0 and V n0 i,s0 ⊈ Kn0(1 ≤ i ≤ ms0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' This implies that the corresponding (Kn0 s0 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) is an element of J .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Furthermore, since U c belongs to (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+, we get that Kn0 s0 ⊆ U and also, we can define a finite subset F of X such that F ∩ V n0 i,s0 ̸= ∅ and F ∩ U = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We conclude that J ∈ ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) ⇒ (1): Following same arguments and employing Lemmas 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='3 and 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='4 as in the first part of (2) ⇒ (1) in the proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='6, we obtain a sequence {Kn : n ∈ N} such that, for each n ∈ N, Kn ∈ Kn, and the collection J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : Kn s ⊆ Kn, V n i,s ⊈ Kn (1 ≤ i ≤ ms)} is a πF (∆)-network of X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define, for each n ∈ N, Dn = Kc n.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then, {Dn : n ∈ N} is a sequence of points in (∆, F) with Dn ∈ Dn, for each n ∈ N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We claim that the collection {St(Dn, Un) : n ∈ N} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Indeed, let A ∈ ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Since J is a πF (∆)-network of X and Ac ∈ ∆c, there exists (Kn0 s0 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) ∈ J (for some n0 ∈ N and some s0 ∈ Sn0) and a finite set F ⊆ X such that F ∩ V n0 i,s0 ̸= ∅ (1 ≤ i ≤ ms0), Kn0 s0 ⊆ Ac and F ∩ Ac = ∅.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The fact (Kn0 s0 ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n0 1,s0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n0 ms0,s0) ∈ J means that Kn0 s0 ⊆ Kn0 and V n0 i,s0 ⊈ Kn0 ( 1 ≤ i ≤ ms0) and hence, Dn0 ∈ (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Moreover, the conditions F ∩ V n0 i,s0 ̸= ∅ (1 ≤ i ≤ ms0), Kn0 s0 ⊆ Ac and F ∩ Ac = ∅ mean that A ∈ (�ms0 i=1 (V n0 i,s0)−) ∩ ((Kn0 s0 )c)+.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Thus, A ∈ St(Dn0, Un0).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In conclusion, {St(Dn, Un) : n ∈ N} is an open cover of (∆, F).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Now, we can characterize the selectively strongly star-Rothberger property on several hyperspaces with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), F) is selSSR if and only if X satisfies SFR(ΠF , ΠF );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), F) is selSSR if and only if X satisfies SFR(ΠF (K(X)), ΠF (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), F) is selSSR if and only if X satisfies SFR(ΠF (CS(X)), ΠF (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), F) is selSSR if and only if X satisfies SFR(ΠF (F(X)), ΠF (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 14 Now we characterize the absolute versions of the Menger-type and Rothberger- type star selection principles with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following principle al- lows us to characterize the absolutely strongly star-Menger property on several hyperspaces with the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: AFM(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ ω} ⊆ ΠF (∆) and each K ∈ KF (∆), there is a sequence {Wn : n ∈ ω} ⊆ [K]<ω such that J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : there exists W ∈ Wn such that Kn s ⊆ W, V n i,s ⊈ W (1 ≤ i ≤ ms)} is an element of ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Adapting the proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='6, one can easily obtain the following result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, F) is aSSM;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies AFM(ΠF (∆), ΠF (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We obtain the following particular cases by taking different choices of our family ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Corollary 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='13.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), F) is aSSM if and only if X satisfies AFM(ΠF , ΠF );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), F) is aSSM if and only if X satisfies AFM(ΠF (K(X)), ΠF (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), F) is aSSM if and only if X satisfies AFM(ΠF (CS(X)), ΠF (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), F) is aSSM if and only if X satisfies AFM(ΠF (F(X)), ΠF (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following principle will give us characterizations of the absolutely strongly star-Rothberger property on hyperspaces considering the Fell topology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Definition 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='14.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' We define: AFR(ΠF (∆), ΠF (∆)): For each sequence {Jn : n ∈ N} ⊆ ΠF (∆) and each K ∈ KF (∆), there is a sequence {Kn : n ∈ N} ⊆ K, such that J = � n∈N{(Kn s ;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' V n 1,s, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' , V n ms,s) ∈ Jn : Kn s ⊆ Kn, V n i,s ⊈ Kn (1 ≤ i ≤ ms)} is an element of ΠF (∆).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Similarly to the proof of Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='9, one can prove the following result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Theorem 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Given a topological space X, the following conditions are equiv- alent: (1) (∆, F) is aSSR;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (2) X satisfies AFR(ΠF (∆), ΠF (∆)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' The following particular cases are obtained by choosing different families ∆.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 15 Corollary 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Let X be a topological space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Then: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CL(X), F) is aSSR if and only if X satisfies AFR(ΠF , ΠF );' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (K(X), F) is aSSR if and only if X satisfies AFR(ΠF (K(X)), ΠF (K(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (CS(X), F) is aSSR if and only if X satisfies AFR(ΠF (CS(X)), ΠF (CS(X)));' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' (F(X), F) is aSSR if and only if X satisfies AFR(ΠF (F(X)), ΠF (F(X))).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Concluding remark.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' It worth remarking that, by doing some slightly modifications to definitions given in Sections 2 and 3, it is possible to easily obtain hit-and-miss generalizations that involve both the Vietoris topology and the Fell topology (similar to how it was done in [6] and [7]).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' However, for con- venience of the reader, the author decided to establish simpler notation results depending of which topology is being considered.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' References [1] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Bonanzinga, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Cammaroto, Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Matveev, On weaker forms of Menger, Rothberger and Hurewicz properties, Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Vesnik 61 (2009) 13-23.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [2] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Casas-de la Rosa, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Garcia-Balan, Variations of star selection prin- ciples on small spaces, Filomat (2022), 36:14, 4903-4917.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [3] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Casas-de la Rosa, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Mart´ınez-Ruiz, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Ram´ırez-P´aramo, Star versions of the Menger property on hyperspaces, Houston J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=', to appear.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [4] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Casas-de la Rosa, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Mart´ınez-Ruiz, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Ram´ırez-P´aramo, Star versions of the Rothberger property on hyperspaces, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 283 (2020), Art.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' ID 107396, 12 pages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [5] A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Caserta, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Di Maio, Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, Versions of properties (a) and (pp), Topology Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 158 (2011) 1360–1368.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [6] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Cruz-Castillo, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Ram´ırez-P´aramo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Tenorio Menger and Menger- type star selection principles for hit-and-miss topology, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 290 (2021), Art.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' ID 107574, 12 pages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [7] R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Cruz-Castillo, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Ram´ırez-P´aramo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Tenorio Star and strong star- type versions of Rothberger and Menger principles for hit-and-miss topology, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 300 (2021), Art.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' ID 107758, 11 pages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [8] J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' D´ıaz-Reyes, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Ram´ırez-P´aramo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Tenorio Rothberger and Rothberger-type star selection principles on hyperspaces, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 287 (2021), Art.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' ID 107448, 9 pages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [9] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Di Maio, Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Meccariello, Selection principles and hy- perspace topologies, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 153 (2005) 912-923.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 16 [10] G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Di Maio, Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, Some covering properties of hyperspaces, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 155 (2008) 1959-1969.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [11] F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hern´andez-Hern´andez, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hruˇs´ak, Topology of Mr´owka-Isbell spaces.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' In Pseudocompact Topological Spaces, Eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hruˇs´ak, Tamariz, Tkachenko.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Springer International Publishing AG, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [12] Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, On Star Selection Principles Theory, Axioms, (2023) 12, 93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [13] Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, Star-Menger and related spaces, Publ.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Debrecen 55 (1999) 421-431.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [14] Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, Star selection principles: A survey, Khayyam J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1 (2015) No.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 1 82-106.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [15] Lj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Koˇcinac, The Reznichenko property and the Pytkeev property in hyperspaces, Acta Math.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Hungar.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 107, 3 (2005), 231–239.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [16] Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Li, Selection principles of the Fell topology and the Vietoris topology, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 212 (2016) 90-104.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' [17] M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Scheepers, Combinatorics of open covers I: Ramsey theory, Topol.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Appl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' 69 (1996) 31-62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Departamento de Matem´aticas, Facultad de Ciencias, UNAM, Circuito Exterior S/N, Ciudad Universitaria, CP 04510, Ciudad de M´exico, M´exico.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Email address: J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content=' Casas-de la Rosa: olimpico.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='25@hotmail.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
+page_content='com 17' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/qNFJT4oBgHgl3EQfaCwn/content/2301.11534v1.pdf'}
diff --git a/qdFIT4oBgHgl3EQfxCs8/content/tmp_files/2301.11354v1.pdf.txt b/qdFIT4oBgHgl3EQfxCs8/content/tmp_files/2301.11354v1.pdf.txt
new file mode 100644
index 0000000000000000000000000000000000000000..94f358040a876ddbe0e23fcb1a5e74327299d902
--- /dev/null
+++ b/qdFIT4oBgHgl3EQfxCs8/content/tmp_files/2301.11354v1.pdf.txt
@@ -0,0 +1,1198 @@
+Permutation-based Hypothesis Testing for Neural Networks
+Francesca Mandel 1 Ian Barnett 1
+Abstract
+Neural networks are powerful predictive models,
+but they provide little insight into the nature of
+relationships between predictors and outcomes.
+Although numerous methods have been proposed
+to quantify the relative contributions of input fea-
+tures, statistical inference and hypothesis testing
+of feature associations remain largely unexplored.
+We propose a permutation-based approach to test-
+ing that uses the partial derivatives of the network
+output with respect to specific inputs to assess
+both the significance of input features and whether
+significant features are linearly associated with the
+network output. These tests, which can be flexibly
+applied to a variety of network architectures, en-
+hance the explanatory power of neural networks,
+and combined with powerful predictive capability,
+extend the applicability of these models.
+1. Introduction
+While neural networks are well known for their predictive ca-
+pability, compared to traditional regression approaches, they
+generally provide little explanatory insight into how they
+make their predictions. While the mathematics of each layer-
+to-layer transformation are relatively simple, how and why a
+network combines information from the inputs to predict the
+outputs becomes more difficult to understand as the network
+architecture grows in complexity. This issue of interpretabil-
+ity of neural networks has been addressed extensively in
+the literature (Gilpin et al., 2018; Zhang et al., 2021). De-
+spite the challenges, there are many settings in which is
+it desirable or necessary to interpret neural networks. In
+applications such as credit, employment, and criminal jus-
+tice, understanding how predictions are made is extremely
+useful for evaluating whether the algorithms are fair and
+non-discriminatory (Bostrom & Yudkowsky, 2014; Hardt
+et al., 2016). Recent laws mandating the “right to explana-
+tion,” a right to information about individual decisions made
+1Department of Biostatistics, Epidemiology, and Informat-
+ics, Perelman School of Medicine, University of Pennsylvania,
+Philadelphia, USA. Correspondence to: Francesca Mandel